issue_owner_repo listlengths 2 2 | issue_body stringlengths 0 262k ⌀ | issue_title stringlengths 1 1.02k | issue_comments_url stringlengths 53 116 | issue_comments_count int64 0 2.49k | issue_created_at stringdate 1999-03-17 02:06:42 2025-06-23 11:41:49 | issue_updated_at stringdate 2000-02-10 06:43:57 2025-06-23 11:43:00 | issue_html_url stringlengths 34 97 | issue_github_id int64 132 3.17B | issue_number int64 1 215k |
|---|---|---|---|---|---|---|---|---|---|
[
"libjxl",
"libjxl"
] | Hello, I hope you doing well.
It seems like samsung galaxy series of the smartphone using
wrong algorithm to their color space compression
(About the description : https://community.adobe.com/t5/camera-raw-discussions/samsung-s24-ultra-phones-don-t-do-raw-anymore-it-s-not-adobes-fault/m-p/14443437)
So I talked with Mr.Jyrki Alakuijala and he required the excat file to analysis.
so I bring up the google drive with problematic .dng RAW file here, and Hope this help
for the contact with adobe or samsung associate.
Thank you, have a nice day. | About the bug of the adobe photoshop cannot handle samsung .dng RAW Camera file with JPEG XL | https://api.github.com/repos/libjxl/libjxl/issues/3640/comments | 21 | 2024-06-11T21:10:26Z | 2024-06-17T07:21:35Z | https://github.com/libjxl/libjxl/issues/3640 | 2,347,319,340 | 3,640 |
[
"libjxl",
"libjxl"
] | **Is your feature request related to a problem? Please describe.**
A clear and concise description of what the problem is. Ex. I'm always frustrated when [...]
i want to be able to encode jfif files
**Describe the solution you'd like**
A clear and concise description of what you want to happen.
implement jfif encoding
**Describe alternatives you've considered**
A clear and concise description of any alternative solutions or features you've considered.
i have not tried alternatives
**Additional context**
Add any other context or screenshots about the feature request here.
i'm trying to encode [this](https://cyfrowe-cdn.mnw.art.pl/upload/multimedia/46/5e/465e5e3e24a24ccdbe76af4186170bc0.jpg) image, which is mislabeled as a jpg file | jfif support | https://api.github.com/repos/libjxl/libjxl/issues/3638/comments | 4 | 2024-06-11T11:38:40Z | 2024-06-25T12:49:56Z | https://github.com/libjxl/libjxl/issues/3638 | 2,346,195,587 | 3,638 |
[
"libjxl",
"libjxl"
] | **Is your feature request related to a problem? Please describe.**
All symbols are being exported for jxl_cms/libjxl_extras_codec
See:
* https://salsa.debian.org/debian-phototools-team/libjxl/-/blob/debian/experimental/debian/libjxl0.9.symbols.amd64#L108
**Describe the solution you'd like**
Same mechanism as with jxl/jxl_threads. Use jxl.version scripts to hide stl related symbols.
**Describe alternatives you've considered**
Solution in Debian is ugly as it depends on a particular GCC/libstdcxx implementation.
| Simplify ABI on jxl_cms/libjxl_extras_codec | https://api.github.com/repos/libjxl/libjxl/issues/3630/comments | 1 | 2024-06-05T06:08:13Z | 2024-06-27T11:28:58Z | https://github.com/libjxl/libjxl/issues/3630 | 2,334,985,364 | 3,630 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Two tests fails on big-endian archs
**To Reproduce**
Full build log on s390x
* https://buildd.debian.org/status/fetch.php?pkg=jpeg-xl&arch=s390x&ver=0.9.2-2&stamp=1717516578&raw=0
powerpc or ppc64 are affected the same way.
**Expected behavior**
Expected that all tests passes as on little-endian archs
**Screenshots**
* https://buildd.debian.org/status/fetch.php?pkg=jpeg-xl&arch=s390x&ver=0.9.2-2&stamp=1717516578&raw=0
**Environment**
- OS: Debian/sid
- Compiler version: GCC-13
- CPU type: s390x
- cjxl/djxl version string: 0.9.2
| Big Endian: Unit test failures | https://api.github.com/repos/libjxl/libjxl/issues/3629/comments | 12 | 2024-06-05T06:05:17Z | 2025-02-06T16:50:32Z | https://github.com/libjxl/libjxl/issues/3629 | 2,334,980,901 | 3,629 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
ffmpeg fails to configure with libjxl being build statically, caused by libjxl_threads
**To Reproduce**
Build libjxl with these options:
` -DBUILD_SHARED_LIBS=OFF -DCMAKE_INSTALL_PREFIX="${WORKSPACE}" -DCMAKE_INSTALL_LIBDIR=lib -DCMAKE_INSTALL_BINDIR=bin -DCMAKE_INSTALL_INCLUDEDIR=include -DENABLE_SHARED=off -DENABLE_STATIC=ON -DCMAKE_BUILD_TYPE=Release -DJPEGXL_ENABLE_BENCHMARK=OFF -DJPEGXL_ENABLE_DOXYGEN=OFF -DJPEGXL_ENABLE_MANPAGES=OFF -DJPEGXL_ENABLE_JPEGLI_LIBJPEG=OFF -DJPEGXL_ENABLE_JPEGLI=OFF -DJPEGXL_TEST_TOOLS=OFF -DJPEGXL_ENABLE_JNI=OFF -DBUILD_TESTING=OFF .`
Build ffmpeg 7.0 and enable libjxl
**Expected behavior**
libjxl being build statically and pkg-config files having correct linking options
**Screenshots**
From the ffmpeg configure log:
```
pkg-config --exists --print-errors libjxl_threads >= 0.7.0
check_func_headers jxl/thread_parallel_runner.h JxlThreadParallelRunner -I/home/stefan/source/ffmpeg-build3/workspace/include -L/home/stefan/source/ffmpeg-build3/workspace/lib -ljxl_threads -lm
test_ld cc -I/home/stefan/source/ffmpeg-build3/workspace/include -L/home/stefan/source/ffmpeg-build3/workspace/lib -ljxl_threads -lm
test_cc -I/home/stefan/source/ffmpeg-build3/workspace/include -L/home/stefan/source/ffmpeg-build3/workspace/lib
BEGIN /tmp/ffconf.GRMmGUbV/test.c
1 #include <jxl/thread_parallel_runner.h>
2 #include <stdint.h>
3 long check_JxlThreadParallelRunner(void) { return (long) JxlThreadParallelRunner; }
4 int main(void) { int ret = 0;
5 ret |= ((intptr_t)check_JxlThreadParallelRunner) & 0xFFFF;
6 return ret; }
END /tmp/ffconf.GRMmGUbV/test.c
gcc -D_ISOC11_SOURCE -D_FILE_OFFSET_BITS=64 -D_LARGEFILE_SOURCE -D_POSIX_C_SOURCE=200112 -D_XOPEN_SOURCE=600 -DPIC -I/home/stefan/source/ffmpeg-build3/workspace/include -std=c17 -fomit-frame-pointer -fPIC -pthread -I/home/stefan/source/ffmpeg-build3/workspace/include -I/home/stefan/source/ffmpeg-build3/workspace/include -pthread -I/home/stefan/source/ffmpeg-build3/workspace/include -I/usr/include/libdrm -I/home/stefan/source/ffmpeg-build3/workspace/include -I/home/stefan/source/ffmpeg-build3/workspace/include/freetype2 -I/home/stefan/source/ffmpeg-build3/workspace/include/libpng16 -I/home/stefan/source/ffmpeg-build3/workspace/include -I/usr/include/harfbuzz -I/usr/include/glib-2.0 -I/usr/lib/x86_64-linux-gnu/glib-2.0/include -DHWY_STATIC_DEFINE -I/home/stefan/source/ffmpeg-build3/workspace/include -I/home/stefan/source/ffmpeg-build3/workspace/include -L/home/stefan/source/ffmpeg-build3/workspace/lib -c -o /tmp/ffconf.GRMmGUbV/test.o /tmp/ffconf.GRMmGUbV/test.c
gcc -L/home/stefan/source/ffmpeg-build3/workspace/lib -Wl,--as-needed -Wl,-z,noexecstack -I/home/stefan/source/ffmpeg-build3/workspace/include -L/home/stefan/source/ffmpeg-build3/workspace/lib -o /tmp/ffconf.GRMmGUbV/test /tmp/ffconf.GRMmGUbV/test.o -ljxl_threads -lm -ldl -lpthread -lm -lz
/usr/bin/ld: /home/stefan/source/ffmpeg-build3/workspace/lib/libjxl_threads.a(thread_parallel_runner.cc.o): in function `JxlThreadParallelRunnerDefaultNumWorkerThreads':
thread_parallel_runner.cc:(.text+0x179): undefined reference to `std::thread::hardware_concurrency()'
/usr/bin/ld: /home/stefan/source/ffmpeg-build3/workspace/lib/libjxl_threads.a(thread_parallel_runner_internal.cc.o): in function `jpegxl::ThreadParallelRunner::Runner(void*, void*, int (*)(void*, unsigned long), void (*)(void*, unsigned int, unsigned long), unsigned int, unsigned int)':
thread_parallel_runner_internal.cc:(.text+0x11e): undefined reference to `std::condition_variable::notify_all()'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x16f): undefined reference to `std::condition_variable::wait(std::unique_lock<std::mutex>&)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x1fa): undefined reference to `std::__throw_system_error(int)'
/usr/bin/ld: /home/stefan/source/ffmpeg-build3/workspace/lib/libjxl_threads.a(thread_parallel_runner_internal.cc.o): in function `jpegxl::ThreadParallelRunner::ThreadFunc(jpegxl::ThreadParallelRunner*, int)':
thread_parallel_runner_internal.cc:(.text+0x34c): undefined reference to `std::condition_variable::wait(std::unique_lock<std::mutex>&)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x425): undefined reference to `std::condition_variable::notify_one()'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x443): undefined reference to `std::__throw_system_error(int)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x44d): undefined reference to `std::__throw_system_error(int)'
/usr/bin/ld: /home/stefan/source/ffmpeg-build3/workspace/lib/libjxl_threads.a(thread_parallel_runner_internal.cc.o): in function `jpegxl::ThreadParallelRunner::ThreadParallelRunner(int)':
thread_parallel_runner_internal.cc:(.text+0x4d9): undefined reference to `std::condition_variable::condition_variable()'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x4f0): undefined reference to `std::condition_variable::condition_variable()'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x541): undefined reference to `operator new(unsigned long)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x588): undefined reference to `operator delete(void*, unsigned long)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x5de): undefined reference to `operator new(unsigned long)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x60f): undefined reference to `std::thread::_M_start_thread(std::unique_ptr<std::thread::_State, std::default_delete<std::thread::_State> >, void (*)())'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x6b0): undefined reference to `operator new(unsigned long)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x6e2): undefined reference to `std::thread::_M_start_thread(std::unique_ptr<std::thread::_State, std::default_delete<std::thread::_State> >, void (*)())'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x74e): undefined reference to `operator delete(void*, unsigned long)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x7e7): undefined reference to `std::condition_variable::wait(std::unique_lock<std::mutex>&)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x886): undefined reference to `operator new(unsigned long)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x8b9): undefined reference to `std::__throw_system_error(int)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x8ca): undefined reference to `std::__throw_length_error(char const*)'
/usr/bin/ld: /home/stefan/source/ffmpeg-build3/workspace/lib/libjxl_threads.a(thread_parallel_runner_internal.cc.o): in function `jpegxl::ThreadParallelRunner::~ThreadParallelRunner()':
thread_parallel_runner_internal.cc:(.text+0x910): undefined reference to `std::thread::join()'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x91d): undefined reference to `std::condition_variable::~condition_variable()'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x927): undefined reference to `std::condition_variable::~condition_variable()'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x9b0): undefined reference to `std::condition_variable::notify_all()'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x9ba): undefined reference to `std::terminate()'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x9c1): undefined reference to `std::__throw_system_error(int)'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text+0x965): undefined reference to `operator delete(void*, unsigned long)'
/usr/bin/ld: /home/stefan/source/ffmpeg-build3/workspace/lib/libjxl_threads.a(thread_parallel_runner_internal.cc.o): in function `std::thread::_State_impl<std::thread::_Invoker<std::tuple<void (*)(jpegxl::ThreadParallelRunner*, int), jpegxl::ThreadParallelRunner*, unsigned int> > >::~_State_impl()':
thread_parallel_runner_internal.cc:(.text._ZNSt6thread11_State_implINS_8_InvokerISt5tupleIJPFvPN6jpegxl20ThreadParallelRunnerEiES5_jEEEEED2Ev[_ZNSt6thread11_State_implINS_8_InvokerISt5tupleIJPFvPN6jpegxl20ThreadParallelRunnerEiES5_jEEEEED5Ev]+0x13): undefined reference to `std::thread::_State::~_State()'
/usr/bin/ld: /home/stefan/source/ffmpeg-build3/workspace/lib/libjxl_threads.a(thread_parallel_runner_internal.cc.o): in function `std::thread::_State_impl<std::thread::_Invoker<std::tuple<void (*)(jpegxl::ThreadParallelRunner*, int), jpegxl::ThreadParallelRunner*, unsigned int> > >::~_State_impl()':
thread_parallel_runner_internal.cc:(.text._ZNSt6thread11_State_implINS_8_InvokerISt5tupleIJPFvPN6jpegxl20ThreadParallelRunnerEiES5_jEEEEED0Ev[_ZNSt6thread11_State_implINS_8_InvokerISt5tupleIJPFvPN6jpegxl20ThreadParallelRunnerEiES5_jEEEEED5Ev]+0x20): undefined reference to `std::thread::_State::~_State()'
/usr/bin/ld: thread_parallel_runner_internal.cc:(.text._ZNSt6thread11_State_implINS_8_InvokerISt5tupleIJPFvPN6jpegxl20ThreadParallelRunnerEiES5_jEEEEED0Ev[_ZNSt6thread11_State_implINS_8_InvokerISt5tupleIJPFvPN6jpegxl20ThreadParallelRunnerEiES5_jEEEEED5Ev]+0x32): undefined reference to `operator delete(void*, unsigned long)'
collect2: error: ld returned 1 exit status
ERROR: libjxl_threads >= 0.7.0 not found using pkg-config
```
**Environment**
- OS: Linux Mint / Ubuntu 22.04 LTS
- Compiler version: gcc 11.4.0
- CPU type: x86_64
- cjxl/djxl version string: 0.10.2
**Additional context**
This is for creating a self-contained ffmpeg binary that can be copied to another machine, so no shared libraries
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| libjxl_threads causes linking errors on static builds | https://api.github.com/repos/libjxl/libjxl/issues/3627/comments | 3 | 2024-06-03T18:34:09Z | 2024-06-06T13:24:17Z | https://github.com/libjxl/libjxl/issues/3627 | 2,331,796,806 | 3,627 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
CMake cannot find GTest after executing deps.sh to get dependencies
**To Reproduce**
Download source zip file, execute deps.sh and run cmake
**Expected behavior**
deps.sh installs all dependencies, cmake generating Makefiles
**Screenshots**
Exact output of cmake:
```CMake Error at /home/stefan/source/ffmpeg-build3/workspace/share/cmake-3.27/Modules/FindPackageHandleStandardArgs.cmake:230 (message):
Could NOT find GTest (missing: GTEST_LIBRARY GTEST_INCLUDE_DIR
GTEST_MAIN_LIBRARY)
Call Stack (most recent call first):
/home/stefan/source/ffmpeg-build3/workspace/share/cmake-3.27/Modules/FindPackageHandleStandardArgs.cmake:600 (_FPHSA_FAILURE_MESSAGE)
/home/stefan/source/ffmpeg-build3/workspace/share/cmake-3.27/Modules/FindGTest.cmake:270 (FIND_PACKAGE_HANDLE_STANDARD_ARGS)
third_party/testing.cmake:57 (find_package)
CMakeLists.txt:358 (include)
-- Configuring incomplete, errors occurred!
Failed to Execute cmake -DBUILD_SHARED_LIBS=OFF -DCMAKE_INSTALL_PREFIX=/home/stefan/source/ffmpeg-build3/workspace -DENABLE_SHARED=off -DCMAKE_BUILD_TYPE=Release -DJPEGXL_ENABLE_BENCHMARK=OFF -DJPEGXL_ENABLE_DOXYGEN=OFF -DJPEGXL_ENABLE_MANPAGES=OFF -DJPEGXL_ENABLE_JPEGLI_LIBJPEG=OFF -DJPEGXL_ENABLE_JPEGLI=OFF -DJPEGXL_TEST_TOOLS=OFF -DJPEGXL_ENABLE_JNI=OFF .
```
**Environment**
- OS: Linux Mint / Ubuntu 22.04 LTS
- Compiler version: gcc 11.4.0
- CPU type: x86_64
- cjxl/djxl version string: 0.10.2
**Additional context**
I would like to include libjxl in an automated script to build a static ffmpeg. Therefore git should be avoided, I have no problem compiling libjxl with git submodules, I want a solution without git.
I am not sure for what GTest is needed and if there is any flag to disable the need for it.
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| deps.sh doesn't download GTest | https://api.github.com/repos/libjxl/libjxl/issues/3621/comments | 2 | 2024-06-01T14:49:20Z | 2024-06-04T05:24:11Z | https://github.com/libjxl/libjxl/issues/3621 | 2,329,182,951 | 3,621 |
[
"libjxl",
"libjxl"
] | _Initially, this was a [discussion](https://discord.com/channels/794206087879852103/794206170445119489/1244930287184511037) in the JXL Discord server; this issue is a shortened version of that._
**Summary**
At some point there was a change to jpegxl that caused a regression in how vibrant colors are represented. The difference in saturation is most noticeable in photographic images compressed at higher distances.
**Testing Methodology**
Factory Image: [Abandoned_Factory_Detroit_Downsampled](https://raw.githubusercontent.com/Galaxy4594/jxl-comp/main/im/Abandoned_Packard_Automobile_Factory_Detroit_Downsampled.png)
The factory image is from this [comparison](https://afontenot.github.io/image-formats-comparison/#abandoned-factory&AV1=m&JPEGXL=m) where the color desaturation is the most obvious. I tested it myself with libjxl 0.10 and the git master (45e688c) at a distance of 2.7 and got similar results. The reference avifs are the same bbp.
**Comparisons**
Zooming in on the yellow leaves shows a very clear destruction of color.
The factory image is definitely a problem case. So here is a more realistic scenario, I used same photographic image from #3530 and tested it at a more reasonable distance of 1.5.
The desaturation is most noticeable on the hair, particularly on the right side in shadow, where it looks almost grey instead of brown. According to preliminary testing in the jxl discord, the colors became [notably worse](https://discord.com/channels/794206087879852103/794206170445119489/1245282113864073266) after 0.6.1. | Excessive desaturation of vibrant colors | https://api.github.com/repos/libjxl/libjxl/issues/3616/comments | 4 | 2024-05-30T01:00:31Z | 2025-04-27T03:17:21Z | https://github.com/libjxl/libjxl/issues/3616 | 2,324,460,962 | 3,616 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Can't decode the following image with jpegli: [334344911-dff6f477-5899-482b-be34-fc4d91dfbaa8.zip](https://github.com/libjxl/libjxl/files/15467346/334344911-dff6f477-5899-482b-be34-fc4d91dfbaa8.zip)
**To Reproduce**
Try to decode it with jpegli
**Expected behavior**
Successful decoding
**Environment**
- OS: Linux
- Compiler version: GCC 12
- CPU type: x86_64
- cjxl/djxl version string: I don't have cjxl/djxl... But I used v0.10.2 tag to build jpegli.
**Additional context**
mozjpeg decodes it just fine
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| Unexpected marker 0xd3 | https://api.github.com/repos/libjxl/libjxl/issues/3609/comments | 0 | 2024-05-28T09:54:36Z | 2024-06-09T16:57:57Z | https://github.com/libjxl/libjxl/issues/3609 | 2,320,629,933 | 3,609 |
[
"libjxl",
"libjxl"
] | The default for Jpegli is to output YUV as 4:4:4 regardless of quality level. However, according to exiftool, XYB outputs as 4:2:0 (blue-channel subsampled) as the default setting.
This can be confusing because, based on the YUV mode, one might assume that XYB also has full chroma resolution. When I noticed a slight color shift while compressing in XYB, I initially thought it was an encoder quirk. However, it turns out that XYB is indeed chroma subsampled.
For consistency, Jpegli should have both modes default to 4:4:4 chroma. A secondary benefit of this change would be that XYB jpegs can be transcoded to Jpegxl.
Another issue is present with XYB: the colors appear significantly different between the default 4:2:0 and the manually defined `chroma_subsampling=420`. My guess is that it’s using YUV subsampling rather than the XYB blue-channel subsampling.
The behavior is pretty strange and it doesn't really make much sense.
 | Inconsistent default behavior for Jpegli | https://api.github.com/repos/libjxl/libjxl/issues/3605/comments | 5 | 2024-05-26T22:11:09Z | 2025-01-26T03:42:34Z | https://github.com/libjxl/libjxl/issues/3605 | 2,317,954,869 | 3,605 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
A clear and concise description of what the bug is.
I've got a JPEG file that cjxl fails to losslessly transcode. I can transcode it from pixels however.
**To Reproduce**
Steps to reproduce the behaviour:
Try and losslessly transcode this JPEG: [unconvertible.zip](https://github.com/libjxl/libjxl/files/15440548/unconvertible.zip)
```shell
cjxl IMG-20240116-WA0000.jpg --disable_output --lossless_jpeg=1
```
**Expected behaviour**
A clear and concise description of what you expected to happen.
As this is not a CMYK JPEG, I'd have expected it to convert successfully.
```
❯ cjxl IMG-20240116-WA0000.jpg output.jxl --lossless_jpeg=1
JPEG XL encoder v0.10.2 0.10.2~alpha20240524121928-0+git45e688c [SSE4,SSE2]
Encoding [JPEG, lossless transcode, effort: 7]
Error while decoding the JPEG image. It may be corrupt (e.g. truncated) or of an unsupported type (e.g. CMYK).
EncodeImageJXL() failed.
```
```
❯ file IMG-20240116-WA0000.jpg
IMG-20240116-WA0000.jpg: JPEG image data, JFIF standard 1.01, aspect ratio, density 1x1, segment length 16, progressive, precision 8, 1042x1798, components 3
```
```
❯ identify -verbose IMG-20240116-WA0000.jpg
Image:
Filename: IMG-20240116-WA0000.jpg
Permissions: rw-r--r--
Format: JPEG (Joint Photographic Experts Group JFIF format)
Mime type: image/jpeg
Class: DirectClass
Geometry: 1042x1798+0+0
Units: Undefined
Colorspace: sRGB
Type: TrueColor
Base type: Undefined
Endianness: Undefined
Depth: 8-bit
Channels: 3.0
Channel depth:
Red: 8-bit
Green: 8-bit
Blue: 8-bit
Channel statistics:
Pixels: 1873516
Red:
min: 0 (0)
max: 255 (1)
mean: 88.4301 (0.346785)
median: 40 (0.156863)
standard deviation: 90.2069 (0.353753)
kurtosis: -0.727428
skewness: 1.06247
entropy: 0.821565
Green:
min: 0 (0)
max: 233 (0.913725)
mean: 90.8236 (0.356171)
median: 61 (0.239216)
standard deviation: 60.0465 (0.235477)
kurtosis: -0.659944
skewness: 1.04197
entropy: 0.789266
Blue:
min: 0 (0)
max: 234 (0.917647)
mean: 104.6 (0.410197)
median: 92 (0.360784)
standard deviation: 44.154 (0.173153)
kurtosis: -0.564884
skewness: 0.506027
entropy: 0.845324
Image statistics:
Overall:
min: 0 (0)
max: 255 (1)
mean: 94.618 (0.371051)
median: 64.3333 (0.252288)
standard deviation: 64.8025 (0.254127)
kurtosis: -0.650752
skewness: 0.870158
entropy: 0.818718
Rendering intent: Perceptual
Gamma: 0.454545
Chromaticity:
red primary: (0.64,0.33,0.03)
green primary: (0.3,0.6,0.1)
blue primary: (0.15,0.06,0.79)
white point: (0.3127,0.329,0.3583)
Matte color: grey74
Background color: white
Border color: srgb(223,223,223)
Transparent color: black
Interlace: JPEG
Intensity: Undefined
Compose: Over
Page geometry: 1042x1798+0+0
Dispose: Undefined
Iterations: 0
Compression: JPEG
Quality: 50
Orientation: Undefined
Properties:
date:create: 2024-05-25T15:03:33+00:00
date:modify: 2024-05-24T22:07:46+00:00
date:timestamp: 2024-05-25T15:04:11+00:00
jpeg:colorspace: 2
jpeg:sampling-factor: 2x2,1x1,1x1
signature: 7386d8fe3a5f471c6cafc801926b4acb425b22f941fe54fb6c5b42a00b2750eb
Artifacts:
verbose: true
Tainted: False
Filesize: 107610B
Number pixels: 1.87352M
Pixel cache type: Memory
Pixels per second: 29.5603MP
Time-to-live: 0:0:0:0 2024-05-25T15:04:11Z
User time: 0.070u
Elapsed time: 0:01.063
Version: ImageMagick 7.1.1-26 Q16-HDRI x86_64 21914 https://imagemagick.org
```
**Environment**
- OS: [e.g. Windows] Fedora 40 / Debian 11
- Compiler version: [e.g. clang 11.0.1] Fedora package / Luca's Debian repo.
- CPU type: [e.g. x86_64] x86_64
- cjxl/djxl version string: [e.g. cjxl [v0.3.7 | SIMD supported: SSE4,Scalar]] cjxl v0.8.2 [AVX2,SSE4,SSSE3,SSE2] and cjxl v0.10.2 0.10.2~alpha20240524121928-0+git45e688c [SSE4,SSE2]
**Additional context**
Add any other context about the problem here.
This image has passed through WhatsApp, though other images that have done the same transcode fine. | YCbCr JPEG cannot be losslessly transcoded | https://api.github.com/repos/libjxl/libjxl/issues/3604/comments | 2 | 2024-05-24T22:31:16Z | 2025-04-27T01:54:24Z | https://github.com/libjxl/libjxl/issues/3604 | 2,316,346,400 | 3,604 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
I am cross compiling for Intel Mac from my M1 mac and I am getting this error while linking. I am using the vcpkg port, `libjxl:x64-osx`.
```
Undefined symbols for architecture x86_64:
"___cpu_model", referenced from:
_JxlFastLosslessPrepareFrame in libjxl.a[21](enc_fast_lossless.cc.o)
_JxlFastLosslessProcessFrame in libjxl.a[21](enc_fast_lossless.cc.o)
ld: symbol(s) not found for architecture x86_64
```
**To Reproduce**
Use the vcpkg port, `libjxl:x64-osx` to cross compile from M1 to intel mac.
**Expected behavior**
It should link just fine.
**Environment**
- OS: MacOS 14.4.1
- Compiler version: clang 15.0.0
- HOST CPU type: aarch64
- TARGET CPU type: x86_64
**Additional Context**
- I am on the latest version of vcpkg, on the master branch
- I got this to link by disabling AVX2 support but that's not ideal.
- If I use the cmake lists `find_package` it also works, as its automatically pulling all the correct flags and stuff for the target, I am probably missing a flag or two.
| Undefined symbols for x86_64: `___cpu_model` | https://api.github.com/repos/libjxl/libjxl/issues/3599/comments | 4 | 2024-05-22T10:50:39Z | 2024-10-22T12:37:59Z | https://github.com/libjxl/libjxl/issues/3599 | 2,310,210,588 | 3,599 |
[
"libjxl",
"libjxl"
] | Currently libjpeg seems to used to decode input jpeg file, but jpegli in this project works perfect to be a drop-in replacement of `libjpeg.a` . In a word, what I do is this:
```sh
cp build/lib/libjpegli-static.a libjpeg.a
ar rcs libjpeg.a build/lib/CMakeFiles/jpegli-libjpeg-obj.dir/jpegli/libjpeg_wrapper.cc.o
cmake -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTING=OFF -DJPEGXL_STATIC=ON -DJPEGXL_ENABLE_BENCHMARK=OFF -DJPEGXL_ENABLE_DOXYGEN=OFF -DJPEGXL_ENABLE_MANPAGES=OFF -DJPEGXL_ENABLE_JNI=OFF -DJPEGXL_ENABLE_SJPEG=OFF -DJPEGXL_ENABLE_OPENEXR=OFF -DJPEGXL_ENABLE_AVX512=OFF \
-DJPEG_LIBRARY=$(pwd)/libjpeg.a
```
Thanks to https://github.com/libjxl/libjxl/issues/3489 . | Just curious why depends on libjpeg, can we use jpegli instead? | https://api.github.com/repos/libjxl/libjxl/issues/3598/comments | 1 | 2024-05-22T05:21:05Z | 2024-08-25T16:39:54Z | https://github.com/libjxl/libjxl/issues/3598 | 2,309,596,139 | 3,598 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Failing to decode the following image with jpegli: [photo_2024-05-21_21-56-01.zip](https://github.com/libjxl/libjxl/files/15393696/photo_2024-05-21_21-56-01.zip)
**To Reproduce**
Try to decode the mentioned image with jpegli
**Expected behavior**
Image is decoded
**Environment**
- OS: Linux
- Compiler version: GCC 12
- CPU type: x86_64
- cjxl/djxl version string: I don't have cjxl/djxl... But I used v0.10.2 tag to build jpegli.
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| DHT marker: no Huffman table found | https://api.github.com/repos/libjxl/libjxl/issues/3597/comments | 10 | 2024-05-21T18:01:57Z | 2025-04-27T03:25:14Z | https://github.com/libjxl/libjxl/issues/3597 | 2,308,825,892 | 3,597 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Converting a PNG to a JXL with `cjxl -d 0`, and then comparing the original and converted images with `ssimulacra2` results in a score < 100.
**To Reproduce**
1. Use `cjxl -d 0` to convert a PNG to a JXL (e.g. `cjxl house.png house.jxl -d 0`)
2. Compare the original and converted image with `ssimulacra2` (e.g. `ssimulacra2 house.png house.jxl`)
3. Observe that the SSIMULACRA2 score is less than 100
4. Convert back to PNG with `djxl` (e.g. `djxl house.jxl house2.png`)
5. Compare the original PNG and new PNG with `ssimulacra2` (e.g. `ssimulacra2 house.png house2.png`)
6. Observe that the SSIMULACRA2 score is 100
**Expected behavior**
The SSIMULACRA2 score between the original PNG and converted JXL should be 100.
**Screenshots**
**Environment**
- OS: Windows
- Compiler version: N/A
- CPU type: x86_64
- cjxl/djxl version string: JPEG XL encoder v0.10.2 e148959 [AVX2,SSE2], JPEG XL decoder v0.10.2 e148959 [AVX2,SSE2]
**Additional context**
- Only steps 1-3 are necessary to reproduce the problem. Steps 4-6 show that the data is not actually lost, despite the <100 score.
- The attached `house.png`, `house.jxl.png`, and `house2.png` are the files used in the examples I've given in the steps to reproduce.
- The exact score I got from SSIMULACRA2 was 96.67210224.
- `cjxl`, `djxl`, and `ssimulacra2` all came from `jxl-x64-windows-static` in release 0.10.2.
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| Bundled SSIMULACRA2 detects differences in losslessly converted JXL | https://api.github.com/repos/libjxl/libjxl/issues/3594/comments | 2 | 2024-05-18T23:44:44Z | 2024-09-14T19:10:57Z | https://github.com/libjxl/libjxl/issues/3594 | 2,304,346,341 | 3,594 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
`cmake -DJPEGXL_STATIC=on` fails on Fedora 40.
**Output**
```
-- The C compiler identification is GNU 14.0.1
-- The CXX compiler identification is GNU 14.0.1
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- CMAKE_SYSTEM_PROCESSOR is x86_64
-- Performing Test CXX_FUZZERS_SUPPORTED
-- Performing Test CXX_FUZZERS_SUPPORTED - Failed
-- Performing Test CXX_MACRO_PREFIX_MAP
-- Performing Test CXX_MACRO_PREFIX_MAP - Success
-- Performing Test CXX_NO_RTTI_SUPPORTED
-- Performing Test CXX_NO_RTTI_SUPPORTED - Success
-- Found PkgConfig: /usr/bin/pkg-config (found version "2.1.0")
-- tcmalloc version -- tcmalloc 2.8.0 disabled due to https://github.com/gperftools/gperftools/issues/1204
-- Performing Test JXL_HWY_DISABLED_TARGETS_FORCED
-- Performing Test JXL_HWY_DISABLED_TARGETS_FORCED - Failed
-- Compiled IDs C:GNU, C++:GNU
-- Disabled AVX512 (set JPEGXL_ENABLE_AVX512 to enable it)
-- Disabled AVX512_SPR (set JPEGXL_ENABLE_AVX512_SPR to enable it)
-- Disabled AVX512_ZEN4 (set JPEGXL_ENABLE_AVX512_ZEN4 to enable it)
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Failed
-- Check if compiler accepts -pthread
-- Check if compiler accepts -pthread - no
-- Looking for pthread_create in pthreads
-- Looking for pthread_create in pthreads - not found
-- Looking for pthread_create in pthread
-- Looking for pthread_create in pthread - not found
CMake Error at /usr/share/cmake/Modules/FindPackageHandleStandardArgs.cmake:230 (message):
Could NOT find Threads (missing: Threads_FOUND)
Call Stack (most recent call first):
/usr/share/cmake/Modules/FindPackageHandleStandardArgs.cmake:600 (_FPHSA_FAILURE_MESSAGE)
/usr/share/cmake/Modules/FindThreads.cmake:226 (FIND_PACKAGE_HANDLE_STANDARD_ARGS)
CMakeLists.txt:225 (find_package)
-- Configuring incomplete, errors occurred!
```
**Environment**
- OS: Fedora 40
- Compiler version: GCC 14, Clang 18, Clang 15
- CPU type: x86_64
**Additional context**
The dynamic libraries and executables build without issue. There have been some recent header changes so maybe it's https://gcc.gnu.org/gcc-14/porting_to.html#header-dep-changes ?
| JPEGXL_STATIC fails to build on Fedora 40 | https://api.github.com/repos/libjxl/libjxl/issues/3593/comments | 5 | 2024-05-17T17:41:32Z | 2024-12-23T19:43:02Z | https://github.com/libjxl/libjxl/issues/3593 | 2,303,283,945 | 3,593 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Errors occur during the build process.
**To Reproduce**
<details>
<summary>my script</summary>
```shell
#!/bin/bash
while getopts 'c' flag ; do
case "${flag}" in
c) CLNFL=1 ;; # clean build
esac
done
set -e
export CC=clang CXX=clang++
CFL='-flto=thin -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -pipe'
LFL='-flto=thin -Wl,--gc-sections'
if [[ ! -d libjxl || $CLNFL == 1 ]]; then
rm -rf libjxl
git clone --depth 1 https://github.com/libjxl/libjxl.git
fi
cd libjxl/third_party
if [[ -z "$(ls -A skcms)" ]]; then git clone --single-branch https://skia.googlesource.com/skcms; cd skcms; git checkout '42030a7'; cd ..; fi
#if [[ -z "$(ls -A skcms)" ]]; then git clone --single-branch https://skia.googlesource.com/skcms; fi
if [[ -z "$(ls -A sjpeg)" ]]; then git clone --depth 1 https://github.com/webmproject/sjpeg.git; fi
if [[ -z "$(ls -A libjpeg-turbo)" ]]; then git clone --branch "3.0.2" --depth 1 https://github.com/libjpeg-turbo/libjpeg-turbo.git; fi
cd ..
rm -rf build
cmake -B build -G Ninja -S ./ \
-DCMAKE_BUILD_TYPE=Release \
-DBUILD_TESTING=OFF \
-DJPEGXL_ENABLE_JPEGLI=ON \
-DJPEGXL_ENABLE_DOXYGEN=OFF \
-DJPEGXL_ENABLE_MANPAGES=OFF \
-DJPEGXL_ENABLE_BENCHMARK=OFF \
-DJPEGXL_ENABLE_EXAMPLES=OFF \
-DJPEGXL_BUNDLE_LIBPNG=OFF \
-DJPEGXL_ENABLE_JNI=OFF \
-DJPEGXL_ENABLE_OPENEXR=OFF \
-DJPEGXL_STATIC=ON \
-DJPEGXL_WARNINGS_AS_ERRORS=OFF \
-DJPEGXL_FORCE_SYSTEM_BROTLI=ON \
-DJPEGXL_FORCE_SYSTEM_LCMS2=ON \
-DJPEGXL_FORCE_SYSTEM_HWY=ON \
-Wno-dev \
-DCMAKE_C_FLAGS="$CFL" -DCMAKE_CXX_FLAGS="$CFL" -DCMAKE_EXE_LINKER_FLAGS="$LFL"
ninja -C build
strip -s build/tools/*.exe
```
</details>
**Errors**
<details>
<summary>with skcms @ 42030a7</summary>
```
[3/256] Building CXX object lib/CMakeFiles/jxl_cms.dir/__/third_party/skcms/skcms.cc.obj
FAILED: lib/CMakeFiles/jxl_cms.dir/__/third_party/skcms/skcms.cc.obj
D:\msys2\clang64\bin\clang++.exe -DFJXL_ENABLE_AVX512=0 -DHWY_DISABLED_TARGETS="(HWY_SSSE3|HWY_AVX3|HWY_AVX3_SPR|HWY_AVX3_ZEN4|HWY_SVE|HWY_SVE2|HWY_SVE_256|HWY_SVE2_128|HWY_RVV)" -D__DATE__=\"redacted\" -D__TIMESTAMP__=\"redacted\" -D__TIME__=\"redacted\" -ID:/msys2/home/Gautama/libjxl -ID:/msys2/home/Gautama/libjxl/third_party/skcms -isystem D:/msys2/home/Gautama/libjxl/build/lib/include -flto=thin -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -pipe -fno-rtti -O3 -DNDEBUG -std=c++11 -fvisibility=hidden -fvisibility-inlines-hidden -fmacro-prefix-map=D:/msys2/home/Gautama/libjxl=. -funwind-tables -Xclang -mrelax-all -Wno-builtin-macro-redefined -Wall -fmerge-all-constants -fno-builtin-fwrite -fno-builtin-fread -Wextra -Wc++11-compat -Warray-bounds -Wformat-security -Wimplicit-fallthrough -Wno-register -Wno-unused-function -Wno-unused-parameter -Wnon-virtual-dtor -Woverloaded-virtual -Wvla -Wdeprecated-increment-bool -Wfloat-overflow-conversion -Wfloat-zero-conversion -Wfor-loop-analysis -Wgnu-redeclared-enum -Winfinite-recursion -Wliteral-conversion -Wno-c++98-compat -Wno-unused-command-line-argument -Wprivate-header -Wself-assign -Wstring-conversion -Wtautological-overlap-compare -Wthread-safety-analysis -Wundefined-func-template -Wunreachable-code -Wunused-comparison -Wno-cast-align -Wno-double-promotion -Wno-float-equal -Wno-format-nonliteral -Wno-shadow -Wno-sign-conversion -Wno-zero-as-null-pointer-constant -Wno-used-but-marked-unused -Wno-unused-template -Wno-unused-member-function -Wno-shadow-field-in-constructor -Wno-language-extension-token -Wno-global-constructors -Wno-c++98-compat-pedantic -DJPEGXL_ENABLE_SKCMS=1 -DHWY_SHARED_DEFINE -Wno-psabi -MD -MT lib/CMakeFiles/jxl_cms.dir/__/third_party/skcms/skcms.cc.obj -MF lib\CMakeFiles\jxl_cms.dir\__\third_party\skcms\skcms.cc.obj.d -o lib/CMakeFiles/jxl_cms.dir/__/third_party/skcms/skcms.cc.obj -c D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc
In file included from D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc:2369:
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:157:15: error: always_inline function '_mm512_cvtph_ps' requires target feature 'evex512', but would be inlined into function 'F_from_Half' that is compiled without support for 'evex512'
157 | return (F)_mm512_cvtph_ps((__m256i)half);
| ^
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:229:12: error: always_inline function '_mm512_mask_floor_ps' requires target feature 'evex512', but would be inlined into function 'floor_' that is compiled without support for 'evex512'
229 | return _mm512_mask_floor_ps(x, (__mmask16)-1, x);
| ^
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:517:18: error: '__builtin_ia32_gathersiv16si' needs target feature avx512f,evex512
517 | U32 v = (U32)_mm512_i32gather_epi32((__m512i)(3*ix), p4, 1);
| ^
D:/msys2/clang64/lib/clang/18/include/avx512fintrin.h:7698:13: note: expanded from macro '_mm512_i32gather_epi32'
7698 | ((__m512i)__builtin_ia32_gathersiv16si((__v16si)_mm512_undefined_epi32(), \
| ^
In file included from D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc:2369:
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:517:18: error: always_inline function '_mm512_undefined_epi32' requires target feature 'evex512', but would be inlined into function 'gather_24' that is compiled without support for 'evex512'
D:/msys2/clang64/lib/clang/18/include/avx512fintrin.h:7698:51: note: expanded from macro '_mm512_i32gather_epi32'
7698 | ((__m512i)__builtin_ia32_gathersiv16si((__v16si)_mm512_undefined_epi32(), \
| ^
In file included from D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc:2369:
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:573:22: error: '__builtin_ia32_gathersiv8di' needs target feature avx512f,evex512
573 | __m512i lo = _mm512_i32gather_epi64(_mm512_extracti32x8_epi32((__m512i)(6*ix), 0), p8, 1),
| ^
D:/msys2/clang64/lib/clang/18/include/avx512fintrin.h:7722:13: note: expanded from macro '_mm512_i32gather_epi64'
7722 | ((__m512i)__builtin_ia32_gathersiv8di((__v8di)_mm512_undefined_epi32(), \
| ^
In file included from D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc:2369:
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:573:22: error: always_inline function '_mm512_undefined_epi32' requires target feature 'evex512', but would be inlined into function 'gather_48' that is compiled without support for 'evex512'
D:/msys2/clang64/lib/clang/18/include/avx512fintrin.h:7722:49: note: expanded from macro '_mm512_i32gather_epi64'
7722 | ((__m512i)__builtin_ia32_gathersiv8di((__v8di)_mm512_undefined_epi32(), \
| ^
In file included from D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc:2369:
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:573:45: error: '__builtin_ia32_extracti32x8_mask' needs target feature avx512dq,evex512
573 | __m512i lo = _mm512_i32gather_epi64(_mm512_extracti32x8_epi32((__m512i)(6*ix), 0), p8, 1),
| ^
D:/msys2/clang64/lib/clang/18/include/avx512dqintrin.h:1256:13: note: expanded from macro '_mm512_extracti32x8_epi32'
1256 | ((__m256i)__builtin_ia32_extracti32x8_mask((__v16si)(__m512i)(A), (int)(imm), \
| ^
In file included from D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc:2369:
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:574:22: error: '__builtin_ia32_gathersiv8di' needs target feature avx512f,evex512
574 | hi = _mm512_i32gather_epi64(_mm512_extracti32x8_epi32((__m512i)(6*ix), 1), p8, 1);
| ^
D:/msys2/clang64/lib/clang/18/include/avx512fintrin.h:7722:13: note: expanded from macro '_mm512_i32gather_epi64'
7722 | ((__m512i)__builtin_ia32_gathersiv8di((__v8di)_mm512_undefined_epi32(), \
| ^
In file included from D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc:2369:
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:574:22: error: always_inline function '_mm512_undefined_epi32' requires target feature 'evex512', but would be inlined into function 'gather_48' that is compiled without support for 'evex512'
D:/msys2/clang64/lib/clang/18/include/avx512fintrin.h:7722:49: note: expanded from macro '_mm512_i32gather_epi64'
7722 | ((__m512i)__builtin_ia32_gathersiv8di((__v8di)_mm512_undefined_epi32(), \
| ^
In file included from D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc:2369:
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:574:45: error: '__builtin_ia32_extracti32x8_mask' needs target feature avx512dq,evex512
574 | hi = _mm512_i32gather_epi64(_mm512_extracti32x8_epi32((__m512i)(6*ix), 1), p8, 1);
| ^
D:/msys2/clang64/lib/clang/18/include/avx512dqintrin.h:1256:13: note: expanded from macro '_mm512_extracti32x8_epi32'
1256 | ((__m256i)__builtin_ia32_extracti32x8_mask((__v16si)(__m512i)(A), (int)(imm), \
| ^
In file included from D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc:2369:
D:/msys2/home/Gautama/libjxl/third_party/skcms/src/Transform_inl.h:184:17: error: '__builtin_ia32_vcvtps2ph512_mask' needs target feature avx512f,evex512
184 | return (U16)_mm512_cvtps_ph((__m512 )f, _MM_FROUND_CUR_DIRECTION );
| ^
D:/msys2/clang64/lib/clang/18/include/avx512fintrin.h:3800:31: note: expanded from macro '_mm512_cvtps_ph'
3800 | #define _mm512_cvtps_ph _mm512_cvt_roundps_ph
| ^
D:/msys2/clang64/lib/clang/18/include/avx512fintrin.h:3786:13: note: expanded from macro '_mm512_cvt_roundps_ph'
3786 | ((__m256i)__builtin_ia32_vcvtps2ph512_mask((__v16sf)(__m512)(A), (int)(I), \
| ^
11 errors generated.
[20/256] Building CXX object lib/CMakeFiles/jxl_cms.dir/jxl/cms/jxl_cms.cc.obj
ninja: build stopped: subcommand failed.
```
</details>
<details>
<summary>with skcms @ 9478a63</summary>
```
[234/256] Linking CXX executable tools\djxl.exe
FAILED: tools/djxl.exe
C:\WINDOWS\system32\cmd.exe /C "cd . && D:\msys2\clang64\bin\clang++.exe -flto=thin -ffunction-sections -fdata-sections -march=native -mtune=native -O3 -pipe -fno-rtti -O3 -DNDEBUG -flto=thin -Wl,--gc-sections -static -static-libgcc -static-libstdc++ tools/CMakeFiles/djxl.dir/djxl_main.cc.obj -o tools\djxl.exe -Wl,--out-implib,tools\libdjxl.dll.a -Wl,--major-image-version,0,--minor-image-version,0 -Wl,-Bstatic -lstdc++ -lpthread -Wl,-Bdynamic lib/libjxl.a lib/libjxl_extras_codec.a lib/libjxl_threads.a tools/libjxl_tool.a lib/libjxl.a lib/libjxl_cms.a D:/msys2/clang64/lib/libbrotlidec.a D:/msys2/clang64/lib/libbrotlicommon.a D:/msys2/clang64/lib/libbrotlienc.a D:/msys2/clang64/lib/libgif.a D:/msys2/clang64/lib/libjpeg.a third_party/sjpeg/libsjpeg.a D:/msys2/clang64/lib/libpng.a D:/msys2/clang64/lib/libz.a -pthread -Wl,-Bstatic -lstdc++ -lpthread -Wl,-Bdynamic D:/msys2/clang64/lib/libhwy.a -lkernel32 -luser32 -lgdi32 -lwinspool -lshell32 -lole32 -loleaut32 -luuid -lcomdlg32 -ladvapi32 && cd ."
ld.lld: error: undefined symbol: skcms_private::baseline::run_program(skcms_private::Op const*, void const**, long long, char const*, char*, int, unsigned long long, unsigned long long)
>>> referenced by D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc
>>> libjxl_cms.a(skcms.cc.obj)
ld.lld: error: undefined symbol: skcms_private::skx::run_program(skcms_private::Op const*, void const**, long long, char const*, char*, int, unsigned long long, unsigned long long)
>>> referenced by D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc
>>> libjxl_cms.a(skcms.cc.obj)
ld.lld: error: undefined symbol: skcms_private::hsw::run_program(skcms_private::Op const*, void const**, long long, char const*, char*, int, unsigned long long, unsigned long long)
>>> referenced by D:/msys2/home/Gautama/libjxl/third_party/skcms/skcms.cc
>>> libjxl_cms.a(skcms.cc.obj)
clang++: error: linker command failed with exit code 1 (use -v to see invocation)
[251/256] Building CXX object lib/CMakeFiles/jxl_extras_core_nocodec-obj.dir/extras/enc/pnm.cc.obj
ninja: build stopped: subcommand failed.
```
</details>
**Environment**
- OS: Windows 11 Pro 22631.3593
- Environment: msys2/clang64
- Compiler version: clang 18.1.5
- CPU type: x86_64
| Failed to build main with skcms in msys2/clang64 | https://api.github.com/repos/libjxl/libjxl/issues/3591/comments | 9 | 2024-05-17T07:55:43Z | 2024-05-28T09:45:20Z | https://github.com/libjxl/libjxl/issues/3591 | 2,302,066,835 | 3,591 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
I can't use jpegli libjpeg.so as a drop-in replacement.
**To Reproduce**
Drop in the libjpeg.so.62.3.0 instead of the system library.
gwenview: symbol lookup error: /usr/lib64/libgwenviewlib.so.5: undefined symbol: jround_up, version LIBJPEG_6.2
**Expected behavior**
For it to work since it is the same major soversion.
**Screenshots**
If applicable, add screenshots or example input/output images to help explain your problem.
**Environment**
- OS: Fedora 40
- Compiler version: gcc (GCC) 14.1.1 20240507 (Red Hat 14.1.1-1)
- CPU type: x86_64
- cjxl/djxl version string: cjxl v0.10.2 [AVX2,SSE4,SSE2]
**Additional context**
I am cheating a bit: the major soname version on my system library is indeed 62, but the minor is not 3 but 4 : libjpeg.so.62.4.0
I have libjpeg-turbo 3.0.2, but it seems you are building with 2.1.5.1 from last year. Is there something I could do to make it effectively a drop in replacement? | JPEG LI drop-in replacement for the system library with the same name | https://api.github.com/repos/libjxl/libjxl/issues/3588/comments | 0 | 2024-05-15T20:09:30Z | 2024-05-16T09:09:27Z | https://github.com/libjxl/libjxl/issues/3588 | 2,298,753,753 | 3,588 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
`benchmark_xl --print_details_csv` does not output the SSIMULACRA2 score as expected. However, `--print_details` without the csv does output it as expected.
**To Reproduce**
`benchmark_xl --print_details_CSV --encode_reps=1 --decode_reps=1 --input "./*.png" --codec jxl:7:d1` outputs the follow headers:
`method,image,error,size,pixels,enc_speed,dec_speed,bpp,dist,psnr,p,bppp,qabpp`
Default output has these headers:
`Encoding kPixels Bytes BPP E MP/s D MP/s Max norm SSIMULACRA2 PSNR pnorm BPP*pnorm QABPP Bugs`
`--print_details` outputs the following columns
`jxl:7:d1 example.png error:0 size: 44264 pixels: 401408 enc_speed:1.97544563 dec_speed:43.35366785 bpp:0.88217474 dist:1.50978231 psnr:43.85183584 ssimulacra2:87.72 p:0.53096151 bppp:0.46840084 qabpp:1.33189174`
**Expected behavior**
Would be great if the headers matched and were in the same order in each output.
**Screenshots**
N/A
**Environment**
- OS: macOS Sonoma 14.4.1
- Compiler version: n/a
- CPU type: M1 Max
- cjxl/djxl version string: benchmark_xl v0.10.2 0.10.2 [NEON]
**Additional context**
N/A
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| Benchmark-xl producing different output depending on parameters called. SSIMULACRA2 missing when --print_details_csv used. | https://api.github.com/repos/libjxl/libjxl/issues/3585/comments | 0 | 2024-05-14T19:16:25Z | 2024-05-30T13:45:24Z | https://github.com/libjxl/libjxl/issues/3585 | 2,296,219,294 | 3,585 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
The following test fails for armv7l-linux platform:
```
2978/7666 Test #2958: TransferFunctionsTargetTestGroup/TransferFunctionsTargetTest.TestPqEncodedFromDisplay/NEON_WITHOUT_AES # GetParam() = 536870912 ...........................................................................................................................................................................................................................................................................................................................................................................***Failed 0.98 sec
Running main() from /build/source/googletest/src/gtest_main.cc
Note: Google Test filter = TransferFunctionsTargetTestGroup/TransferFunctionsTargetTest.TestPqEncodedFromDisplay/NEON_WITHOUT_AES
[==========] Running 1 test from 1 test suite.
[----------] Global test environment set-up.
[----------] 1 test from TransferFunctionsTargetTestGroup/TransferFunctionsTargetTest
[ RUN ] TransferFunctionsTargetTestGroup/TransferFunctionsTargetTest.TestPqEncodedFromDisplay/NEON_WITHOUT_AES
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31825029850006104
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31586802005767822
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35572302341461182
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.32755899429321289
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.34616672992706299
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.38591563701629639
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.30870497226715088
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31915187835693359
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.48413765430450439
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.34998428821563721
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.32587313652038574
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31491947174072266
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31358742713928223
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.37572312355041504
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31834566593170166
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.34484386444091797
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31863033771514893
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31600654125213623
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.4836958646774292
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.36299705505371094
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35342562198638916
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.81288421154022217
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.32098913192749023
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.43078625202178955
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.32766711711883545
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.46281552314758301
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.3615947961807251
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.50101768970489502
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35244190692901611
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.32313978672027588
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.32157790660858154
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.32209169864654541
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.33173489570617676
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.33122944831848145
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.39967000484466553
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.32911968231201172
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.34171700477600098
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31078994274139404
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35325610637664795
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35478127002716064
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.43785154819488525
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.30984222888946533
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.96046448e-07 vs 5e-07
f = 0.97221016883850098
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.50119113922119141
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35653603076934814
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.49654769897460938
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.64983105659484863
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35459482669830322
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.36024773120880127
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31262421607971191
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35773515701293945
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.34757530689239502
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.96046448e-07 vs 5e-07
f = 0.94286859035491943
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31473541259765625
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.33544600009918213
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.32465457916259766
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.34389722347259521
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31026196479797363
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31200778484344482
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.32381975650787354
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31950807571411133
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31250262260437012
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.33498513698577881
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35062885284423828
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.34454870223999023
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35588574409484863
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.36063051223754883
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.46181261539459229
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.31478917598724365
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.36363577842712402
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.33343064785003662
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.30853712558746338
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.86675345897674561
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.35502970218658447
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.36441803e-07 vs 5e-07
f = 0.38460981845855713
./lib/jxl/cms/transfer_functions_test.cc:37: Failure
Expected: (abs_err) < (5e-7), actual: 5.96046448e-07 vs 5e-07
f = 0.97885000705718994
max abs err 5.960464e-07
[ FAILED ] TransferFunctionsTargetTestGroup/TransferFunctionsTargetTest.TestPqEncodedFromDisplay/NEON_WITHOUT_AES, where GetParam() = 536870912 (969 ms)
[----------] 1 test from TransferFunctionsTargetTestGroup/TransferFunctionsTargetTest (969 ms total)
[----------] Global test environment tear-down
[==========] 1 test from 1 test suite ran. (969 ms total)
[ PASSED ] 0 tests.
[ FAILED ] 1 test, listed below:
[ FAILED ] TransferFunctionsTargetTestGroup/TransferFunctionsTargetTest.TestPqEncodedFromDisplay/NEON_WITHOUT_AES, where GetParam() = 536870912
1 FAILED TEST
```
**To Reproduce**
Run armv7l-linux test suite on aarch64 kernel
**Expected behavior**
All tests pass
**Environment**
- OS: NixOS unstable
- Compiler version: gcc 13.2.0
- CPU type: aarch64 kernel, armv7l userspace
- cjxl/djxl version string: ???
**Additional context**
Full build log:
[libjxl-0.10.2.log](https://github.com/libjxl/libjxl/files/15287092/libjxl-0.10.2.log)
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| NEON_WITHOUT_AES test fails on armv7l-linux (0.10.2) | https://api.github.com/repos/libjxl/libjxl/issues/3571/comments | 0 | 2024-05-12T17:08:37Z | 2024-05-15T08:03:08Z | https://github.com/libjxl/libjxl/issues/3571 | 2,291,420,867 | 3,571 |
[
"libjxl",
"libjxl"
] | I found a JPEG image that cannot be decompressed by djpegli.exe. I can't say what exactly is wrong with it. It has unoptimal huffman encoding. It works after optimization, which removes restart marker 0xFFDD.
https://coverartarchive.org/release/9d628829-6a1c-47b5-9d72-3f0be929cd55/25011158016.jpg
> t:\working>djpegli.exe "25011158016.jpg" "25011158016.pnm"
> Read 2734380 compressed bytes.
> jpegli decoding failed | djpegli: Can't decompress an image | https://api.github.com/repos/libjxl/libjxl/issues/3570/comments | 3 | 2024-05-12T00:39:24Z | 2024-05-13T11:12:10Z | https://github.com/libjxl/libjxl/issues/3570 | 2,291,112,974 | 3,570 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
I had a hard time compiling libjxl. First, it failed because could not find “-lgmock”. Second, it failed because could not find the header files for highway.
**To Reproduce**
In addition to the OS mentioned, libjxl-0.10.2, my Clang is Apple clang version 14.0.0 (clang-1400.0.29.202); my gtest (gmock) is 1.14.0; my highway is 1.1.0; and my gcc is gcc (the real one) is 10.2.0 (do not think is relevant, but list it anyway).
**Expected behavior**
I expected the libraries and associated files to compile.
**Solution**
I was able to compile everything I wanted successfully, by doing the following:
1. Running:
./deps.sh
3. Running:
cmake -S . -B build -DJPEGXL_FORCE_SYSTEM_BROTLI=ON -DJPEGXL_FORCE_SYSTEM_LCMS2=ON -DJPEGXL_FORCE_SYSTEM_HWY=ON -DJPEGXL_ENABLE_DEVTOOLS=ON -DJPEGXL_TEST_TOOLS=OFF -DJPEGXL_ENABLE_JNI=OFF -DJPEGXL_ENABLE_JPEGLI=OFF -DJPEGXL_ENABLE_SKCMS=OFF -DJPEGXL_ENABLE_MANPAGES=OFF -DCMAKE_INSTALL_RPATH=/usr/local -DCMAKE_BUILD_TYPE=Release
4. Replacing the string “-lgmock” with the full path to the library:
find ./ -name link.txt -exec sed -i '.bak' 's/\-lgmock/\/usr\/local\/lib\/libgmock\.a/g' {} \;
5. Adding a link to the directory hwy to the directory the contained the CMakeLists.txt (in this case libjxl-0.10.2)
ln -s third_party/highway/hwy hwy
6. After these steps, running:
cmake --build build
built everything without issues.
**Environment**
- OS: macos 12.7.4
- Compiler version: Apple clang-1400.0.29.202
- CPU type: x86_64
- cjxl/djxl version string: cjxl v0.10.2 [AVX2,SSE4,SSE2] (but irrelevant because post hoc)
ZP | Trouble compiling on macos (Monterey 12.7.4) | https://api.github.com/repos/libjxl/libjxl/issues/3568/comments | 0 | 2024-05-10T08:57:17Z | 2024-11-02T23:10:53Z | https://github.com/libjxl/libjxl/issues/3568 | 2,289,309,252 | 3,568 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
```
[ 0%] Generating cjxl.1
File "/home/p/.asdf/shims/a2x", line 3
exec /home/p/.asdf/bin/asdf exec "a2x" "$@" # asdf_allow: ' asdf '
SyntaxError: invalid syntax
```
**To Reproduce**
```
which a2x
/home/p/.asdf/shims/a2x
mkdir build
cd build
cmake ..
cmake --build .
```
**Expected behavior**
```
[ 0%] Generating cjxl.1
[ 0%] Generating djxl.1
[ 0%] Built target manpages
```
**Environment**
- OS: WSL2
- Compiler version: Debian clang 17.0.6
- CPU type: x86_64
- cjxl/djxl version string: v0.10.2 661fec1b [AVX2,SSE4,SSE2]
**Additional context**
Line 435f of CMakeLists.txt:
```
file(STRINGS "${ASCIIDOC}" ASCIIDOC_SHEBANG LIMIT_COUNT 1)
if(ASCIIDOC_SHEBANG MATCHES "/sh|/bash" OR MINGW)
```
.asdf/shims/a2x
```
#!/usr/bin/env bash
exec /home/p/.asdf/bin/asdf exec "a2x" "$@" # asdf_allow: ' asdf '
```
I think the problem is the SheBang line which has "/usr/bin/env bash",
so it isn't recognized as a Shell script.
| a2x Syntax Error with .asdf/PyEnv | https://api.github.com/repos/libjxl/libjxl/issues/3564/comments | 1 | 2024-05-07T18:02:32Z | 2024-05-13T10:41:53Z | https://github.com/libjxl/libjxl/issues/3564 | 2,283,932,167 | 3,564 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
I tried to encode black png generated with `imagemagick convert` but failed:
```
$ cjpegli c.png o.jpg
Read 100x100 image, 260 bytes.
Encoding [YUV d1.000 AQ p2 OPT]
jpegli encoding failed
```
**To Reproduce**
Steps to reproduce the behavior:
cjpegli c.png o.jpg
**Expected behavior**
jpg created by cjpegli
**Screenshots**
Test image:
**Environment*
*
- OS: Linux
- Compiler version: gcc 13
- CPU type: x86_64
- cjxl/djxl version string: cjxl v0.10.2 e1489592 [AVX2,SSE4,SSE2]
**Additional context**
Add any other context about the problem here.
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| jpegli encoding failed for black png | https://api.github.com/repos/libjxl/libjxl/issues/3560/comments | 4 | 2024-05-07T05:59:28Z | 2025-06-13T06:20:30Z | https://github.com/libjxl/libjxl/issues/3560 | 2,282,349,100 | 3,560 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
In relased zip archives tgere are backslashes in paths. Zip uses forward slash as path separator.
It's a problem add not all un/zip clients assume `\` to be a `/`.
**To Reproduce**
Get the [released](https://github.com/libjxl/libjxl/releases) archives and see the path names, like `cmake\hwy\hwy-config-release.cmake`.
**Expected behavior**
To use forward slashed as path separator in zips.
**Environment**
- OS: Linux
**Additional context**
[specification](https://pkware.cachefly.net/webdocs/APPNOTE/APPNOTE-6.3.10.TXT) is clear about that:
> 4.4.17.1 The name of the file, with optional relative path.
> The path stored MUST NOT contain a drive or
> device letter, or a leading slash. All slashes
MUST be forward slashes '/'
| Backslashes in released zip archives paths | https://api.github.com/repos/libjxl/libjxl/issues/3559/comments | 9 | 2024-05-06T21:33:01Z | 2025-03-02T01:06:16Z | https://github.com/libjxl/libjxl/issues/3559 | 2,281,817,873 | 3,559 |
[
"libjxl",
"libjxl"
] | 1. [Here](https://github.com/libjxl/libjxl/blob/9d5477dc7c2604c596980de1bca072489d4b5aed/lib/include/jxl/color_encoding.h#L100), `Rendering` is misspelled as `Renderig`.
2. There seems to be a copy-paste error [here](https://github.com/libjxl/libjxl/blob/9d5477dc7c2604c596980de1bca072489d4b5aed/lib/include/jxl/color_encoding.h#L81), as the enum constant is called `JXL_TRANSFER_FUNCTION_709`, but the comment refers to `SMPTE RP 431-2` instead of `ITU-R BT.709-6`. | Typos in color_encoding.h | https://api.github.com/repos/libjxl/libjxl/issues/3553/comments | 1 | 2024-05-03T16:19:24Z | 2024-05-07T12:29:26Z | https://github.com/libjxl/libjxl/issues/3553 | 2,278,016,705 | 3,553 |
[
"libjxl",
"libjxl"
] | Why does cjpegli ignore sRGB ICC profiles? Is this intended?
Works with all other profiles, but all sRGB variants are missing after conversion.
| cjpegli ignores sRGB ICC profiles | https://api.github.com/repos/libjxl/libjxl/issues/3549/comments | 3 | 2024-05-02T03:53:24Z | 2024-05-27T07:42:31Z | https://github.com/libjxl/libjxl/issues/3549 | 2,274,539,369 | 3,549 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
When I do a batch jpeg transcoding, some of these images in CMYK color space cannot be transcoded
**To Reproduce**
cjxl .\1.jpg 1.jxl
JPEG XL encoder v0.10.2 e148959 [AVX2,SSE2]
Note: Implicit-default for JPEG is lossless-transcoding. To silence this message, set --lossless_jpeg=(1|0).
Encoding [JPEG, lossless transcode, effort: 7]
Error while decoding the JPEG image. It may be corrupt (e.g. truncated) or of an unsupported type (e.g. CMYK).
EncodeImageJXL() failed.
cjxl .\1.jpg 1.jxl --lossless_jpeg=0 -d 1 -v -v -v -v
JPEG XL encoder v0.10.2 e148959 [AVX2,SSE2]
Getting pixel data failed.
**Expected behavior**
Images are correctly transcoded. Provide more effective information in case of an error occurred.
**Environment**
- OS: Windows 11 Professional Workstation
- Compiler version: (https://github.com/libjxl/libjxl/releases/download/v0.10.2/jxl-x64-windows-static.zip)
- CPU type: x86_64
- cjxl/djxl version string: JPEG XL encoder v0.10.2 e148959 [AVX2,SSE2]
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| Unable to transcode a CMYK jpeg to jxl | https://api.github.com/repos/libjxl/libjxl/issues/3540/comments | 1 | 2024-04-29T13:05:08Z | 2024-05-14T10:58:53Z | https://github.com/libjxl/libjxl/issues/3540 | 2,268,997,608 | 3,540 |
[
"libjxl",
"libjxl"
] | **Describe the solution you'd like**
If the input PNG image contains a physical resolution tag in 'pHYs', it should be written to the output in a JFIF marker after conversion of pixels per meter into pixels per inch. JFIF implies YCbCr colors, so I accept that it may not be present in XYB encoding, and it adds only 18 bytes of overhead.
**Describe alternatives you've considered**
Use a 3rd party image optimizer afterwards. | cjpegli: Pass on the DPI resolution into a JFIF marker | https://api.github.com/repos/libjxl/libjxl/issues/3533/comments | 2 | 2024-04-26T03:35:45Z | 2024-04-29T11:58:24Z | https://github.com/libjxl/libjxl/issues/3533 | 2,264,892,635 | 3,533 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
faster_decoding 1 and 2 when encoding with lossless results in extremely large file sizes. faster_decoding 3 and 4 do not
**To Reproduce**
`cjxl img-in.png img-out.jxl -d 0 -e 7 --faster_decoding=1` or `cjxl img-in.jpeg img.jxl -d 0 -e 7 --lossless_jpeg=0 --faster_decoding=1`
**Expected behavior**
Create reasonably sized files
**Screenshots**
Not Applicable
**Environment**
- OS: Arch Linux
- Compiler version: Unkown
- CPU type: x86_64
- cjxl/djxl version string: cjxl v0.10.2 e1489592 [AVX2,SSE4,SSE2]
**Additional context**
The issue does not occur when doing lossless-transcoding or during normally lossy transcoding `cjxl mona.png mona-f1.jxl -d 0.1 -e 5 --faster_decoding=1` (though the overall hit is very large still)
results of lossless compression
```
➜ jxl-debug l mona*
Permissions Size User Date Modified Name
.rw-r--r-- 17M quack 25 Apr 13:13 mona-f0.jxl
.rw-r--r-- 100M quack 25 Apr 13:14 mona-f1.jxl
.rw-r--r-- 100M quack 25 Apr 13:15 mona-f2.jxl
.rw-r--r-- 26M quack 25 Apr 13:16 mona-f3.jxl
.rw-r--r-- 26M quack 25 Apr 13:16 mona-f4.jxl
.rw-r--r-- 27M quack 25 Apr 13:13 mona.png
```
```
➜ jxl-debug l wind*
Permissions Size User Date Modified Name
.rw-r--r-- 51M quack 25 Apr 13:17 windwaker-0.jxl
.rw-r--r-- 533M quack 25 Apr 13:18 windwaker-1.jxl
.rw-r--r-- 533M quack 25 Apr 13:19 windwaker-2.jxl
.rw-r--r-- 90M quack 25 Apr 13:20 windwaker-3.jxl
.rw-r--r-- 90M quack 25 Apr 13:21 windwaker-4.jxl
.rw-r--r-- 23M quack 25 Apr 13:17 windwaker35xnative.jpg
```
results of lossy compression
```
➜ jxl-debug l
Permissions Size User Date Modified Name
.rw-r--r-- 4.2M quack 25 Apr 13:27 mona-f0.jxl
.rw-r--r-- 26M quack 25 Apr 13:27 mona-f1.jxl
.rw-r--r-- 26M quack 25 Apr 13:27 mona-f4.jxl
.rw-r--r-- 27M quack 25 Apr 13:13 mona.png
```
| faster_decoding=1 or faster_decoding=2 has extremely bad size with lossless images. | https://api.github.com/repos/libjxl/libjxl/issues/3532/comments | 2 | 2024-04-25T17:32:16Z | 2025-05-01T23:48:08Z | https://github.com/libjxl/libjxl/issues/3532 | 2,264,133,896 | 3,532 |
[
"libjxl",
"libjxl"
] | ### **The first two posts are just talking about metrics (which don't really matter), for visual comparisons you can skip [here](https://github.com/libjxl/libjxl/issues/3530#issuecomment-2083474040)**
---
**Summary**
In my limited testing, libjxl 0.8.2 outperforms libjxl 0.10.2 in nonphotographic images like artwork/screenshots. Similar behavior can be seen in photographic images but the gap in quality is smaller and could be with in margin of error.
**Environment**
Libjxl 0.10.2 and 0.8.2 are downloaded from the releases page
Identical behavior on Windows 10 and Ubuntu 22.04
**Testing Methodology**
Nonphotographic image: [_ultraman](https://github.com/libjxl/libjxl/assets/164440799/cc207f8d-3ebb-4683-be8d-152c9e60253a)
Photographic image: [donald-michael-chambers-x2d-xcd55v-4](https://github.com/libjxl/libjxl/assets/164440799/3304778b-14a6-4714-bc06-196f09790fc3)
I libjxl 0.10.2 distances 0.75, 1.5, and 3 to represent high-medium-low quality benchmarks. These 0.10.2 images will be used as reference, I adjusted the distance setting in 0.8.2 to match the bbp of the reference images. Then I measured the subsequent outputs using four metrics, ssimulacra2, butteraugli, butteraugli (3-norm), and dssim. Since dssim does not allow .jxl input, I converted the images into a 16-bit png for the most quality and consistency.
I noticed that in benchmark_xl 0.8.2, ssimularcra2 and butteraugli gave different results for the same image when compared to those in the 0.10.2 library. For the sake of consistency, I have decided that tools provided in 0.10.2 are more accurate than those in 0.8.2, therefore I have replaced the ssimulacra2 and butteraugli numbers with the those provided in the latest release.
_Side note:_ Since I'm replacing certain variables for 0.8.2 tables, the "Aggregate" row is not accurate, it shouldn't be necessary anyways since the difference is quite noticeable.
Nonphotographic image comparisons:
**0.8.2**
```
_ultraman.png
Encoding kPixels Bytes BPP E MP/s D MP/s Max norm SSIMULACRA2 pnorm BPP*pnorm Bugs
-------------------------------------------------------------------------------------------------------------------------
jxl:d0.594 1387 312905 1.8035754 5.065 88.485 1.05268490 88.36280502 0.475509 0.857616334878 0
jxl:d1.22 1387 198023 1.1413989 5.118 106.640 1.67951285 83.06127014 0.756590 0.863580964129 0
jxl:d2.62 1387 122924 0.7085304 5.472 113.922 3.41860055 74.48578629 1.146457 0.812299869103 0
Aggregate: 1387 196753 1.1340793 5.215 102.439 1.82155991 82.18873767 0.744380 0.844186007485 0
```
**0.10.2**
```
_ultraman.png
Encoding kPixels Bytes BPP E MP/s D MP/s Max norm SSIMULACRA2 PSNR pnorm BPP*pnorm QABPP Bugs
----------------------------------------------------------------------------------------------------------------------------------------
jxl:d0.75 1387 313129 1.8048665 4.101 86.106 1.20206707 87.80768917 44.70 0.46107022 0.832170203284 2.170 0
jxl:d1.5 1387 199238 1.1484021 4.211 98.007 2.00189194 82.26755929 41.30 0.75180506 0.863374498072 2.299 0
jxl:d3 1387 123075 0.7094007 4.376 110.069 3.62111758 72.93681785 37.61 1.22165818 0.866645225771 2.569 0
Aggregate: 1387 197282 1.1371303 4.228 97.570 2.05780390 80.76737921 41.10 0.75094364 0.853920767103 2.340 0
```
**dssim metrics**
```
High quality (~1.80 bpp)
0.00028548 _ultraman-0.8.2
0.00029972 _ultraman-0.10.2
Medium quality (~1.14 bpp)
0.00075359 _ultraman-0.8.2
0.00076116 _ultraman-0.10.2
Low quality (~0.71 bpp)
0.00174440 _ultraman-0.8.2
0.00192789 _ultraman-0.10.2
```
Photographic image comparisons:
**0.8.2**
```
donald-michael-chambers-x2d-xcd55v-4.png
Encoding kPixels Bytes BPP E MP/s D MP/s Max norm SSIMULACRA2 pnorm BPP*pnorm Bugs
--------------------------------------------------------------------------------------------------------------------------
jxl:d0.614 6370 620830 0.7796923 5.601 161.527 1.07826220 88.77737185 0.41565914 0.323904888692 0
jxl:d1.39 6370 342875 0.4306122 5.771 169.412 1.85036587 83.73489551 0.67712914 0.291309491353 0
jxl:d3.17 6370 169212 0.2125110 5.545 106.820 3.31106091 73.83532398 1.18592679 0.252022503415 0
Aggregate: 6370 330089 0.4145544 5.638 142.982 1.87781044 82.08119056 0.69367425 0.287565746072 0
```
**0.10.2**
```
donald-michael-chambers-x2d-xcd55v-4.png
Encoding kPixels Bytes BPP E MP/s D MP/s Max norm SSIMULACRA2 PSNR pnorm BPP*pnorm QABPP Bugs
----------------------------------------------------------------------------------------------------------------------------------------
jxl:d0.75 6370 620482 0.7792548 7.007 122.930 1.18638035 88.24965401 48.27 0.44016851 0.343003408582 0.924 0
jxl:d1.5 6370 342590 0.4302540 7.295 149.051 2.14467233 83.09617108 44.74 0.73977772 0.318292357622 0.923 0
jxl:d3 6370 169512 0.2128878 5.143 163.187 3.78650298 73.32046639 41.27 1.24507709 0.265061695613 0.806 0
Aggregate: 6370 330295 0.4148124 6.406 144.066 2.12785027 81.31545211 44.67 0.74012570 0.307013311650 0.883 0
```
**dssim metrics**
```
High quality (~0.78 bbp)
0.00016938 donald-0.8.2-hq.png
0.00020154 donald-0.10.2-hq.png
Medium quality (~0.43 bbp)
0.00053243 donald-0.8.2-mq.png
0.00056884 donald-0.10.2-mq.png
Low quality (~0.21 bbp)
0.00165394 donald-0.8.2-lq.png
0.00167708 donald-0.10.2-lq.png
```
**Analysis/Conclusion**
For artwork, it's clear to see that 0.8.2 has consistently scored higher in all tested metrics by significant amount. Libjxl 0.10.2 lags behind in quality and in speed for the same bbp. The speed regression is really strange since 0.10.2 should be much faster. I guess it slowed down for small and medium sized images. (should this is a separate issue?) On the other hand, the photo consistently scored higher in all metrics as well, but the gap is smaller, so its probably within margin of error. Most likely there is no regression for the photo. Also, here is where 0.10.2 outperforms 0.8.2 in terms of speed, maybe because it's a larger image?
_PS: I also noticed that 0.9.2 gives similar subpar results compared to 0.8.2 but I didn't have the time to go into it with detail. So it seems that the regression happened in between 0.8.2 and 0.9.2. So more testing is needed regardless. Also, if somebody can provide me with a script that automates this process, that would make my life a lot easier._ | Regression in lossy nonphotographic image quality | https://api.github.com/repos/libjxl/libjxl/issues/3530/comments | 23 | 2024-04-25T10:26:56Z | 2025-02-19T17:01:01Z | https://github.com/libjxl/libjxl/issues/3530 | 2,263,233,465 | 3,530 |
[
"libjxl",
"libjxl"
] | null | API for region of interest decoding | https://api.github.com/repos/libjxl/libjxl/issues/3519/comments | 1 | 2024-04-22T09:58:12Z | 2025-06-20T11:00:43Z | https://github.com/libjxl/libjxl/issues/3519 | 2,256,107,373 | 3,519 |
[
"libjxl",
"libjxl"
] | Encode_effort.md is not accurate anymore with the release of libjxl 0.10.x. Outside of the "brute force" effort now being 11, does VarDCT e10 do anything special? Or does it simply do more Butteraugli iterations? What about e9 vs e10 lossless? I think it should also mention which efforts use the new streaming mode by default in 0.10.x. Jon Sneyers in the JXL discord [mentioned](https://discord.com/channels/794206087879852103/794206170445119489/1215017451931045948) that e10 disables streaming by default.
I should also mention that for benchmarking.md, "glacier" is equivalent to e10, so what is e11? _Edit: Apparently e11 is called "tectonic plate"._ And finally the target bitrate command `-r` is depreciated. | The effort documentation is outdated | https://api.github.com/repos/libjxl/libjxl/issues/3517/comments | 9 | 2024-04-21T10:44:50Z | 2025-01-05T14:16:50Z | https://github.com/libjxl/libjxl/issues/3517 | 2,255,012,945 | 3,517 |
[
"libjxl",
"libjxl"
] | When it comes to manga (grayscale Japanese comic) files, many of the final production files that publishers end up encoding are JPEGs, due to their inclusion in EPUB files. When doing so, the majority of publishers do not encode these as grayscale (8BPP) JPEGs, even though this would result in no loss of data.
For a project I am working on, I would like to losslessly convert these JPEG files to JXL files to reduce their size, and have done so using `cjxl input.jpg output.jxl`. I would like to, however, also reduce the colorspace of these images at the same time, making them grayscale. I have confirmed that the images unsurprisingly remain in color after the conversion:
```
jxlinfo output.jxl
JPEG XL file format container (ISO/IEC 18181-2)
JPEG XL image, 900x1350, (possibly) lossless, 8-bit RGB
Color space: 3144-byte ICC profile, CMM type: "Lino", color space: "RGB ", rendering intent: 0
JPEG bitstream reconstruction data available
Brotli-compressed Exif metadata: 74 compressed bytes
```
Would it somehow be possible to losslessly reduce the colorspace during the JPEG --> JXL conversion? | Losslessly converting color JPEG to grayscale JXL | https://api.github.com/repos/libjxl/libjxl/issues/3516/comments | 5 | 2024-04-21T00:54:07Z | 2025-03-02T00:29:22Z | https://github.com/libjxl/libjxl/issues/3516 | 2,254,786,104 | 3,516 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
`-d 0` is producing an ouput which is not visually identical to the original png file.
**To Reproduce**
Steps to reproduce the behavior:
* `cjxl FILE0_0_copy.png FILE0_0_copy.jxl --distance=0 --effort=9`
**Expected behavior**
jxl image is visually identical to the png image.
**Screenshots**
**Environment**
- OS: macOS 14.4.1
- Compiler version: ?
- CPU type: Apple M1
- cjxl/djxl version string: cjxl v0.10.2
**Additional context**
cjxl installed using homebrew.
Original png file:
| cjxl lossless (`-d 0`) is not identical. | https://api.github.com/repos/libjxl/libjxl/issues/3515/comments | 5 | 2024-04-20T15:36:59Z | 2024-04-22T10:29:47Z | https://github.com/libjxl/libjxl/issues/3515 | 2,254,603,379 | 3,515 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
When using `cjxl` to encode an image with the Alpha channel to a lossy JPEG XL file, the default value for `keep_invisible` should be 0. The results show a possible default as 1 instead.
```
--keep_invisible=0|1
disable/enable preserving color of invisible pixels (default: 1 if lossless, 0 if lossy).
```
**To Reproduce**
```
cjxl file.png file.jxl -d 1
cjxl file.png file.jxl -d 1 --keep_invisible=0
```
**Expected behavior**
The default value of `keep_invisible` for lossy JPEG XL images should be 0.
**Screenshots**
<img width="431" alt="截圖 2024-04-20 23 18 17" src="https://github.com/libjxl/libjxl/assets/161316/75c34982-d9ee-4e99-b155-a80fce8b6c0f">
**Environment**
- OS: macOS Sonoma 14.4.1 (23E224)
- Compiler version: Apple clang version 15.0.0 (clang-1500.3.9.4)
- CPU type: x86_64
- cjxl/djxl version string: JPEG XL encoder v0.10.2 0.10.2 [AVX2,SSE4,SSE2] | Lossy cjxl with Alpha channel not using --keep_invisible=0 as default | https://api.github.com/repos/libjxl/libjxl/issues/3514/comments | 5 | 2024-04-20T15:22:21Z | 2024-11-08T14:45:48Z | https://github.com/libjxl/libjxl/issues/3514 | 2,254,597,867 | 3,514 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
djpegli should decode xyb-jpeg to rgb-png to ensure that most programs can read it.
**To Reproduce**
```bash
❯ cjpegli --xyb in.png out.jpeg
Read 1920x1080 image, 2048379 bytes.
Encoding [XYB d1.000 AQ p2 OPT]
Compressed to 187502 bytes (0.723 bpp).
1920 x 1080, 52.273 MP/s [52.27, 52.27], , 1 reps, 1 threads.
❯ djpegli out.jpeg out.png
Read 187502 compressed bytes.
1920 x 1080, 56.085 MP/s [56.09, 56.09], , 1 reps, 1 threads.
❯ ffmpeg -i out.png -i in.png -filter_complex libvmaf -f null -
[Parsed_libvmaf_0 @ 0x5d4fcc17d600] VMAF score: 8.031800
```
**Screenshots**
**Environment**
- OS: Arch Linux + CachyOS&ALHP Repository
- Compiler version: GCC 13.2.1
- CPU type: x86_64-v3
- cjxl/djxl version string: v0.10.2 e1489592 [AVX2,SSE4,SSE2] | djpegli should decode xyb-jpeg to rgb-png | https://api.github.com/repos/libjxl/libjxl/issues/3513/comments | 2 | 2024-04-20T07:26:49Z | 2024-04-25T09:23:55Z | https://github.com/libjxl/libjxl/issues/3513 | 2,254,429,479 | 3,513 |
[
"libjxl",
"libjxl"
] | **Is your feature request related to a problem? Please describe.**
While XYB JPEGs work in many browsers, colors appear wrong in most other apps.
**Describe the solution you'd like**
Most apps are expecting an Adobe APP14 segment. This is supported in other encoders such as libjpeg-turbo. They [describe](https://github.com/libjpeg-turbo/libjpeg-turbo/blob/2dfe6c0fe9e18671105e94f7cbf044d4a1d157e6/libjpeg.txt#L1193):
> If TRUE, an Adobe APP14 marker is emitted. jpeg_set_defaults() and
> jpeg_set_colorspace() set this TRUE if JPEG color space RGB, CMYK,
> or YCCK is selected, otherwise FALSE.
The specific format can be found [here](https://github.com/libjpeg-turbo/libjpeg-turbo/blob/2dfe6c0fe9e18671105e94f7cbf044d4a1d157e6/jcmarker.c#L390).
**Additional context**
I have tested adding this segment to XYB JPEGs and they show correct colors in other apps like Photoshop, Safari, and macOS Finder/Preview. | Add Adobe APP14 tag to JPEG images with specific colorspaces (fix XYB appearance in most apps) | https://api.github.com/repos/libjxl/libjxl/issues/3512/comments | 2 | 2024-04-19T22:15:39Z | 2025-01-25T23:17:34Z | https://github.com/libjxl/libjxl/issues/3512 | 2,254,103,229 | 3,512 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Nightly builds result in bpp higher than the uncompressed source. Using 0.10.2 correctly reduces file size.
```
cjxl metro_noord_2k.hdr.pfm metro.hdr.jxl -d 0
JPEG XL encoder v0.10.2 e148959 [AVX2,SSE2]
Encoding [Modular, lossless, effort: 7]
Compressed to 4144.0 kB (15.808 bpp).
2048 x 1024, 0.200 MP/s [0.20, 0.20], , 1 reps, 16 threads.
cjxl metro_noord_2k.hdr.pfm metro.hdr.jxl -d 0
JPEG XL encoder v0.10.2 60dfa1b [AVX2,SSE2]
Encoding [Modular, lossless, effort: 7]
Compressed to 15722.2 kB (59.976 bpp).
2048 x 1024, 0.548 MP/s [0.55, 0.55], , 1 reps, 16 threads.
```
**To Reproduce**
Run `cjxl -d 0 metro.hdr.jxl Test.jxl` on 0.10.2 and nightly
**Environment**
- OS: Windows
- CPU type: x86_64
- cjxl/djxl version string: JPEG XL encoder v0.10.2 60dfa1b [AVX2,SSE2] | 32bit Float Density Regression | https://api.github.com/repos/libjxl/libjxl/issues/3511/comments | 5 | 2024-04-19T17:07:13Z | 2025-04-11T18:19:21Z | https://github.com/libjxl/libjxl/issues/3511 | 2,253,477,716 | 3,511 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
MotionPhotos are a cursed file where a video appended to a jpeg file. `cjxl` will currently fail to transcode the image with lossless bitstream reconstruction data with the error "JPEG bitstream reconstruction data could not be created. Possibly there is too much tail data."
**To Reproduce**
`cjxl filein.jpeg fileout.jxl`
File provided
**Expected behavior**
A) CJXL could discard the data and throw a message to the user. While discarding this data isn't *technically* lossless as far as the entire image file is concerned. This extra data doesn't pertain to the actual jpeg itself needed for proper reconstruction of the "image". CJXL should likely by default exit with an "Error" to bail out of any scripts unless an extra flag is passed.
B) CJXL could append the same video file to the end of the JXL file in the same manor as a motionphoto. This would retain "true" losslessness of the entire file and not just the jpeg data.
**Screenshots**
Not Aplicable
**Environment**
- OS: Linux
- cjxl/djxl version string: cjxl v0.10.2 e1489592 [AVX2,SSE4,SSE2]
**Additional context**
MotionPhotos are cursed, but there are a good number of them out in the wild and failing to transcode these with lossless reconstruction is a pain point for at least some users
Exiftool can identify multiple types of motionphotos "Motion Photo" is the most common, However some "Micro Videos" exist too. exiftool can find the appropriate offset for those using `exiftool "$FILENAME" | grep "Micro Video Offset" | sed -E 's/^[^:]+: ([0-9]+)$/\1/'`
Almost all motion photos seem to be mp4 based and have a `ftypmp42` signature to them. Identification based on this seems reliable.
to extract the video from a MotionPhoto `exiftool -b -EmbeddedVideoFile motionphoto.jpg > out.mp4`
to extract the JPEG data from a motionphoto (replaces the current file, creates a new "original file") `exiftool -b -trailer:all= motionphoto.jpeg`
Example Image
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| cjxl fails on "MotionPhoto" jpegs | https://api.github.com/repos/libjxl/libjxl/issues/3502/comments | 1 | 2024-04-18T04:18:38Z | 2024-04-18T08:26:10Z | https://github.com/libjxl/libjxl/issues/3502 | 2,249,702,392 | 3,502 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Running `cjpegli -h` shows the following:
```
cjpegli -h | head -n 3
Usage: cjpegli INPUT OUTPUT [OPTIONS...]
INPUT
the input can be PNG, APNG, GIF, EXR, PPM, PFM, or PGX
```
However, the program does support JXL input, though it is not listed here.
**To Reproduce**
Steps to reproduce the behavior:
1. Run `cjpegli -h` and note the lack of JXL in the list of supported input formats.
2. Try encoding a JXL with it, e.g. `cjpegli foo.jxl foo.jpg`, and observe it working just fine.
**Expected behavior**
Include all supported formats in the list supplied by the help. JXL is not listed but supported, and possibly there are others?
**Screenshots**
N/A
**Environment**
- OS: Linux
- Compiler version: clang version 16.0.6
- CPU type: x86_64
- cjxl/djxl version string: cjpegli doesn't appear to print a version string, but I am using the one from libjxl v0.10.2
**Additional context**
N/A | JXL input is supported by cjpegli, but not listed in the help | https://api.github.com/repos/libjxl/libjxl/issues/3498/comments | 0 | 2024-04-17T09:03:57Z | 2024-04-22T09:59:34Z | https://github.com/libjxl/libjxl/issues/3498 | 2,247,769,014 | 3,498 |
[
"libjxl",
"libjxl"
] | **Is your feature request related to a problem? Please describe.**
Would be really nice two variants of libjpeg in build output - both shared and static
**Describe the solution you'd like**
Some `add_library(jpeg-static STATIC ...)`
**Describe alternatives you've considered**
I do
```
cp build/lib/libjpegli-static.a $DESTDIR/usr/local/lib64/libjpeg.a
ar rcs $DESTDIR/usr/local/lib64/libjpeg.a build/lib/CMakeFiles/jpegli-libjpeg-obj.dir/jpegli/libjpeg_wrapper.cc.o
```
right now which is not really convenient or portable... | Static libjpeg | https://api.github.com/repos/libjxl/libjxl/issues/3489/comments | 0 | 2024-04-16T11:52:04Z | 2024-04-22T07:18:44Z | https://github.com/libjxl/libjxl/issues/3489 | 2,245,850,328 | 3,489 |
[
"libjxl",
"libjxl"
] | 这个压缩有分压缩等级吗？需要怎么修改啊！ | 请教个问题！ | https://api.github.com/repos/libjxl/libjxl/issues/3488/comments | 3 | 2024-04-16T08:56:56Z | 2024-04-17T08:33:59Z | https://github.com/libjxl/libjxl/issues/3488 | 2,245,508,539 | 3,488 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
I have noticed https://github.com/libjxl/libjxl/commit/0f0997c9d7db7bbc161341af972ebb6d4094d8f0. Please don't make it into a release, jpegs are almost always untrusted input and aborting on them is a no-go. The commit message doesn't even explain why this is done... Lots of projects (at least Qt for sure) do setjmp/longjmp trick to recover from error_exit with libjpeg and its forks. The original libjpeg and its forks exit in the default error_exit handler, not after it.
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| Applications shouldn't abort while reading jpegs | https://api.github.com/repos/libjxl/libjxl/issues/3487/comments | 4 | 2024-04-16T01:44:43Z | 2025-03-02T09:45:39Z | https://github.com/libjxl/libjxl/issues/3487 | 2,244,924,672 | 3,487 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Getting the error mentioned in title with the following image
[image.zip](https://github.com/libjxl/libjxl/files/14985844/image.zip)
**To Reproduce**
Steps to reproduce the behavior:
I get the error while reading the image through Qt's QImageReader when linking it with jpegli. But I guess it should be reproducible with the utilities from libjxl repo just as well?
**Expected behavior**
Image is decoded
**Environment**
- OS: Linux
- Compiler version: GCC 12
- CPU type: x86_64
- cjxl/djxl version string: I don't have cjxl/djxl... But I used v0.10.2 tag to build jpegli.
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| Quantization table with index 0 not found | https://api.github.com/repos/libjxl/libjxl/issues/3486/comments | 8 | 2024-04-16T01:39:15Z | 2024-04-18T12:00:29Z | https://github.com/libjxl/libjxl/issues/3486 | 2,244,920,195 | 3,486 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
`djpegli` fails to decode a CMYK JPG to PNG.
**To Reproduce**
1. Obtain a CMYK JPG, for example [this one](https://i.postimg.cc/ZTB4nKrW/crash.jpg) (from [this issue](https://gitlab.com/wg1/jpeg-xl/-/issues/122)).
2. `./djpegli image.jpg output.png`
**Actual behavior**
```
Read 2249602 compressed bytes.
jpegli decoding failed
```
**Expected behavior**
A PNG is produced.
**Environment**
* OS: Linux
* CPU type: x86_64
* [Binary used](https://artifacts.lucaversari.it/libjxl/libjxl/latest/jxl-linux-x86_64-static.zip) (`c04b3ba`) | djpegli fails to decode CMYK image | https://api.github.com/repos/libjxl/libjxl/issues/3474/comments | 0 | 2024-04-11T10:33:43Z | 2024-04-22T09:49:02Z | https://github.com/libjxl/libjxl/issues/3474 | 2,237,413,206 | 3,474 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
```
libjxl/lib/extras/dec/jpg.cc:219:5: error: ‘jpeg_mem_src’ was not declared in this scope; did you mean ‘jpeg_stdio_src’?
219 | jpeg_mem_src(&cinfo, reinterpret_cast<const unsigned char*>(bytes.data()),
| ^~~~~~~~~~~~
| jpeg_stdio_src
make[2]: *** [lib/CMakeFiles/jxl_extras_core-obj.dir/build.make:232: lib/CMakeFiles/jxl_extras_core-obj.dir/extras/dec/jpg.cc.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:953: lib/CMakeFiles/jxl_extras_core-obj.dir/all] Error 2
make: *** [Makefile:141: all] Error 2
``` | Compilation failure. | https://api.github.com/repos/libjxl/libjxl/issues/3472/comments | 3 | 2024-04-10T18:51:42Z | 2024-04-24T15:51:32Z | https://github.com/libjxl/libjxl/issues/3472 | 2,236,216,482 | 3,472 |
[
"libjxl",
"libjxl"
] | Decoder already support image reorientation but there is no way to force certain orientation except for disabling reorientation entirely, at least as far as I can tell from looking at documentation and headers. This is useful for example in my case where after decoding JXL I have to flip it upside down because I have to deal with coordinate system where positive Y is up instead of down. | Add ability to force orientation when decoding JxlDecoder | https://api.github.com/repos/libjxl/libjxl/issues/3470/comments | 1 | 2024-04-10T10:49:31Z | 2024-04-11T06:10:24Z | https://github.com/libjxl/libjxl/issues/3470 | 2,235,295,635 | 3,470 |
[
"libjxl",
"libjxl"
] | Hey guys, Is there any offline gui app for windows to compress photos with this lib?
https://github.com/libjxl/libjxl/tree/main/lib/jpegli
I have webp convertor and I want to try this format too
I have no knowledge about programming so I can't build by myself
Thanks in advance | Offline convertor window app? | https://api.github.com/repos/libjxl/libjxl/issues/3468/comments | 1 | 2024-04-08T18:37:31Z | 2024-04-18T23:13:37Z | https://github.com/libjxl/libjxl/issues/3468 | 2,231,855,193 | 3,468 |
[
"libjxl",
"libjxl"
] | I'm doing some experiments with converting my processed images to jpg using jpegli instead of mozjpeg.
Now the option to encode more than 8 bits sounds great to me, as it might make the images very futureproof.
I just can't find any documentation on it.
Here are a few questions:
- what is the max number of bits? I'm working with 16 bit TIFF / PNG
- what should the input format be? is 16 bit png the only option?
- do I have to adjust the quality / distance setting? or would 8 bit decoding be the same, just file size bigger?
- is there currently any way to decode / check if it worked? if not, can we be sure that there will be a way in the future?
any info very much appreciated, thanks
| jpegli 10+ bit documentation | https://api.github.com/repos/libjxl/libjxl/issues/3464/comments | 3 | 2024-04-06T06:57:53Z | 2024-04-13T21:39:05Z | https://github.com/libjxl/libjxl/issues/3464 | 2,229,112,874 | 3,464 |
[
"libjxl",
"libjxl"
] | successful build on Ubuntu 22.04.1 server, no errors
then library can't be find >:{
fix, add to .bashrc:
`export LD_LIBRARY_PATH=/usr/local/lib` | errand in documentation - compiled library not found | https://api.github.com/repos/libjxl/libjxl/issues/3463/comments | 0 | 2024-04-05T18:47:41Z | 2024-04-10T08:06:22Z | https://github.com/libjxl/libjxl/issues/3463 | 2,228,629,069 | 3,463 |
[
"libjxl",
"libjxl"
] | failed compilation on Ubuntu 22.04.1
Steps to reproduce the behavior = following the manual to build from the source.
the fix:
`sudo apt install libstdc++-12-dev` | errand in documentation - missing library | https://api.github.com/repos/libjxl/libjxl/issues/3462/comments | 1 | 2024-04-05T18:44:10Z | 2024-04-10T08:05:34Z | https://github.com/libjxl/libjxl/issues/3462 | 2,228,622,780 | 3,462 |
[
"libjxl",
"libjxl"
] | Some photographers (like me) intentionally [use digital noise in their images for effect](https://www.quora.com/Does-anyone-intentionally-use-digital-noise-in-their-photography). Reducing noise should be a function of my image editing program, not my file compression algorithm. I realize that sacrificing image fidelity for compression is nothing new, but doing so in a way that makes specific aesthetic choices (i.e. ["to make the internet faster and more beautiful"](https://opensource.googleblog.com/2024/04/introducing-jpegli-new-jpeg-coding-library.html)) is crossing the line into image manipulation. Between this and all the AI enhancement my cell phone camera does, we'll soon be living in a world where it is literally impossible to create a photo that isn't "beautiful". | jpegli reduces noise in images | https://api.github.com/repos/libjxl/libjxl/issues/3461/comments | 7 | 2024-04-05T16:18:45Z | 2024-04-07T19:34:59Z | https://github.com/libjxl/libjxl/issues/3461 | 2,228,401,323 | 3,461 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
As noted by `eddie.sato` on this message https://discord.com/channels/794206087879852103/803645746661425173/1225342521014485103
The generation loss of the jxl exibit an issue which might be induced by a regression on the gaborish filter (encoding part, as stated by `_wb_`)
**To Reproduce**
Encode and decode the same image twice in a row
**Expected behavior**
The second encoding should insert a minimal distortion
**Screenshots**
[genloss.webm](https://github.com/libjxl/libjxl/assets/20285309/af5166f2-5f2a-4b9d-89e5-93e4b2101506)
**Environment**
**Additional context**
cf. the dicussion following the post on discord:
`_wb_` — Looks like jpegli zeroes AC quite quickly. And inv_gaborish(gaborish) should probably be made closer to the identity function.
`_wb_` — [...] the encoder side gaborish has been changed since then, I suspect causing a regression in terms of generation loss.
| Gaborish filter regression | https://api.github.com/repos/libjxl/libjxl/issues/3458/comments | 4 | 2024-04-05T08:59:27Z | 2025-03-01T19:54:46Z | https://github.com/libjxl/libjxl/issues/3458 | 2,227,436,687 | 3,458 |
[
"libjxl",
"libjxl"
] | Hi!
Recently I checked optimizations like Profile-Guided Optimization (PGO) and Post-Link Optimizations (PLO) improvements on multiple projects. The results are available [here](https://github.com/zamazan4ik/awesome-pgo/). According to the tests, all these optimizations can help with achieving better performance in many cases for many applications. I think trying to enable them for `libjxl` can be a good idea. I read [an article](https://www.phoronix.com/news/Google-Jpegli-Library) on Phoronix about a new JPEG encoding/decoding library - Jpegli - and decided to optimize it with PGO.
I already did some benchmarks and want to share my results here. Hopefully, they will be helpful.
## Test environment
* Fedora 39
* Linux kernel 6.7.6
* AMD Ryzen 9 5900x
* 48 Gib RAM
* SSD Samsung 980 Pro 2 Tib
* Compiler - Clang 17.0.6
* libjxl version: the latest for now from the `main` branch on commit `680d0e38683b6485e39807772c579252fe91f3a4`
* Disabled Turbo boost (for better results consistency across runs)
## Benchmark
I didn't find a good benchmark suite to evaluate performance gains on a large dataset. Instead, I use [these](https://sample-videos.com/download-sample-png-image.php) image samples. In all cases, an image for 30 Mib is used. In all cases, the library is configured with `cmake -DCMAKE_BUILD_TYPE=Release -DBUILD_TESTING=OFF -DENABLE_JPEGLI_DEFAULT=ON ..`. For the PGO training phase, additional flag `-fprofile-generate` is passed to the compiler, for the PGO optimization phase - `-fprofile-use` flag. The PGO training phase is done with the following command: `cjpegli Sample-png-image-30mb.png converted.jpeg -q 90`, where `cjpegli` - Jpegli's encoder, `Sample-png-image-30mb.png` - an input image.
All tests are done on the same machine, done multiple times, with the same background "noise" (as much as I can guarantee of course) - the results are reproducible at least on my machine. `taskset -c 0` is used for better stability across runs (to reduce OS scheduler influence).
## Results
Here are the results:
* Release: https://gist.github.com/zamazan4ik/95b1472be0042086c6d4646d6edab053
* PGO-optimized compared to Release: https://gist.github.com/zamazan4ik/3159661ef3943f8fad270c8634cd738d
* (just for reference) PGO-instrumented compared to Release: https://gist.github.com/zamazan4ik/1e9c37de3e95a95cf06c9f3135c9b7e3
Also, I tested the case when training and actual workloads differ. Here are the PGO optimized compared to a regular release benchmark, when another sample image is used (not the same as during the training phase): https://gist.github.com/zamazan4ik/4750fa6424a53e83638f4ab422f901a9
At least to the simple benchmarks above, PGO allows achieving better performance.
## Further steps
I can suggest the following action points:
* Perform more PGO benchmarks on libjxl. If it shows improvements - add a note to the documentation about possible improvements in libjxl performance with PGO.
* Providing an easier way (e.g. a build option) to build scripts with PGO can be helpful for the end-users and maintainers since they will be able to optimize libjxl according to their workloads.
* Optimize pre-built libjxl binaries (if any)
Here are some examples of how PGO optimization is integrated into other projects:
* Rustc: a CI [script](https://github.com/rust-lang/rust/blob/master/src/ci/stage-build.py) for the multi-stage build
* GCC:
- Official [docs](https://gcc.gnu.org/install/build.html), section "Building with profile feedback" (even AutoFDO build is supported)
- A [part](https://github.com/gcc-mirror/gcc/blob/4832767db7897be6fb5cbc44f079482c90cb95a6/configure#L7818) in a "wonderful" `configure` script
* Clang: [Docs](https://llvm.org/docs/HowToBuildWithPGO.html)
* Python:
- CPython: [README](https://github.com/python/cpython#profile-guided-optimization)
- Pyston: [README](https://github.com/pyston/pyston#building)
* Go: [Bash script](https://github.com/golang/go/blob/master/src/cmd/compile/profile.sh)
* V8: [Bazel flag](https://github.com/v8/v8/blob/main/BUILD.gn#L184)
* ChakraCore: [Scripts](https://github.com/chakra-core/ChakraCore/tree/master/Build/scripts/pgo)
* Chromium: [Script](https://chromium.googlesource.com/chromium/src/build/config/+/refs/heads/main/compiler/pgo/BUILD.gn)
* Firefox: [Docs](https://firefox-source-docs.mozilla.org/build/buildsystem/pgo.html)
- Thunderbird has PGO support too
* PHP - [Makefile command](https://github.com/php/php-src/blob/master/build/Makefile.global#L138) and old Centminmod [scripts](https://github.com/centminmod/php_pgo_training_scripts)
* MySQL: [CMake script](https://github.com/mysql/mysql-server/blob/8.0/cmake/fprofile.cmake)
* YugabyteDB: [GitHub commit](https://github.com/yugabyte/yugabyte-db/commit/34cb791ed9d3d5f8ae9a9b9e9181a46485e1981d)
* FoundationDB: [Script](https://github.com/apple/foundationdb/blob/1a6114a66f3de508c0cf0a45f72f3687ba05750c/contrib/generate_profile.sh)
* Zstd: [Makefile](https://github.com/facebook/zstd/blob/dev/programs/Makefile#L232)
* [Foot](https://codeberg.org/dnkl/foot): [Scripts](https://codeberg.org/dnkl/foot/src/branch/master/pgo)
* Windows Terminal: [GitHub PR](https://github.com/microsoft/terminal/pull/10071)
* Pydantic-core: [GitHub PR](https://github.com/pydantic/pydantic-core/pull/741)
* file.d: [GitHub PR](https://github.com/ozontech/file.d/pull/469)
* OceanBase: [CMake flag](https://github.com/oceanbase/oceanbase/blob/master/cmake/Env.cmake#L55)
I have some examples of how PGO information looks in the documentation:
* ClickHouse: https://clickhouse.com/docs/en/operations/optimizing-performance/profile-guided-optimization
* Databend: https://databend.rs/doc/contributing/pgo
* Vector: https://vector.dev/docs/administration/tuning/pgo/
* Nebula: https://docs.nebula-graph.io/3.5.0/8.service-tuning/enable_autofdo_for_nebulagraph/
* GCC: Official [docs](https://gcc.gnu.org/install/build.html), section "Building with profile feedback" (even AutoFDO build is supported)
* Clang:
- https://llvm.org/docs/HowToBuildWithPGO.html
- https://llvm.org/docs/AdvancedBuilds.html
* tsv-utils: https://github.com/eBay/tsv-utils/blob/master/docs/BuildingWithLTO.md
Please, do not treat the issue like a bug or smth like that. It's just a benchmark report with possible improvement idea for the project. | Profile-Guided Optimization (PGO) benchmark report | https://api.github.com/repos/libjxl/libjxl/issues/3456/comments | 0 | 2024-04-04T13:11:52Z | 2024-04-10T08:04:39Z | https://github.com/libjxl/libjxl/issues/3456 | 2,225,475,925 | 3,456 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
I use cjxl to recompress some JPEG files to JXL.
Then I use djxl to reconstruct them to JPEG, but it silently failed (no files were produced).
But I can successfully decode the JXL files into PNG files.
**To Reproduce**
JPEG file sample: [1c94fa30eea811e597bc25dc24f9c26c.zip](https://github.com/libjxl/libjxl/files/14865227/1c94fa30eea811e597bc25dc24f9c26c.zip)
```
$ cjxl 1c94fa30eea811e597bc25dc24f9c26c.jpg 1.jxl
JPEG XL encoder v0.10.2 e148959 [AVX2,SSE2]
Note: Implicit-default for JPEG is lossless-transcoding. To silence this message, set --lossless_jpeg=(1|0).
Encoding [JPEG, lossless transcode, effort: 7]
Compressed to 2746.1 kB including container
```
```
$ djxl 1.jxl 2.jpg
JPEG XL decoder v0.10.2 e148959 [AVX2,SSE2]
3677 x 5549, 282.590 MP/s [282.59, 282.59], , 1 reps, 16 threads.
```
The return code is 0, but nothing is produced.
```
$ djxl 1.jxl 2.png
JPEG XL decoder v0.10.2 e148959 [AVX2,SSE2]
Decoded to pixels.
3677 x 5549, 280.952 MP/s [280.95, 280.95], , 1 reps, 16 threads.
```
The content of 2.png is the same as the original JPEG.
**Environment**
- OS: tested on Debian and Windows
- Compiler version: all download from GitHub releases, so I guess clang
- CPU type: x86_64
- cjxl/djxl version string: see command line output above
| djxl silently failed to reconstruct some JPEG files | https://api.github.com/repos/libjxl/libjxl/issues/3455/comments | 1 | 2024-04-04T11:00:46Z | 2024-04-11T13:27:31Z | https://github.com/libjxl/libjxl/issues/3455 | 2,225,161,791 | 3,455 |
[
"libjxl",
"libjxl"
] | **Is your feature request related to a problem? Please describe.**
In @discourse we added a WASM version of MozJPEG, that is used to re-encode uploads in the user device before the actual upload to make uploads faster and save user bandwidth. You can read about it over https://blog.discourse.org/2021/07/faster-user-uploads-on-discourse-with-rust-webassembly-and-mozjpeg/.
This was heavily inspired by the work done at https://github.com/GoogleChromeLabs/squoosh, which ships many image codecs as WASM, like MozJPEG and JXL itself.
Since Jpegli provides better compression, I'd love to swap to this, provided we can get a WASM build of it's encoder that accepts a RGBA array as input.
**Describe the solution you'd like**
The project provides WASM files of it's releases, or documents a way for downstream projects to build it themselves.
Optionally, publish said package in NPM.
**Describe alternatives you've considered**
- Build it myself. Hard to justify the time investment given the unfamiliarity with this stack and that existing MozJPEG still works.
- Wait for squoosh project to add it. Looks like squoosh is in maintenance mode, as key people left the company.
**Additional context**
Discourse is OSS, and widely used. Compressing images from user uploads in forums have a huge impact as those images are downloaded many times over decades after they are uploaded due to the open nature of forums on the internet.
Also, I was asked to open this issue at https://twitter.com/jyzg/status/1775590308585034090
| Release a Jpegli WASM build | https://api.github.com/repos/libjxl/libjxl/issues/3454/comments | 0 | 2024-04-03T18:42:36Z | 2024-04-04T12:40:26Z | https://github.com/libjxl/libjxl/issues/3454 | 2,223,657,854 | 3,454 |
[
"libjxl",
"libjxl"
] | I just git cloned today for libjxl's source.
Then I ran cmake.
First it complained that google gtest was not available. I compiled it as well.
Upon re-running this, after removing the cache, I get this:
cmake -DCMAKE_INSTALL_PREFIX=/usr ..
-- The C compiler identification is GNU 14.0.1
-- The CXX compiler identification is GNU 14.0.1
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working C compiler: /usr/bin/cc - skipped
-- Detecting C compile features
-- Detecting C compile features - done
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++ - skipped
-- Detecting CXX compile features
-- Detecting CXX compile features - done
-- CMAKE_SYSTEM_PROCESSOR is x86_64
-- Performing Test CXX_FUZZERS_SUPPORTED
-- Performing Test CXX_FUZZERS_SUPPORTED - Failed
-- Performing Test CXX_MACRO_PREFIX_MAP
-- Performing Test CXX_MACRO_PREFIX_MAP - Success
-- Performing Test CXX_NO_RTTI_SUPPORTED
-- Performing Test CXX_NO_RTTI_SUPPORTED - Success
-- Found PkgConfig: /usr/bin/pkg-config (found version "0.29.2")
-- Performing Test JXL_HWY_DISABLED_TARGETS_FORCED
-- Performing Test JXL_HWY_DISABLED_TARGETS_FORCED - Failed
-- Compiled IDs C:GNU, C++:GNU
-- Disabled AVX512 (set JPEGXL_ENABLE_AVX512 to enable it)
-- Disabled AVX512_SPR (set JPEGXL_ENABLE_AVX512_SPR to enable it)
-- Disabled AVX512_ZEN4 (set JPEGXL_ENABLE_AVX512_ZEN4 to enable it)
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD
-- Performing Test CMAKE_HAVE_LIBC_PTHREAD - Success
-- Found Threads: TRUE
-- Performing Test ATOMICS_LOCK_FREE_INSTRUCTIONS
-- Performing Test ATOMICS_LOCK_FREE_INSTRUCTIONS - Success
-- Found GTest: /usr/lib/cmake/GTest/GTestConfig.cmake (found version "1.14.0")
-- Found HWY: /usr/lib/libhwy.so (found suitable version "1.1.0", minimum required is "1.0.7")
-- Checking for module 'libbrotlicommon'
-- Found libbrotlicommon, version 1.1.0
-- Checking for module 'libbrotlienc'
-- Found libbrotlienc, version 1.1.0
-- Checking for module 'libbrotlidec'
-- Found libbrotlidec, version 1.1.0
-- Found Brotli: /usr/include
CMake Error at third_party/CMakeLists.txt:100 (message):
Please run /Depot/jjjj/libjxl-27.03.2024/deps.sh to fetch the build
dependencies.
-- Configuring incomplete, errors occurred!
So, the first error in regards to libjxl's current handling is:
(1) it tells the user that deps.sh should be used. But I don't want to. I want
to be able to control what I install at all times. The correct solution here is
to report what is missing. For some reason, libjxl does not tell the user
what is missing now - look at the above output. It does not tell me what
is missing now. This leads me to the second error:
(2) libjxl should tell the user what exactly is missing. After I installed
googletest, it does not tell me EXACTLY what is missing; it only tells
me to use deps.sh, which I do not want to. So both (1) and (2) in
combination is wrong right now.
Yes, some users may want to use deps.sh - that is fine. I'd still reason
that this behaviour is wrong, but I am ok if cmake warns us and then
tells us that deps.sh should be used. But it should not assume that
everyone wants to use some random .sh file, even less so when there
is now no information shown what exactly is missing. That is, IMO,
an incorrectly set up build system situation. Other cmake based projects
report what is missing; I compile all of KDE from source, for instance,
and they do that, always telling the user what is missing.
| [Cmake issue] Please run /Depot/jjjj/libjxl-27.03.2024/deps.sh to fetch the build dependencies. | https://api.github.com/repos/libjxl/libjxl/issues/3443/comments | 1 | 2024-03-27T17:46:07Z | 2024-04-02T11:53:32Z | https://github.com/libjxl/libjxl/issues/3443 | 2,211,415,285 | 3,443 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
libjxl v0.6.0 (25ef914) compresses better than v0.10.2 e1489592
**To Reproduce**
```
cjxl -e 9 -I 100 -d 0 -E 3 -g 3 --patches 1 0.6.0.jxl
JPEG XL encoder v0.6.0 25ef914 [AVX2,SSE4,SSSE3,Scalar]
Read 10240x5760 image, 189.6 MP/s
Encoding [Modular, lossless, tortoise], 8 threads.
Compressed to 8949 bytes (0.001 bpp).
10240 x 5760, 0.16 MP/s [0.16, 0.16], 1 reps, 8 threads.
cjxl -e 10 -I 100 -d 0 -E 3 -g 3 --patches 1 0.10.2.jxl
JPEG XL encoder v0.10.2 e1489592 [AVX2,SSE4,SSE2]
Encoding [Modular, lossless, effort: 10]
Compressed to 28162 bytes (0.004 bpp).
10240 x 5760, 0.315 MP/s [0.32, 0.32], , 1 reps, 16 threads.
```
**Expected behavior**
Latest libjxl should compress as well as the older version.
**Screenshots**
A zip file of both jxl files:
[images.zip](https://github.com/libjxl/libjxl/files/14764954/images.zip)
**Environment**
- OS: Arch linux
- Compiler version:
17.0.6 for v0.10.2
12.0.1 for v0.6.0
- CPU type: x86_64, ryzen 2700x
- cjxl/djxl version string:
cjxl v0.6.0 25ef914 [AVX2,SSE4,SSSE3,Scalar]
cjxl v0.10.2 e1489592 [AVX2,SSE4,SSE2]
| Lossless density regression | https://api.github.com/repos/libjxl/libjxl/issues/3441/comments | 3 | 2024-03-26T21:47:34Z | 2024-03-28T10:23:52Z | https://github.com/libjxl/libjxl/issues/3441 | 2,209,392,560 | 3,441 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Jpegli compilation fails on Mac (`FROM_JXL_BOOL` not found). `brew install jpeg-xl` does not give jpegli support, so compiling from source is the only way forward on Mac.
**To Reproduce**
```shell
LIBJXL_VERSION="0.10.2"
git clone --depth 1 --branch "v$LIBJXL_VERSION" https://github.com/libjxl/libjxl .
./deps.sh
# Adapted from https://github.com/Homebrew/homebrew-core/blob/c5aeee48e2f2539182c8e83537368495dc2531cf/Formula/j/jpeg-xl.rb
cmake -S . -B build \
-DBUILD_TESTING=OFF \
-DCMAKE_BUILD_TYPE=Release \
-DCMAKE_INSTALL_LIBDIR=lib \
-DCMAKE_VERBOSE_MAKEFILE=ON \
-DJPEGXL_FORCE_SYSTEM_BROTLI=ON \
-DJPEGXL_FORCE_SYSTEM_HWY=ON \
-DJPEGXL_FORCE_SYSTEM_LCMS2=ON \
-DJPEGXL_ENABLE_JNI=OFF \
-DJPEGXL_ENABLE_JPEGLI=ON \
-DJPEGXL_ENABLE_SKCMS=OFF \
-DJPEGXL_VERSION="$LIBJXL_VERSION" \
-DJPEGXL_ENABLE_MANPAGES=ON \
-DPython_EXECUTABLE="$(which python)" \
-DPython3_EXECUTABLE="$(which python3)" \
-Wno-dev
cmake --build build
cmake --build build --target install
```
Observe the following error message:
```
In file included from /tmp/libjxl/lib/extras/dec/decode.cc:13:
In file included from /tmp/libjxl/lib/extras/dec/jpg.h:13:
In file included from /tmp/libjxl/lib/extras/codec.h:22:
In file included from /tmp/libjxl/lib/jxl/codec_in_out.h:20:
In file included from /tmp/libjxl/lib/jxl/frame_header.h:26:
In file included from /tmp/libjxl/lib/jxl/image_metadata.h:21:
/tmp/libjxl/lib/jxl/color_encoding_internal.h:348:12: error: use of undeclared identifier 'FROM_JXL_BOOL'
return FROM_JXL_BOOL(cms_.run(cms_data_, thread, buf_src, buf_dst, xsize));
^
1 error generated.
make[2]: *** [lib/CMakeFiles/jxl_extras_core-obj.dir/extras/dec/decode.cc.o] Error 1
make[1]: *** [lib/CMakeFiles/jxl_extras_core-obj.dir/all] Error 2
make: *** [all] Error 2
```
**Expected behavior**
Jpegli compilation should succeed on Mac.
**Screenshots**
**Environment**
- OS: Mac OS X Sonoma 14.4
- Compiler version: Apple clang 15.0.0
- CPU type: arm64
- cjxl/djxl version string: N/A
**Additional context**
| Jpegli compilation fails on Mac (`FROM_JXL_BOOL` not found) | https://api.github.com/repos/libjxl/libjxl/issues/3440/comments | 0 | 2024-03-25T21:27:07Z | 2024-04-02T11:53:23Z | https://github.com/libjxl/libjxl/issues/3440 | 2,206,752,311 | 3,440 |
[
"libjxl",
"libjxl"
] | [Encode_effort.md](https://github.com/libjxl/libjxl/blob/main/doc/encode_effort.md) is not accurate anymore with respect to libjxl 0.10.x. Outside of the "brute force" effort now being 11, does VarDCT e10 do anything special outside of more Butteraugli iterations than e9? What about e9 vs e10 lossless? I think it should also mention which efforts use the new streaming mode by default in 0.10.x. | Outdated effort documentation | https://api.github.com/repos/libjxl/libjxl/issues/3436/comments | 0 | 2024-03-22T03:42:11Z | 2024-04-21T10:39:50Z | https://github.com/libjxl/libjxl/issues/3436 | 2,201,634,264 | 3,436 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Trying to use lossy palette results in an identical output file as without
**To Reproduce**
`cjxl -m 1 --modular_palette_colors=0 --modular_lossy_palette Test.png Test.jxl`
`cjxl -m 1 Test.png Test2.jxl`
Compare output files
**Expected behavior**
Files should be different with significantly reduced colors
**Environment**
- OS: Windows 10
- CPU type: x86_64
- cjxl/djxl version string: JPEG XL encoder v0.10.2 e148959 [AVX2,SSE2] | Lossy/Delta Palette doesn't work | https://api.github.com/repos/libjxl/libjxl/issues/3433/comments | 2 | 2024-03-18T18:00:33Z | 2025-03-01T19:26:15Z | https://github.com/libjxl/libjxl/issues/3433 | 2,192,904,598 | 3,433 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Both `fcTL->IDAT->fcTL` and `IDAT->fcTL` chunks ordering seems to be [spec-conforming](https://wiki.mozilla.org/APNG_Specification#Chunk_Sequence_Numbers) ways to encode apng.
But neither ffmpeg, imagemagick, nor apngasm produces a cjxl acceptable apng with `IDAT->fcTL` ordering
* `ffmpeg -f concat -i ./concat_demuxer -plays 0 -c:v apng out.apng`
* `magick convert *.png APNG:out.apng`
* `apngasm *.png`
**To Reproduce**
Steps to reproduce the behavior:
```bash
convert -size 512x512 canvas:white '1.png'
convert -size 512x512 canvas:gray '2.png'
convert -delay 10 '1.png' '2.png' 'anim.apng'
cjxl 'anim.apng' 'out.jxl'
```
```log
JPEG XL encoder v0.10.0 b49cdd5 [AVX2,SSE4,SSE2]
./lib/extras/dec/apng.cc:739: JXL_FAILURE: IDAT chunk after fcTL chunk
./lib/extras/dec/decode.cc:142: JXL_FAILURE: Codecs failed to decode
Getting pixel data failed.
```
**Additional context**
Related commit: https://github.com/libjxl/libjxl/commit/4a9aaa83ad3c7b9d28140aa9a0b3be8d96548d74
| Apng encoding: `IDAT chunk after fcTL chunk` error when using apng created with imagemagick and ffmpeg | https://api.github.com/repos/libjxl/libjxl/issues/3430/comments | 1 | 2024-03-17T12:30:00Z | 2024-04-11T14:48:53Z | https://github.com/libjxl/libjxl/issues/3430 | 2,190,643,981 | 3,430 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
From my understanding, in lossy mode, JPEG XL transforms images to the XYB color space. Additionally, the ICC profile embedded within the image is always in the XYB color space. Subsequently, the image viewer directly converts it from XYB to the color profile of the display.
**To Reproduce**
I observed two distinct behaviors when comparing two identical PNG images, one in AdobeRGB and the other in DisplayP3.
Here is the jxlinfo for a lossy image converted from PNG in AdobeRGB
```
dbtsai@DBs-Mac ~ % jxlinfo 2016-12-28\ 21.18.45.00.jxl
JPEG XL file format container (ISO/IEC 18181-2)
JPEG XL image, 7952x5304, lossy, 16-bit RGB
Color space: RGB, D65, Custom primaries: red(x=0.639997,y=0.329997), green(x=0.210005,y=0.710005), blue(x=0.149998,y=0.060004)gamma(0.454707) transfer function, rendering intent: Perceptual
Brotli-compressed Exif metadata: 19980 compressed bytes
Brotli-compressed xml metadata: 505 compressed bytes
```
Here is the jxlinfo for a lossy image converted from PNG in DisplayP3
```
dbtsai@DBs-Mac ~ % jxlinfo 2016-12-28\ 21.18.45.00.jxl
JPEG XL file format container (ISO/IEC 18181-2)
JPEG XL image, 7952x5304, lossy, 16-bit RGB
Color space: RGB, D65, P3 primaries, sRGB transfer function, rendering intent: Perceptual
Brotli-compressed Exif metadata: 19933 compressed bytes
Brotli-compressed xml metadata: 505 compressed bytes
```
Steps to reproduce the behavior:
**Expected behavior**
I assumed that the lossy JXL files would share identical `Custom primaries` for the XYB color space. Yet, in the case of the file converted from DisplayP3, the resulting lossy JXL file exhibits P3 primaries.
**Environment**
- OS: Mac
- CPU type: Apple M1
- cjxl/djxl version string: JPEG XL encoder v0.10.2 0.10.2 [NEON]
| JPEG XL color space in lossy mode | https://api.github.com/repos/libjxl/libjxl/issues/3429/comments | 15 | 2024-03-17T05:57:39Z | 2024-04-22T12:31:41Z | https://github.com/libjxl/libjxl/issues/3429 | 2,190,494,874 | 3,429 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
When using CJXL to convert a JXL back to GIF file it doesn't convert the animated jxl back to a gif, it just makes a copy
**To Reproduce**
cjxl image.gif image.jxl
cjxl image.jxl image2.gif
**Expected behavior**
should be able to convert back the processed jxl back to the gif original file like it does with jpg image files. | JXL back to GIF issue | https://api.github.com/repos/libjxl/libjxl/issues/3428/comments | 5 | 2024-03-16T23:37:20Z | 2024-04-02T11:10:41Z | https://github.com/libjxl/libjxl/issues/3428 | 2,190,355,944 | 3,428 |
[
"libjxl",
"libjxl"
] | **Is your feature request related to a problem? Please describe.**
When inputting the unoptimized image, the file size of the output image will be larger than when inputting the optimized image.
**Describe the solution you'd like**
Optimize the output image by removing the alpha (if unnecessary), reducing color depth (if possible), etc.
I think either of the following is better:
1. Add an option to do this.
2. Make this the default behavior and add an option to not optimize (preserve the alpha, color depth, etc. of the input image).
**Describe alternatives you've considered**
Optimize the input image in advance.
**Additional context**
See #3413 for sample images and details.
| Feature request: cjxl: Optimize the output image when the input image is unoptimized | https://api.github.com/repos/libjxl/libjxl/issues/3415/comments | 1 | 2024-03-12T09:05:08Z | 2024-10-17T09:38:56Z | https://github.com/libjxl/libjxl/issues/3415 | 2,181,082,206 | 3,415 |
[
"libjxl",
"libjxl"
] | This is unoptimized image (752,757 B):
This is optimized image (653,559 B):
The optimization method is as follows:
```sh
oxipng -o max electrichearts_20201224A_kiki_c1-1024x512.png
zopflipng -my electrichearts_20201224A_kiki_c1-1024x512.png electrichearts_20201224A_kiki_c1-1024x512.png
```
Also, these images are completely identical:
```text
$ compare -metric AE unoptimized.png optimized.png /dev/null
0
```
I compressed these images losslessly:
```sh
cjxl -q 100 electrichearts_20201224A_kiki_c1-1024x512.png electrichearts_20201224A_kiki_c1-1024x512.jxl
```
The output sizes are as follows:
- 458,778 B (unoptimized)
- 458,631 B (optimized)
I expected the output size to be the same, but it was not. Is this the expected behavior?
## Environment
- OS: Arch Linux on Docker on openSUSE Leap 15.5
- CPU type: x86_64
- cjxl/djxl version string: cjxl v0.10.2 e1489592 [AVX2,SSE4,SSE2] | The output size is different when inputting optimized PNG image or unoptimized it | https://api.github.com/repos/libjxl/libjxl/issues/3413/comments | 4 | 2024-03-12T04:52:35Z | 2024-03-12T09:06:57Z | https://github.com/libjxl/libjxl/issues/3413 | 2,180,725,857 | 3,413 |
[
"libjxl",
"libjxl"
] | This is just FYI.
The good news:
kdenlive video editor on Windows has JXL support bundled-in. https://download.kde.org/stable/kdenlive/24.02/windows/
The bad news:
Some JXL images works but some cause crash of the whole application. Similar situation happened in the past with other apps built using MinGW.
kdenlive from the link uses libjxl 0.9.1, I reproduced the crash with an other app using 0.10.2 too.
```
kolourpaint.exe caused an Access Violation at location 00007FF9683615C8 in module libjxl.dll Reading from location FFFFFFFFFFFFFFFF.
AddrPC Params
00007FF9683615C8 0000019330E98560 0000019330E43D30 0000000000000000 libjxl.dll!DecodeGroupImpl+0x9c8 [C:/_/6be5d956/libjxl/lib/jxl/dec_group.cc @ 415]
413: HWY_ALIGN float* const block = group_dec_cache->dec_group_block;
414: // Dequantize and add predictions.
> 415: dequant_block(
416: acs, inv_global_scale, row_quant[bx], dec_state->x_dm_multiplier,
417: dec_state->b_dm_multiplier, x_cc_mul, b_cc_mul, acs.RawStrategy(),
00007FF968367152 0000019330E98560 0000008576FFFB80 0000000000000001 libjxl.dll!DecodeGroup+0x842 [C:/_/6be5d956/libjxl/lib/jxl/dec_group.cc @ 778]
776: dec_state, first_pass));
777:
> 778: JXL_RETURN_IF_ERROR(HWY_DYNAMIC_DISPATCH(DecodeGroupImpl)(
779: frame_header, get_block.get(), group_dec_cache, dec_state, thread,
780: group_idx, render_pipeline_input, decoded, draw));
00007FF968342483 0000019330C63760 0000000000000000 0000019330C637C8 libjxl.dll!ProcessACGroup+0x713 [C:/_/6be5d956/libjxl/lib/jxl/dec_frame.cc @ 491]
489: JXL_RETURN_IF_ERROR(group_dec_caches_[thread].InitOnce(
490: frame_header_.passes.num_passes, dec_state_->used_acs));
> 491: JXL_RETURN_IF_ERROR(DecodeGroup(frame_header_, br, num_passes, ac_group_id,
492: dec_state_, &group_dec_caches_[thread],
493: thread, render_pipeline_input, decoded_,
00007FF968342651 0000000000000000 0000019330C63FB8 0000019330CD21A0 libjxl.dll!JxlEncoderCreate+0x56261
00007FF9D4271F40 0000019330DD8400 0000000000000000 0000000000000000 libjxl_threads.dll!RunRange+0x90 [C:/_/6be5d956/libjxl/lib/threads/thread_parallel_runner_internal.cc @ 145]
143: }
144: for (uint32_t task = my_begin; task < my_end; ++task) {
> 145: self->data_func_(self->jpegxl_opaque_, task, thread);
146: }
147: }
00007FF9D4272017 0000000000000000 0000000000000000 000079C11682CB92 libjxl_threads.dll!ThreadFunc+0xa7 [C:/_/6be5d956/libjxl/lib/threads/thread_parallel_runner_internal.cc @ 175]
173: default:
174: lock.unlock();
> 175: RunRange(self, command, thread);
176: break;
177: }
00007FF9A7E608B1 0000019330DD38B8 0000000000000000 0000000000000000 libstdc++-6.dll!atomic_flag_test_and_set_explicit+0x1a1
0000000064944E11 0000019330DD3880 0000000000000000 0000019330F55910 libwinpthread-1.dll!pthread_create_wrapper+0x131
00007FF9DF74AF5A 00007FF9DF7A06D0 0000019330F55910 0000000000000000 msvcrt.dll!_beginthreadex+0x12a
00007FF9DF74B02C 0000000000000000 0000000000000000 0000000000000000 msvcrt.dll!_endthreadex+0xac
00007FF9E0CB7344 0000000000000000 0000000000000000 0000000000000000 KERNEL32.DLL!BaseThreadInitThunk+0x14
00007FF9E16C26B1 0000000000000000 0000000000000000 0000000000000000 ntdll.dll!RtlUserThreadStart+0x21
``` | libjxl built using MinGW-W64 may crash | https://api.github.com/repos/libjxl/libjxl/issues/3412/comments | 3 | 2024-03-11T17:37:33Z | 2024-04-07T04:59:01Z | https://github.com/libjxl/libjxl/issues/3412 | 2,179,766,431 | 3,412 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
--output_extra_channels does not work and only a single PNG is written
**To Reproduce**
djxl suz-multichannel-800px.jxl --output_extra_channels multichannel.png
**Environment**
- OS: Windows 10
- CPU type: x86_64
- cjxl/djxl version string: JPEG XL decoder v0.10.2 e148959 [AVX2,SSE2] | djxl does not output extra channels for png | https://api.github.com/repos/libjxl/libjxl/issues/3407/comments | 2 | 2024-03-10T22:09:46Z | 2025-04-30T18:07:00Z | https://github.com/libjxl/libjxl/issues/3407 | 2,177,927,260 | 3,407 |
[
"libjxl",
"libjxl"
] | For example this image from the QOI testset:
```
OXO_emulated_portion.png
Encoding kPixels Bytes BPP E MP/s D MP/s Max norm SSIMULACRA2 PSNR pnorm BPP*pnorm QABPP Bugs
----------------------------------------------------------------------------------------------------------------------------------------
jxl:d0 51 956 0.1471250 3.328 23.892 681.09784854 -9209.858444 -11.95 388.13174541 57.103891443587 100.206 0
jxl 51 14907 2.2941346 2.702 28.994 681.19316895 -9209.641460 -11.85 388.14985862 890.46802876689 1562.74 0
png 51 1500 0.2308447 44.937 187.411 954.11460537 -4701.709204 -10.92 543.61281000 125.49013562060 220.252 0
Aggregate: 51 2775 0.4271120 7.394 50.635 762.12600173 0.00000000 0.00 434.26626476 185.48033643120 325.513 0
```
Converting it to a 1-bit pgm (`convert OXO_emulated_portion.png OXO_emulated_portion.png.pgm`) and using that as input gives even weirder results; looks like the jxl is only an image header here?
```
OXO_emulated_portion.png.pgm
Encoding kPixels Bytes BPP E MP/s D MP/s Max norm SSIMULACRA2 PSNR pnorm BPP*pnorm QABPP Bugs
----------------------------------------------------------------------------------------------------------------------------------------
jxl:d0 51 22 0.0033857 6.083 41.520 365.31238053 -1297.587367 36.53 208.73440368 0.706716715984 1.237 0
jxl 51 413 0.0635592 5.733 60.241 364.90230844 -1283.162731 36.42 208.49507614 13.251788691794 23.193 0
png 51 1508 0.2320759 37.703 215.548 nan 100.00000000 99.99 0.00000000 0.000000000000 0.232 0
Aggregate: 51 239 0.0368259 10.955 81.389 365.10728692 100.00000000 51.05 208.61470559 3.060271325418 1.881 0
```
| benchmark_xl does something very wrong with 1-bit input images | https://api.github.com/repos/libjxl/libjxl/issues/3406/comments | 0 | 2024-03-09T21:56:56Z | 2024-03-11T14:50:55Z | https://github.com/libjxl/libjxl/issues/3406 | 2,177,417,973 | 3,406 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Valid PFM files cannot be opened by the streaming encoder on Windows.
**To Reproduce**
Download: https://drive.google.com/file/d/17Sldh7liNN4FZDcjRnbncNlb7ZBbNKsd/view?usp=sharing
Run: `cjxl sun.pfm sun.jxl -d 0.1 -e 5 --modular=1 --container=1 --streaming_input --streaming_output -x color_space=RGB_D65_SRG_Per_Lin --intensity_target=65000`
**Expected behavior**
streaming input should work (since the encoding works without `--streaming_input`
**Screenshots**
````
cjxl G:\Temp\compression\sun.pfm "G:\Temp\compression\sun_d0.1_e5_i65k.jxl" -d 0.1 -e 5 --modular=1 --container=1 --streaming_input --streaming_output -x color_space=RGB_D65_SRG_Per_Lin --intensity_target=65000
JPEG XL encoder v0.10.2 e148959 [AVX2,SSE2]
PNM decoding failed.
````
**Environment**
Environment
OS: Win11
CPU type: IAMD Ryzen 7 6800HS Creator Edition x86_64
cjxl v10.0.2 from releases
| Stream encoder fails on PFM | https://api.github.com/repos/libjxl/libjxl/issues/3403/comments | 0 | 2024-03-08T14:54:57Z | 2024-03-11T11:03:07Z | https://github.com/libjxl/libjxl/issues/3403 | 2,176,200,876 | 3,403 |
[
"libjxl",
"libjxl"
] | @eustas, @szabadka , @mo271
**Describe the bug**
-d 0.0 -e 10 output file is larger then -e 9
**To Reproduce**
compress attached file with -d 0.0 -e 9 and -e 10 flags
**Expected behavior**
with -e 10 flag final file shouldn`t be larger then using -e 9
**Screenshots**
**Environment**
- OS: win11
- CPU type: x64
- cjxl/djxl version string: v0.10.2 e148959 [AVX2,SSE2]
- sample:
-
| -d 0.0 -e 10 output file is larger then -e 9 | https://api.github.com/repos/libjxl/libjxl/issues/3401/comments | 1 | 2024-03-08T13:43:51Z | 2024-04-20T13:31:02Z | https://github.com/libjxl/libjxl/issues/3401 | 2,176,072,176 | 3,401 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Truncat PPM files succeed when using cjxl (v 10.0.1) with streaming mode but resulting jxl is invalid (according to djxl).
**To Reproduce**
Download and extract: https://drive.google.com/file/d/1oHlYjxcGrKNEV0e5pCB3NMP-lWUrm-er/view?usp=drive_link
Remove the last 99483648 bytes
Endcode: `cjxl G:\Temp\8kx170k_.ppm G:\Temp\8kx170k.jxl -d 0.4 -e 5 --intensity_target=64000 --streaming_input --streaming_output --container=1`
`JPEG XL encoder v0.10.1 5f67ebc [AVX2,SSE2]`
`Encoding [Container | VarDCT, d0.400, effort: 5]`
Decode: `djxl G:\Temp\8kx170k.jxl G:\Temp\8kx170k_decode.ppm`
`JPEG XL decoder v0.10.1 5f67ebc [AVX2,SSE2]`
`Failed to decode image`
`DecompressJxlToPackedPixelFile failed`
Resulting jxl: https://drive.google.com/file/d/18ubwZ_BBpXmX-JoRK8mvgKT9A6KrxauS/view?usp=sharing
```
jxlinfo G:\Temp\8kx170k.jxl
JPEG XL file format container (ISO/IEC 18181-2)
unknown box: type: "" size: 0
Decoder error
Error reading file: G:\Temp\8kx170k.jxl
```
**Expected behavior**
cjxl should not create an invalid jxl without warning
**Environment**
OS: Win11
CPU type: IAMD Ryzen 7 6800HS Creator Edition x86_64
cjxl v10.0.1 from releases
| Stream encoder: Truncated PPM not properly handeled | https://api.github.com/repos/libjxl/libjxl/issues/3400/comments | 1 | 2024-03-08T11:55:58Z | 2024-03-11T11:16:35Z | https://github.com/libjxl/libjxl/issues/3400 | 2,175,894,539 | 3,400 |
[
"libjxl",
"libjxl"
] | Example of large (200MB) JPEG - https://commons.wikimedia.org/wiki/File:Van_Gogh_-_Starry_Night_-_Google_Art_Project.jpg
A simple lossless (accepting all defaults) encode produces a JXL that cannot be opened - XnView ('File can't be loaded), GIMP ('Opening 'D:\jxl-x64-windows-static\bbb.jxl' failed: Unexpected event 1 instead of JXL_DEC_NEED_IMAGE_OUT_BUFFER')
https://drive.google.com/file/d/1uVqbDiXIE_J8A4NnrmXiaUHwrpp_bmz7/view?usp=sharing
Same result when outputting a lossy conversion and seeing the same with other, large (200MB+??) JPEGs.
Windows 11 beta, 5800X3D, 16GB. | (v0.10.2) Large JPEG encode creates invalid JXL with no error warning | https://api.github.com/repos/libjxl/libjxl/issues/3398/comments | 6 | 2024-03-08T10:10:36Z | 2024-03-18T16:04:01Z | https://github.com/libjxl/libjxl/issues/3398 | 2,175,705,305 | 3,398 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
A clear and concise description of what the bug is.
`jxl_extras_codec` is a static library, but the codes [`add_library(jxl_extras_codec`](https://github.com/libjxl/libjxl/blob/491ee5ab4f17d35ec92a5e38ae2098683cf24ffa/lib/jxl_extras.cmake#L154) don't have `STATIC` property, so it will not create lib file when build dynamic, it just create dll file. This cause build tools failed, the [tools](https://github.com/libjxl/libjxl/blob/491ee5ab4f17d35ec92a5e38ae2098683cf24ffa/tools/CMakeLists.txt#L135-L139) need link library `jxl_extras_codec`.
**To Reproduce**
Steps to reproduce the behavior:
Open option `JPEGXL_ENABLE_TOOLS` and build `libjxl`.
**Expected behavior**
A clear and concise description of what you expected to happen.
Build succeed.
**Screenshots**
If applicable, add screenshots or example input/output images to help explain your problem.
**Environment**
- OS: [e.g. Windows]
- Compiler version: [e.g. clang 11.0.1]
- CPU type: [e.g. x86_64]
- cjxl/djxl version string: [e.g. cjxl [v0.3.7 | SIMD supported: SSE4,Scalar]]
**Additional context**
Add any other context about the problem here.
The error log:
```
cmd.exe /C "cd . && "F:\Program Files\CMake\bin\cmake.exe" -E vs_link_exe --intdir=tools\CMakeFiles\djxl.dir --rc=C:\PROGRA~2\WI3CF2~1\10\bin\100226~1.0\x64\rc.exe --mt=C:\PROGRA~2\WI3CF2~1\10\bin\100226~1.0\x64\mt.exe --manifests -- C:\PROGRA~1\MICROS~3\2022\ENTERP~1\VC\Tools\MSVC\1438~1.331\bin\Hostx64\x64\link.exe tools\CMakeFiles\djxl.dir\djxl_main.cc.obj /out:tools\djxl.exe /implib:tools\djxl.lib /pdb:tools\djxl.pdb /version:0.0 /machine:x64 /nologo /debug /INCREMENTAL /subsystem:console lib\jxl_extras_codec.lib lib\jxl_threads.lib tools\jxl_tool.lib lib\jxl.lib lib\jxl_cms.lib F:\0218\installed\x64-windows\debug\lib\hwy.lib kernel32.lib user32.lib gdi32.lib winspool.lib shell32.lib ole32.lib oleaut32.lib uuid.lib comdlg32.lib advapi32.lib && cd ."
FAILED: tools/djxl.exe
cmd.exe /C "cd . && "F:\Program Files\CMake\bin\cmake.exe" -E vs_link_exe --intdir=tools\CMakeFiles\djxl.dir --rc=C:\PROGRA~2\WI3CF2~1\10\bin\100226~1.0\x64\rc.exe --mt=C:\PROGRA~2\WI3CF2~1\10\bin\100226~1.0\x64\mt.exe --manifests -- C:\PROGRA~1\MICROS~3\2022\ENTERP~1\VC\Tools\MSVC\1438~1.331\bin\Hostx64\x64\link.exe tools\CMakeFiles\djxl.dir\djxl_main.cc.obj /out:tools\djxl.exe /implib:tools\djxl.lib /pdb:tools\djxl.pdb /version:0.0 /machine:x64 /nologo /debug /INCREMENTAL /subsystem:console lib\jxl_extras_codec.lib lib\jxl_threads.lib tools\jxl_tool.lib lib\jxl.lib lib\jxl_cms.lib F:\0218\installed\x64-windows\debug\lib\hwy.lib kernel32.lib user32.lib gdi32.lib winspool.lib shell32.lib ole32.lib oleaut32.lib uuid.lib comdlg32.lib advapi32.lib && cd ."
LINK Pass 1: command "C:\PROGRA~1\MICROS~3\2022\ENTERP~1\VC\Tools\MSVC\1438~1.331\bin\Hostx64\x64\link.exe tools\CMakeFiles\djxl.dir\djxl_main.cc.obj /out:tools\djxl.exe /implib:tools\djxl.lib /pdb:tools\djxl.pdb /version:0.0 /machine:x64 /nologo /debug /INCREMENTAL /subsystem:console lib\jxl_extras_codec.lib lib\jxl_threads.lib tools\jxl_tool.lib lib\jxl.lib lib\jxl_cms.lib F:\0218\installed\x64-windows\debug\lib\hwy.lib kernel32.lib user32.lib gdi32.lib winspool.lib shell32.lib ole32.lib oleaut32.lib uuid.lib comdlg32.lib advapi32.lib /MANIFEST /MANIFESTFILE:tools\CMakeFiles\djxl.dir/intermediate.manifest tools\CMakeFiles\djxl.dir/manifest.res" failed (exit code 1104) with the following output:
LINK : fatal error LNK1104: cannot open file 'lib\jxl_extras_codec.lib'
```
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| Should add STATIC property for jxl_extras_codec | https://api.github.com/repos/libjxl/libjxl/issues/3389/comments | 2 | 2024-03-07T10:17:08Z | 2025-03-14T19:42:40Z | https://github.com/libjxl/libjxl/issues/3389 | 2,173,502,061 | 3,389 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Hi, I help maintain packages in [conda-forge](https://conda-forge.org/), a cross-platform distribution used in many places (e.g. scientific computing). We generally try to support our packages - wherever feasible - on `linux-64`, `linux-aarch64`, `linux-ppc64le`, `osx-64`, `osx-arm64`, `win-64` (and `win-arm64` hopefully not too far away).
While trying to [enable](https://github.com/conda-forge/libjxl-split-feedstock/pull/21) ppc-builds in conda-forge I'm seeing that our cmake-based [build script](https://github.com/conda-forge/libjxl-split-feedstock/blob/main/recipe/build.sh) of libjxl manages to force GCC 12 & 13 into an internal compiler error (ICE)
```
[ 44%] Building CXX object lib/CMakeFiles/jxl_enc-obj.dir/jxl/enc_ac_strategy.cc.o
during RTL pass: expand
In file included from $SRC_DIR/lib/jxl/dct-inl.h:21,
from $SRC_DIR/lib/jxl/dec_transforms-inl.h:19,
from $SRC_DIR/lib/jxl/dec_group.cc:31,
from $PREFIX/include/hwy/foreach_target.h:266,
from $SRC_DIR/lib/jxl/dec_group.cc:19:
$SRC_DIR/lib/jxl/transpose-inl.h: In function 'void jxl::N_PPC10::{anonymous}::NoInlineWrapper(const T&, const Args& ...) [with T = void(const DCTFrom&, const DCTTo&, long unsigned int, float*); Args = {DCTFrom, DCTTo, long unsigned int, float* __restrict__}]':
$SRC_DIR/lib/jxl/transpose-inl.h:40:11: internal compiler error: in rs6000_sibcall_aix, at config/rs6000/rs6000.cc:25823
40 | return f(args...);
| ~^~~~~~~~~
0x6947a1 rs6000_sibcall_aix(rtx_def*, rtx_def*, rtx_def*, rtx_def*)
../../gcc/config/rs6000/rs6000.cc:25823
0x14b393c gen_sibcall(rtx_def*, rtx_def*, rtx_def*)
../../gcc/config/rs6000/rs6000.md:11691
0x108ca85 target_gen_sibcall
../../gcc/config/rs6000/rs6000.md:13394
0x959d10 emit_call_1
../../gcc/calls.cc:472
0x9601ad expand_call(tree_node*, rtx_def*, int)
../../gcc/calls.cc:3652
0xa705ce expand_expr_real_1(tree_node*, rtx_def*, machine_mode, expand_modifier, rtx_def**, bool)
../../gcc/expr.cc:11868
0x96e246 expand_expr(tree_node*, rtx_def*, machine_mode, expand_modifier)
../../gcc/expr.h:310
0x96e246 expand_call_stmt
../../gcc/cfgexpand.cc:2831
0x96e246 expand_gimple_stmt_1
../../gcc/cfgexpand.cc:3880
0x96e246 expand_gimple_stmt
../../gcc/cfgexpand.cc:4044
0x973247 expand_gimple_tailcall
../../gcc/cfgexpand.cc:4090
0x973247 expand_gimple_basic_block
../../gcc/cfgexpand.cc:6080
0x974866 execute
../../gcc/cfgexpand.cc:6841
Please submit a full bug report, with preprocessed source (by using -freport-bug).
Please include the complete backtrace with any bug report.
See <https://github.com/conda-forge/ctng-compilers-feedstock/issues/new/choose> for instructions.
make[2]: *** [lib/CMakeFiles/jxl_dec-obj.dir/build.make:342: lib/CMakeFiles/jxl_dec-obj.dir/jxl/dec_group.cc.o] Error 1
make[1]: *** [CMakeFiles/Makefile2:302: lib/CMakeFiles/jxl_dec-obj.dir/all] Error 2
```
**To Reproduce**
Run conda-forge recipe build; I'm happy to provide instructions how to do this locally, or to test patches in our infrastructure
**Expected behavior**
At first I wasn't sure if ppc is even considered supported, but given previous issues here on the bugtracker, I get the impression that this is the case. In this case, I assume it's of interest to the project to fix this.
**Environment**
- OS: linux
- Compiler version: gcc 12.3.0 & 13.2.0
- CPU type: ppc64le
**Additional context**
I'm slightly perplexed to see `rs6000_sibcall_aix` in the stacktrace. AIX and PPC are related but different things. AIX is an OS, ppc is an architecture, though both trace back to IBM. Is it conceivable that some wrong conclusion is being drawn somewhere in config/code?
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| Potential bug? ICE with GCC 12 & 13 on ppc64le | https://api.github.com/repos/libjxl/libjxl/issues/3387/comments | 3 | 2024-03-06T21:38:57Z | 2024-10-02T19:52:06Z | https://github.com/libjxl/libjxl/issues/3387 | 2,172,479,971 | 3,387 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Using the libjpeg.so replacement built from this repo, loading images from libvips always fails with the error "No SOS marker found."
**To Reproduce**
Steps to reproduce the behavior:
```sh
$ LD_LIBRARY_PATH=$PATH_TO_NEW_LIBJPEG vipsheader $IMAGE.jpg
vipsheader: VipsJpeg: ./lib/jpegli/decode.cc:742: No SOS marker found.
```
So far I've gotten the same error on every image I've tried. Using non-libvips programs (e.g. imagemagick) works fine. Encoding jpegs from libvips also appears to work fine.
**Environment**
- OS: Linux
- Compiler version: gcc 13.2.1
- CPU type: x86_64
- cjxl/djxl version string: v0.10.1 461cc01e [AVX2,SSE4,SSE2]
**Additional context**
Related: [libvips feature request](https://github.com/libvips/libvips/issues/3872), but not sure if anyone has actually given it a try yet.
I'd normally consider this a libvips bug instead, but if the goal of this libjpeg is to be a drop-in replacement, then this probably ought to work, right? :) | libvips + libjpeg loading fails: "No SOS marker found." | https://api.github.com/repos/libjxl/libjxl/issues/3378/comments | 0 | 2024-03-05T12:36:46Z | 2024-03-07T15:34:57Z | https://github.com/libjxl/libjxl/issues/3378 | 2,169,108,225 | 3,378 |
[
"libjxl",
"libjxl"
] | After Update to extra/libjxl 0.10.1-1 telegram-desktop will not work correctly. A Downgrade to libjxl-0.9.2-1
all works fine | After Update to extra/libjxl 0.10.1-1 extra/telegram-desktop 4.15.0-1 won't work | https://api.github.com/repos/libjxl/libjxl/issues/3372/comments | 1 | 2024-03-05T08:53:46Z | 2024-03-05T13:33:38Z | https://github.com/libjxl/libjxl/issues/3372 | 2,168,648,508 | 3,372 |
[
"libjxl",
"libjxl"
] | After Update to extra/libjxl 0.10.1-1 mpv will not work correctly. A Downgrade to libjxl-0.9.2-1
all works fine | extra/mpv 1:0.37.0-2 won't work after update to extra/libjxl 0.10.1-1 | https://api.github.com/repos/libjxl/libjxl/issues/3371/comments | 1 | 2024-03-05T08:52:17Z | 2024-03-05T13:33:38Z | https://github.com/libjxl/libjxl/issues/3371 | 2,168,645,795 | 3,371 |
[
"libjxl",
"libjxl"
] | After Update to extra/libjxl 0.10.1-1 calibre will not work correctly. A Downgrade to libjxl-0.9.2-1
all works fine | After Update to extra/libjxl 0.10.1-1 calibre won't work (extra/calibre 7.6.0-1) | https://api.github.com/repos/libjxl/libjxl/issues/3370/comments | 3 | 2024-03-05T08:50:10Z | 2024-03-06T13:57:44Z | https://github.com/libjxl/libjxl/issues/3370 | 2,168,642,127 | 3,370 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Under certain Conditions an undecodable Image is produced with modular jxl.
[bugreport.tar.gz](https://github.com/libjxl/libjxl/files/14482752/bugreport.tar.gz)
**To Reproduce**
(intended to be reproduced on an debian system)
Uncompress the file bugreport.tar.gz
cd bugreport
bash bugreport.sh
**Expected behavior**
Image can be decoded
**Screenshots**
Output on my system see bugreport.log.
**Environment**
- OS: [Windows,Linux]
- (tried with static linux version from github)
- CPU type: [x86_64]
- cjxl/djxl version string: cjxl v0.10.1 5f67ebc [AVX2,SSE4,SSE2]
**Additional context**
Problem is since V0.10.x. Problem only happens with special quality settings.
| Image encoded with modular jxl cannot be decoded | https://api.github.com/repos/libjxl/libjxl/issues/3362/comments | 0 | 2024-03-04T14:38:53Z | 2024-03-04T18:12:28Z | https://github.com/libjxl/libjxl/issues/3362 | 2,167,006,605 | 3,362 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Was reading about the fast decode times of jpegli on Cloudinary and wanted to test, if it has a positive impact on browser rendering times in comparison to mozjpg.
Installed Jpeg-XL via Homebrew without any errors and ran the cjpegli command.
zsh: command not found: cjpegli
**To Reproduce**
- Have Homebrew installed
- Open Terminal
- Run "brew install jpeg-xl"
- Run "cjpegli input.jpg output.jpg"
**Expected behavior**
After a successful install through homebrew, I was expecting the cjpegli command to be available.
**Screenshots**
**Environment**
- OS: MacOs Monterey 12.7.3
- CPU type: Intel Core i7
- cjxl/djxl version string: cjxl v0.10.1 0.10.1 [AVX2,SSE4,SSE2]
**Additional context**
Not sure if it has anything to do with it, but on line 62 the brew formula has the following option
"-DJPEGXL_ENABLE_JPEGLI=OFF",
https://github.com/Homebrew/homebrew-core/blob/6b90bd87b55c14dd8cc04829be39f41e59b90b78/Formula/j/jpeg-xl.rb
Other people experiencing the same problem:
https://news.ycombinator.com/item?id=39568890
https://github.com/orgs/Homebrew/discussions/5074
| cjpegli not included in Homebrew ? | https://api.github.com/repos/libjxl/libjxl/issues/3360/comments | 3 | 2024-03-04T12:13:23Z | 2024-03-22T17:47:53Z | https://github.com/libjxl/libjxl/issues/3360 | 2,166,700,698 | 3,360 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Just what the title described, the JXL file generated by the latest v0.10..1 cannot be opened by Image Viewer (eog) in GNOME 42.5 on Ubuntu-based PopOS 22.04. All five DEB packages contained in the downloaded tar.gz were installed correctly. Actually, all those JXL files generated by v0.9.x that used to work with Image Viewer now just stop working (interestingly, GIMP can still open the file). So I guess there's probably something wrong with libjxl-gdk-pixbuf.
**To Reproduce**
Just download the official tar.gz file for Ubuntu 22.04 from GitHub
**Expected behavior**
Image Viewer (eog) in GNOME should be able to open and view JXL file correctly, as all previous versions of libjxl do.
**Screenshots**
**Environment**
- OS: Pop_OS (Ubuntu-based) 22.04
- CPU type: x86_64
- cjxl/djxl version v0.10.1
**Additional context**
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| JXL file generated by the latest v0.10..1 cannot be opened by Image Viewer in GNOME 42.5 on Ubuntu-based PopOS 22.04 | https://api.github.com/repos/libjxl/libjxl/issues/3358/comments | 10 | 2024-03-04T05:00:39Z | 2024-03-05T16:21:04Z | https://github.com/libjxl/libjxl/issues/3358 | 2,165,937,428 | 3,358 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Lossless encode/decode of [qoi_benchmark_suite/images/textures_pk/pkw_wall02bf.png](https://qoiformat.org/benchmark/) is not lossless
**To Reproduce**
See encoder/decoder code at https://github.com/sandflow/libench/blob/ff5ec6959ee27b24a5185cc78c972ccc52f07a16/src/main/cpp/jxl_codec.cpp
**Expected behavior**
Lossless encode/decode should be lossless.
**Environment**
- OS: Ubuntu 20
- Compiler version: gcc (Ubuntu 9.4.0-1ubuntu1~20.04.1) 9.4.0
- CPU type: x86_64
- cjxl/djxl version string: [libjxl @ 5f67ebc](https://github.com/libjxl/libjxl/tree/5f67ebced96df5c0696ec7d1822bdd3e4e11a1a1)
| Lossless encode/decode is not lossless | https://api.github.com/repos/libjxl/libjxl/issues/3357/comments | 6 | 2024-03-04T04:25:01Z | 2024-03-04T13:41:47Z | https://github.com/libjxl/libjxl/issues/3357 | 2,165,893,272 | 3,357 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
cjxl lossless -e8 and -e9 sometimes create an invalid image according to the spec.
**To Reproduce**
Encode this image using `-d 0 -e 8`: [1225036754354229250_1](https://github.com/libjxl/libjxl/assets/2322996/6e607820-542f-443f-ae2a-5d6c4db39b5e). Pass group 10 will have a Palette transform with 557 colors with LZ77 enabled. libjxl uses `dist_multiplier = 557` to encode and decode, which doesn't match the spec (18181-1).
**Expected behavior**
LZ77 `dist_multiplier` should be 256.
**Environment**
cjxl version: `cjxl v0.10.0 6d041617 [AVX2,SSE4,SSE2]`
**Additional context**
First reported in tirr-c/jxl-oxide#268.
| cjxl lossless -e8 and -e9 sometimes create an invalid image | https://api.github.com/repos/libjxl/libjxl/issues/3356/comments | 5 | 2024-03-03T10:06:05Z | 2024-09-16T06:38:59Z | https://github.com/libjxl/libjxl/issues/3356 | 2,165,284,325 | 3,356 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
cjpegli prints `jpegli encoding failed` and quits when trying to encode a particular file.
This file is a PNG with grayscale colors.
**To Reproduce**
Encode the following image with `--xyb` option:
**Expected behavior**
The image is encoded and `cjpegli` does not halt with an error message.
**Screenshots**
```
PS C:\Users\Firepal\Pictures> cjpegli -d 1 --xyb c.png c.jpg
Read 150x150 image, 5853 bytes.
Encoding [XYB d1.000 AQ p2 OPT]
jpegli encoding failed
```
**Environment**
- OS: Windows 10
- Compiler version: (0.10.1 build downloaded from Github)
- CPU type: x64
<!--
Currently github does not allow uploading files that end in `.jxl`, but when you
rename them for example as `image.jxl.jpg`, it will be possible to upload them
and also view them in browsers that are configured to support it.
See https://github.com/orgs/github-community/discussions/18139
-->
| jpegli XYB encode fails on grayscale image | https://api.github.com/repos/libjxl/libjxl/issues/3355/comments | 4 | 2024-03-02T18:46:56Z | 2024-03-08T15:38:48Z | https://github.com/libjxl/libjxl/issues/3355 | 2,164,925,833 | 3,355 |
[
"libjxl",
"libjxl"
] | **Is your feature request related to a problem? Please describe.**
I'm confused about the difference between "Effort" parameters and "Speed" parameters.
**Describe the solution you'd like**
- Explanation what the value for the parameter mean.
- What the parameter do?
- What is the difference between using "Effort" parameters and "Speed" Parameters? | Clarification on "Speed" parameters. | https://api.github.com/repos/libjxl/libjxl/issues/3354/comments | 4 | 2024-03-02T17:24:11Z | 2024-03-03T10:13:38Z | https://github.com/libjxl/libjxl/issues/3354 | 2,164,898,014 | 3,354 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
When testing a set of a few thousand images, this ones fails to be converted by cjxl.
it does not seem to be an issue with the PNG file itself, as re-saving it with different programs or running it through `oxipng` does not change the outcome.
**To Reproduce**
Obtain the image file [1337453268931244032_1.zip](https://github.com/libjxl/libjxl/files/14468228/1337453268931244032_1.zip)
Attempt to compress it with cjxl, neither lossless or lossy work.
**Expected behavior**
A JPEG-XL file of the image is created.
**Environment**
- OS: Linux
- Compiler version: gcc version 13.2.1 20230801 (GCC)
- CPU type: x86_64
- cjxl/djxl version string: JPEG XL encoder v0.10.1 5f67ebce [AVX2,SSE4,SSE2]
| Attempt to compress png wit cjxl results in error "Illegal instruction (core dumped)" | https://api.github.com/repos/libjxl/libjxl/issues/3353/comments | 5 | 2024-03-02T06:07:52Z | 2024-03-04T14:22:03Z | https://github.com/libjxl/libjxl/issues/3353 | 2,164,627,737 | 3,353 |
[
"libjxl",
"libjxl"
] | The readme file doesn't list `.jxl` as a file type that can be read.
https://github.com/libjxl/libjxl/blob/cae805d8b68f973fcac3b4235029b82366f02a8b/README.md?plain=1#L76-L77
Just wondering if `cjxl` can recompress a JXL image. | Can cjxl read JXL files? | https://api.github.com/repos/libjxl/libjxl/issues/3349/comments | 2 | 2024-02-29T20:16:16Z | 2024-03-01T14:04:38Z | https://github.com/libjxl/libjxl/issues/3349 | 2,162,030,284 | 3,349 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Consider an HDR jxl file (attached as .jpg, actually a jxl)
**To Reproduce**
```
$ djxl --num_threads=0 --disable_output --color_space=RGB_D65_SRG_Rel_SRG hdr.jxl
JPEG XL decoder v0.10.0 cae805d8 [AVX2]
./lib/jxl/render_pipeline/stage_cms.cc:47: JXL_ASSERT: xsize == xsize_
Illegal instruction (core dumped)
```
**Expected behavior**
The file should decode as normal. It can also be reproduced by decoding to PNG instead of with `--disable_output`. It can also be reproduced with default threading options, but disabling threading prevents the error from triggering once per thread. It cannot be reproduced without `--color_space=RGB_D65_SRG_Rel_SRG`, i.e. decoding to PQ as it is tagged works as expected.
**Environment**
- OS: Arch Linux x86_64
- Compiler version: GCC 13.2.1
- cjxl/djxl version string: JPEG XL decoder v0.10.0 cae805d8 [AVX2]
**Additional context**
Valgrind Log: [valgrind.log](https://github.com/libjxl/libjxl/files/14452254/valgrind.log)
| assert fails when decoding HDR jxl to RGB_D65_SRG_Rel_SRG | https://api.github.com/repos/libjxl/libjxl/issues/3348/comments | 0 | 2024-02-29T18:53:46Z | 2024-03-07T11:12:36Z | https://github.com/libjxl/libjxl/issues/3348 | 2,161,901,773 | 3,348 |
[
"libjxl",
"libjxl"
] | **About me**
Hi. I'm maintaining a Apple project integration (For GUI developer who don't use CMake or bazel, using SPM/CocoaPods/Xcodeproj these tools).
You can on-click to build libjxl on any Apple platforms (even include visionOS), and easy to use as a downstream dependency in GUI App (written in Swift code). It's simple and native for cross-compile (without need hacking with CMake)
https://github.com/SDWebImage/libjxl-Xcode
This is not a real pain, but annoying (because people using real Apple Watch, not simulator for shipping App) and maybe not hard to fix.
**Describe the bug**
Compiler error on platform that does not support C++ thread_local. Need provide a way to disable this.
```
/Applications/Xcode-15.2.0.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang -x c++ -ivfsstatcache /Users/user/Library/Developer/Xcode/DerivedData/SDKStatCaches.noindex/watchsimulator10.2-21S355-964dc38f2d3161e1a057f32e04cadc53.sdkstatcache -target x86_64-apple-watchos2.0-simulator -fmessage-length\=0 -fdiagnostics-show-note-include-stack -fmacro-backtrace-limit\=0 -fno-color-diagnostics -std\=gnu++14 -fmodules -gmodules -fmodules-cache-path\=/Users/user/Library/Developer/Xcode/DerivedData/ModuleCache.noindex -fmodules-prune-interval\=86400 -fmodules-prune-after\=345600 -fbuild-session-file\=/Users/user/Library/Developer/Xcode/DerivedData/ModuleCache.noindex/Session.modulevalidation -fmodules-validate-once-per-build-session -fmodule-name\=libjxl -Wno-trigraphs -fpascal-strings -O0 -fno-common -Wno-missing-field-initializers -Wno-missing-prototypes -Werror\=return-type -Wdocumentation -Wunreachable-code -Wquoted-include-in-framework-header -Werror\=deprecated-objc-isa-usage -Werror\=objc-root-class -Wno-non-virtual-dtor -Wno-overloaded-virtual -Wno-exit-time-destructors -Wno-missing-braces -Wparentheses -Wswitch -Wunused-function -Wno-unused-label -Wno-unused-parameter -Wunused-variable -Wunused-value -Wempty-body -Wuninitialized -Wconditional-uninitialized -Wno-unknown-pragmas -Wno-shadow -Wno-four-char-constants -Wno-conversion -Wconstant-conversion -Wint-conversion -Wbool-conversion -Wenum-conversion -Wno-float-conversion -Wnon-literal-null-conversion -Wobjc-literal-conversion -Wshorten-64-to-32 -Wno-newline-eof -Wno-c++11-extensions -Wno-implicit-fallthrough -DDEBUG\=1 -DDEBUG\=1 -DCMS_NO_REGISTER_KEYWORD\=1 -isysroot /Applications/Xcode-15.2.0.app/Contents/Developer/Platforms/WatchSimulator.platform/Developer/SDKs/WatchSimulator10.2.sdk -fasm-blocks -fstrict-aliasing -Wdeprecated-declarations -Winvalid-offsetof -g -fvisibility-inlines-hidden -Wno-sign-conversion -Winfinite-recursion -Wmove -Wcomma -Wblock-capture-autoreleasing -Wstrict-prototypes -Wrange-loop-analysis -Wno-semicolon-before-method-body -Wunguarded-availability -index-store-path /Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Index.noindex/DataStore -iquote /Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/jxl-generated-files.hmap -I/Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/jxl-own-target-headers.hmap -I/Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/jxl-all-non-framework-target-headers.hmap -ivfsoverlay /Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/all-product-headers.yaml -iquote /Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/jxl-project-headers.hmap -I/Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Products/Debug-watchsimulator/include -I/Users/user/Documents/GitHub/libjxl-Xcode/libjxl -I/Users/user/Documents/GitHub/libjxl-Xcode/libjxl/lib/include -I/Users/user/Documents/GitHub/libjxl-Xcode/generate -I/Users/user/Documents/GitHub/libjxl-Xcode/libjxl/third_party/brotli/c/include -I/Users/user/Documents/GitHub/libjxl-Xcode/libjxl/third_party/highway -I/Users/user/Documents/GitHub/libjxl-Xcode/libjxl/third_party/lcms/include -I/Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/DerivedSources-normal/x86_64 -I/Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/DerivedSources/x86_64 -I/Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/DerivedSources -Wno-shorten-64-to-32 -F/Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Products/Debug-watchsimulator -fnew-alignment\=8 -fno-cxx-exceptions -fno-slp-vectorize -fno-vectorize -MMD -MT dependencies -MF /Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/Objects-normal/x86_64/jxl_cms.d --serialize-diagnostics /Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/Objects-normal/x86_64/jxl_cms.dia -c /Users/user/Documents/GitHub/libjxl-Xcode/libjxl/lib/jxl/cms/jxl_cms.cc -o /Users/user/Library/Developer/Xcode/DerivedData/libjxl-gigldylknrjspbfnhqzpqmrpxzto/Build/Intermediates.noindex/libjxl.build/Debug-watchsimulator/libjxl.build/Objects-normal/x86_64/jxl_cms.o -index-unit-output-path /libjxl.build/Debug-watchsimulator/libjxl.build/Objects-normal/x86_64/jxl_cms.o
/Users/user/Documents/GitHub/libjxl-Xcode/libjxl/lib/jxl/cms/jxl_cms.cc:794:10: error: thread-local storage is not supported for the current target
static thread_local void* context_;
^
1 error generated.
```
**To Reproduce**
1. Use a Mac and Install Xcode 15.2(https://developer.apple.com/xcode/)
2. Clone this repo: https://github.com/SDWebImage/libjxl-Xcode, with git submodule update
3. Checkout to `0.10.0` tag
4. Open `libjxl.xcodeproj`
5. Choose the `Scheme` (See button in screenshot) to `Any watch simulator device`
6. Build using Command + B
7. See the compile error in build log
**Expected behavior**
Should provide a way to disable C++ thread_local usage. Like that `register` keyword under `DCMS_NO_REGISTER_KEYWORD`
**Screenshots**
<img width="326" alt="截屏2024-02-29 01 12 33" src="https://github.com/libjxl/libjxl/assets/6919743/d3ebbdb5-c953-4701-a520-85685f6097dc">
<img width="747" alt="截屏2024-02-29 01 10 37" src="https://github.com/libjxl/libjxl/assets/6919743/09ea9ce5-10c2-499d-b0f5-03c6afc299f1">
**Environment**
- OS: watchOS (target tripple: x86_64-apple-watchos2.0-simulator)
- Compiler version: clang-15 (bundled in Xcode 15.2)
- CPU type: x86_64 (watchOS simulator)
- cjxl/djxl version string: Not about this
**Additional context**
Seems there are many way to workaroud without need of thread_local
Like: https://github.com/realm/realm-core/pull/5341
| Compiler error when platform (etc, Apple watchOS simulator x86_64) does not support C++ thread_local | https://api.github.com/repos/libjxl/libjxl/issues/3345/comments | 3 | 2024-02-28T17:12:03Z | 2024-03-04T10:34:40Z | https://github.com/libjxl/libjxl/issues/3345 | 2,159,439,465 | 3,345 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Lossless encode of a grayscale jpg with a gray ICC profile makes an RGB jxl
**To Reproduce**
Steps to reproduce the behavior:
1. Download [test.jpg](https://github.com/libjxl/libjxl/assets/19401176/093b8aa1-1a7f-4a63-aa02-f4cfc7dab2a6)
2. `cjxl test.jpg --lossless_jpeg=1 test.jxl`
3. `jxlinfo test.jxl -v` returns 8-bit RGB num_color_channels: 3
**Expected behavior**
jxlinfo should return 8-bit Grayscale num_color_channels: 1
**Environment**
- OS: ubuntu 23.04
- Compiler version: clang 15.0.7
- CPU type: x86_64
- cjxl/djxl version string: cjxl v0.10.0 5c75066a [AVX2,SSE4,SSE2]
**Additional context**
- Stripping the icc makes it grayscale with num_color_channels: 1
- `djxl test.jxl test.jpg` creates the correct grayscale jpg
- `djxl test.jxl test.png` creates an RGB png
- libpng warning: profile '1': 'GRAY': Gray color space not permitted on RGB PNG
- libjxl built-in cms doesn't use the icc
| Lossless encode of grayscale JPG with icc profile turns into RGB image | https://api.github.com/repos/libjxl/libjxl/issues/3338/comments | 0 | 2024-02-27T07:56:36Z | 2024-03-04T08:01:05Z | https://github.com/libjxl/libjxl/issues/3338 | 2,155,907,733 | 3,338 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
When running `cjxl -h`, the following is shown:
```
...
-e EFFORT, --effort=EFFORT
Encoder effort setting. Range: 1 .. 9.
...
```
However, the actual range of efforts are 1 through 10.
**To Reproduce**
Run `cjxl -h`
**Expected behavior**
Show the full range of 1 .. 10; 11 does not need to be shown as that requires a separate option just to enable it.
**Screenshots**
N/A
**Environment**
- OS: Linux
- Compiler version: clang version 16.0.6
- CPU type: x86_64
- cjxl/djxl version string: cjxl v0.10.0 [AVX2,SSE4,SSE2]
**Additional context**
For what it's worth, the range is shown correctly if you induce an error by trying to use a value outside the actual accepted range. E.g., running `cjxl foo.png bar.jxl -e 0` results in an error message that states: `Invalid flag value for --effort: Valid range is {1, 2, ..., 10}.` | cjxl help (-h) shows incorrect range for effort levels. | https://api.github.com/repos/libjxl/libjxl/issues/3328/comments | 2 | 2024-02-26T08:52:49Z | 2024-02-26T10:25:49Z | https://github.com/libjxl/libjxl/issues/3328 | 2,153,608,506 | 3,328 |
[
"libjxl",
"libjxl"
] | **Is your feature request related to a problem? Please describe.**
On the release page, for Windows, there is 2 zip file options which is Windows-static and another one is non-static. I'm confused which one should I download and why there is 2 options for Windows?
This is not about x64 or x86.
**Describe the solution you'd like**
An explanation the difference between windows-static and non windows-static zip file.
**Describe alternatives you've considered**
1 zip file for Windows which is either static only or non static.
**Additional context**
| Windows file zip clarification. | https://api.github.com/repos/libjxl/libjxl/issues/3327/comments | 2 | 2024-02-25T10:13:41Z | 2024-02-26T04:07:47Z | https://github.com/libjxl/libjxl/issues/3327 | 2,152,692,690 | 3,327 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Background story: I've been trying to convert a 14 gigapixel (169,984 x 84,992 pixels) 16-Bit sRGB panorama for quite a while. I tried different input formats (png, streaming pfm), versions of cjxl (v0.9.1-x64-windows-static, latest master compiled with clang on windows) and threading (default number of threads, no threading).
Even when assigning sufficient working memory (up to 1 TB), cjxl always crashes. I now found where it is crashing and could crop the input image to a handable size so that the bug still occurs.
**To Reproduce**
Download and unpack: https://drive.google.com/file/d/1oHlYjxcGrKNEV0e5pCB3NMP-lWUrm-er/view?usp=sharing
(CRC32 hash of the image: F8753C7D)
`cjxl 8kx170k_.ppm 8kx170k.jxl -e 5 -d 0.4 --intensity_target=64000 --streaming_output --streaming_input --container=1 --num_threads=0`
**Expected behavior**
cjxl should convert the image or gracefully abort the conversion.
**Screenshots**
Stack trace and causing DC group (top left: x=14336 y=6144, size: 2048x2048; if I generate an image only containing this group, the crash does not occur)
**Environment**
- OS: Win10
- Compiler version: 16.0.5 with Visual Studio
- CPU type: Intel i7-9700 x86_64
- cjxl compiled from commit bf2b7655db56bdec39fa81b150ccde36e654372a
| Access violation in ANSCoder::PutSymbol | https://api.github.com/repos/libjxl/libjxl/issues/3326/comments | 0 | 2024-02-24T15:06:08Z | 2024-02-26T13:38:55Z | https://github.com/libjxl/libjxl/issues/3326 | 2,152,353,740 | 3,326 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
`-E` increases the encoding time if the value of the argument is larger than the number of channels in the image.
As far as I understand, `-E` describes how many channels in the image will be looked at at a time in the input image.
It does not make sens for it to do more work when `-E` exceeds the number of input channels of the image.
The outputs produced with different `-E` options are identical.
**To Reproduce**
```
time cjxl img.png -e 9 -d 0 --patches 1 --disable_output
JPEG XL encoder v0.10.0 bf2b7655 [AVX2]
Encoding [Modular, lossless, effort: 9]
Compressed to 2340.7 kB (2.774 bpp).
3000 x 2250, 0.106 MP/s [0.11, 0.11], 1 reps, 16 threads.
66.56user 0.89system 1:03.93elapsed 105%CPU (0avgtext+0avgdata 587916maxresident)k
0inputs+0outputs (0major+67501minor)pagefaults 0swaps
time cjxl img.png -e 9 -d 0 -E 3 --patches 1 --disable_output
JPEG XL encoder v0.10.0 bf2b7655 [AVX2]
Encoding [Modular, lossless, effort: 9]
Compressed to 2340.7 kB (2.774 bpp).
3000 x 2250, 0.076 MP/s [0.08, 0.08], 1 reps, 16 threads.
91.80user 1.07system 1:29.41elapsed 103%CPU (0avgtext+0avgdata 607396maxresident)k
0inputs+0outputs (0major+74815minor)pagefaults 0swaps
time cjxl img.png -e 9 -d 0 -E 11 --patches 1 --disable_output
JPEG XL encoder v0.10.0 bf2b7655 [AVX2]
Encoding [Modular, lossless, effort: 9]
Compressed to 2340.7 kB (2.774 bpp).
3000 x 2250, 0.040 MP/s [0.04, 0.04], 1 reps, 16 threads.
172.36user 0.86system 2:49.79elapsed 102%CPU (0avgtext+0avgdata 781596maxresident)k
0inputs+0outputs (0major+98965minor)pagefaults 0swaps
```
**Expected behavior**
`-E` should be capped to the number of channels in the input image so that it takes the same amount of time if it's larger
**Screenshots**
grayscale image, 1 channel:
**Environment**
- OS: Arch Linux
- Compiler version: clang version 16.0.6
- CPU type: x86_64 Ryzen 2700x
- cjxl/djxl version string: cjxl v0.10.0 bf2b7655 [AVX2]
| cap `-E`/`--modular_nb_prev_channels` to the number of channels of the image | https://api.github.com/repos/libjxl/libjxl/issues/3324/comments | 4 | 2024-02-23T23:14:20Z | 2024-02-26T18:14:39Z | https://github.com/libjxl/libjxl/issues/3324 | 2,151,938,540 | 3,324 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
`-e 9` is slower than `-e 10` with `--num_threads 0` with lossless encoding
**To Reproduce**
```
hyperfine 'cjxl -e 9 -d 0 --num_threads 0 img.png /dev/null' 'cjxl -e 10 -d 0 --num_threads 0 img.png /dev/null'
Benchmark 1: cjxl -e 9 -d 0 --num_threads 0 img.png /dev/null
Time (mean ± σ): 36.188 s ± 0.217 s [User: 36.117 s, System: 0.033 s]
Range (min … max): 35.786 s … 36.494 s 10 runs
Benchmark 2: cjxl -e 10 -d 0 --num_threads 0 img.png /dev/null
Time (mean ± σ): 32.581 s ± 0.243 s [User: 32.398 s, System: 0.143 s]
Range (min … max): 32.240 s … 32.983 s 10 runs
Summary
cjxl -e 10 -d 0 --num_threads 0 img.png /dev/null ran
1.11 ± 0.01 times faster than cjxl -e 9 -d 0 --num_threads 0 img.png /dev/null
```
**Expected behavior**
`-e 9` should be faster than `-e 10`
**Screenshots**
Input file:
**Environment**
- OS: Arch Linux
- Compiler version: clang 16.0.6
- CPU type: x86_64 Ryzen 2700x
- cjxl/djxl version string: cjxl v0.10.0 b38469de [AVX2]
| Modular encoding, `-e 9` slower than `-e 10` with `--num_threads 0` | https://api.github.com/repos/libjxl/libjxl/issues/3323/comments | 6 | 2024-02-23T14:34:20Z | 2024-02-26T16:57:22Z | https://github.com/libjxl/libjxl/issues/3323 | 2,151,216,970 | 3,323 |
[
"libjxl",
"libjxl"
] | MS-SSIM is a popular metric for the team [AVIF](https://storage.googleapis.com/demos.webmproject.org/webp/cmp/2024_01_25/index.html), could the bit of code necessary to compute it be added to the libjxl repo and the result of the MS-SSIM score be added to the output of a benchmark_xl run ?
This would add a new column, yet gives way to setup adequate comparisons and new arguments to tackle the proponents of video based image formats :) | add MS-SSIM to benchmark_xl | https://api.github.com/repos/libjxl/libjxl/issues/3312/comments | 1 | 2024-02-20T15:47:55Z | 2024-07-19T12:11:17Z | https://github.com/libjxl/libjxl/issues/3312 | 2,144,688,024 | 3,312 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
I downloaded [jxl-linux-x86_64-static-v0.9.2.tar.gz](https://github.com/libjxl/libjxl/releases/download/v0.9.2/jxl-linux-x86_64-static-v0.9.2.tar.gz). When I run cjxl, I get
```
cjxl: error while loading shared libraries: libjxl_extras_codec.so.0.9: cannot open shared object file: No such file or directory
```
**To Reproduce**
Download on a system without that library pre-installed, and try to use cjxl to convert a JPEG to JXL losslessly. It may be possible to produce it other ways, that's the only one I tried.
**Expected behavior**
Statically linked binary is self contained.
**Environment**
- OS: Ubuntu 24 via WSL
- cjxl/djxl version string: `JPEG XL encoder v0.9.2 41b8cda [AVX2,SSE4,SSE2]`
| v0.9.2 static Linux binary is broken | https://api.github.com/repos/libjxl/libjxl/issues/3309/comments | 2 | 2024-02-20T12:46:34Z | 2024-02-21T14:06:53Z | https://github.com/libjxl/libjxl/issues/3309 | 2,144,301,381 | 3,309 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
Trying to use a PPM input file with --streaming_input results in "PNM decoding failed."
```
cjxl.exe Doggo.ppm Doggo.jxl -d 0 --streaming_input --streaming_output
JPEG XL encoder v0.10.0 a070f82 [AVX2,SSE2]
PNM decoding failed.
```
**To Reproduce**
cjxl.exe Doggo.ppm Doggo.jxl -d 0 --streaming_input --streaming_output
**Expected behavior**
The file should be readable
**Environment**
- OS: Windows 10
- CPU type: x86_64
- cjxl/djxl version string: JPEG XL encoder v0.10.0 a070f82 [AVX2,SSE2]
Source file
PPM file made with ffmpeg (With .png appended for upload)
| Streaming input PPM decoding fails on Windows | https://api.github.com/repos/libjxl/libjxl/issues/3303/comments | 4 | 2024-02-19T14:15:29Z | 2024-02-21T06:36:22Z | https://github.com/libjxl/libjxl/issues/3303 | 2,142,473,340 | 3,303 |
[
"libjxl",
"libjxl"
] | **Describe the bug**
I used to be able to compile from the repo without issues, but recently attempting to compile creates binaries which cause a "illegal hardware instruction (core dumped)" crash, indicating that something is compiling with the wrong architecture flags enabled
**To Reproduce**
Steps to reproduce the behavior, how I build the program:
mkdir build && cd build
cmake -G Ninja ..
ninja install
now cjxl will always crash with an illegal hardware instruction
**Environment**
- OS: Arch Linux (kernel 6.7.4)
- Compiler version: g++ 13.2.1 20230801
- CPU type: AMD Ryzen 3950X
- cjxl/djxl version string: v0.10.0 ef9a6fb4 [AVX2] | Illegal Hardware Instruction | https://api.github.com/repos/libjxl/libjxl/issues/3299/comments | 1 | 2024-02-18T18:59:33Z | 2024-02-19T23:15:43Z | https://github.com/libjxl/libjxl/issues/3299 | 2,141,138,569 | 3,299 |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.