=== rsalveti` is now known as rsalveti === deffrag__ is now known as deffrag [06:19] c/ === phh_ is now known as phh === doko_ is now known as doko [13:31] oh, ah ! [13:32] stgraber, while concentrating on getting the framebuffer to work i didnt actually notice that tablet is already booting fine [13:32] * ogra_ is just reading /var/log/dmesg from the SD [13:32] [ 17.361] ABI class: X.Org Video Driver, version 13.0 [13:32] [ 17.361] (EE) open /dev/fb0: No such file or directory [13:32] [ 17.361] (EE) No devices detected. [13:32] [ 17.361] [13:32] HA ! [13:35] ogra_: yay! [13:36] ogra_: what did you end up doing to have it boot? [13:36] damned, but that was an older boot, not the kernel i'm currently trying [13:37] stgraber, dunno, i'm wildly poking kernel config opts mainly to get some kind of console (HDMI, USB serial or LCD) [13:37] so if i can get back to that working kernel i should be able to set up wlan in /e/n/i and use ssh [13:38] ogra_: that'd be great! [13:38] * ogra_ goes back to the defconfig and starts over [13:39] luckily cross building only takes a few seconds :) [13:40] ogra_: yeah, once you get through the initial 30-45min build, rebuilding with different options is really quick :) [13:41] 30-40 min ? [13:42] what kind of desktop do you have ? [13:42] i think building it with modules took under 5 min initially here [13:42] (though i run my development folders in a ramdisk....) [13:43] Hi, i am failing to compile latest ti-omap kernel because of a "implicit function declaration" warning. Should I try and disable this or actually instantiate the function? [13:45] sigh, and now i cant get a single working boot anymore [13:45] fun [13:45] i should have checked /var/log eariler [13:50] ogra_: When u use a ramdisk do u have a local repo to commit the changes or is this unversioned development? [13:51] unversioned === cmagina_ is now known as cmagina [13:51] ok [13:51] i have the git tree on disk though ... and copy it into the ramdisk as needed === alkisg1 is now known as alkisg [14:21] ogra_: so I hear you have something booting? that's awesome :) [14:21] i had [14:21] and i'm not at all sure which kernel that was [14:22] the timestamps on the logs are 4h old [15:18] ogra_: What system are you working on? [15:18] zatab [15:28] sigh, i really would like to know how it booted the last time [15:29] i think i went throuh all combos of u-boot and uImae now [15:29] *uImage [15:32] So, why not just run Ubunto on Android? [15:32] why would i do such insanity ? :) [15:33] * ogra_ doesnt really want to carry a monitor with the tablet all the time to make it run ubuntu :) [15:33] I'd better just not comment. May not bode well for future relations. [15:34] i want to run native ubuntu :) [15:35] Is that even possible on arm still? [15:35] (yea, I went there). [15:35] hmm ? [15:35] sure it is [15:36] we're on a channel called ubuntu-arm, hint hint [15:36] :D [15:37] gildean: If you knew me and my background, you would better understand my reference. [15:39] i was just joking, don't take it too seriously [15:42] * GrueMaster takes nothing seriously on these channels anymore. [15:43] hallo pple! I am trying to compile kernel armel from scratch... cause i need to have CONFIG_AEABI=y and CONFIG_OABI_COMPAT=Y and CONFIG_ARM_THUMB=y (in order to get and older etch ARM(not armel!) chroot working with a needed toolchaing glibc 2.3.2) - Question: Why i fail compiling inside scratchbox with "/scratchbox/compilers/cs2007q3-glibc2.5-arm7/bin/sbox-arm-none-linux-gnueabi-strip: Unable to recognise the format of the input f [15:44] here is the output http://pastebin.com/e0uETWfg ... it seems to do its job, but it stops on 'debbing' the headers [15:44] i doubt anyone in ubuntu land has ever used scratchbox [15:44] well, me [15:44] ;D [15:44] was just about to say the same - why not use the standard toolchain? [15:44] LetoThe2nd: tell me more [15:45] geotec: apt-cache search gnueabi ;) [15:45] right, just a std cross compiler or a plain chroot with qemu-arm-static [15:45] it has to be armel [15:46] geotec: so whats the problem? [15:46] geotec: just check what arm toolchains are in the repo and pick what ever you like. [15:48] interesting.. approach.. [15:51] but i need a toolchaing like emdebian? [15:52] no, why should you? [15:53] the kernel is one of the things you can compile with mostly just the bare compiler (and some accesory tools) [15:53] (plus, emdebian is not really what i would call a "toolchain") [15:54] yeah right. But how do i make a kernel package of it [15:54] ? [15:55] i mean, a debian kernel package [15:55] hmh, easiest is if the source is debianzied. [15:56] the source is debianized [15:56] \o/ for hrw then: [15:56] http://marcin.juszkiewicz.com.pl/2012/03/26/ubuntu-12-04-precise-and-cross-compilation-of-arm-kernels/ [15:56] yeah [15:56] it gots all its rules and staff goes it way to the kernel package .. ok tnx i'll have a look [15:57] we should be more precise on arm architectures.. there are so many differences oabi legacy abi eabi and now hard eabi [15:57] ubuntu only supports one :) [15:59] "one architecture to run them all" :P [16:00] (well, there is technically a rotting armv5 port in the archive nobody uses ... but thats not officially supported) [16:00] LetoThe2nd: thaz my problem, i am working with an arm binary which is linked to 2.3.2 glibc [16:00] but my kernel is armel and has no old eabi supporto [16:00] o_O this is drivin'me crazy [16:01] geotec: i still do not see the problem. enable OABI and recompile the kernel. [16:01] yeah trying to do that, i am working on your tip [16:03] dpkg-buildpackage -b -aarmhf (this armhf toolchaing frightens me .. still gonna give it a try) [16:03] toolchain* sorry for the tuple errors [16:03] well what ubuntu are you running? [16:04] if you're running a hf one, better stick to it. if not, then... not. [16:16] geotec: I'd suggest that if you're trying to make OABI binaries work, there's probably something going fundamentally wrong in your life, and you might want to reexamine the choices that led here. :P [16:17] lol [16:17] geotec: (Also, OABI support will likely drop out of the kernel very soon, we've already torn it out of glibc) [16:19] * ogra_ curses this tablet [16:19] it definitely loads the kernel [16:20] ogra_: shamanic curse? [16:20] (the LCD brightens up if the kernel initializes it) [16:20] but all my kernels dont seem to get thruh to booting a rootfs [16:20] ogra_: I'm guessing you can't pop the back off and slap a JTAG in? [16:21] Pretty hard to debug early kernel failures without some sort of serial love. :/ [16:21] infinity, well, i just know that stgraber broke his display when opening his device .... [16:21] i dont really want to do that, the HW is fragile enough [16:21] Get stgraber to send you his broken one to attach a JTAG to? ;) [16:21] infinity, well, i did it on the ac100 and succeeded [16:21] Does the device enable any serial ports on the host when usb is plugged in? [16:21] its not rocket science, just a matter of luck and time [16:21] infinity, ogra_: you can actually get a serial port on the emmc connector, the problem is that it's the same port you're booting from :) [16:22] GrueMaster: If it did, that would almost certainly be a kernel driver that does that kind of magic, not firmware, so it would be useless for debugging the kernel itself. [16:22] stgraber, yeah, i'm experimenting with g_serial compiled in [16:22] infinity: Blaze had that feature w/o kernel booting. [16:22] stgraber: Oh, that's a special "design". [16:22] sounds like that thing is totally braindead. apart from the fact that they obviously have been forced to admit right on their webpage that hdmi out is disfunctional. [16:22] GrueMaster, yeah, TI is lovely, isnt it :) [16:23] GrueMaster: Blaze was also wildly overengineered for, like, engineers. I've never seen a consumer device nearly as cool. [16:23] infinity: i need to run an old arm binary linked against glibc <=2.3.2 [16:23] infinity, well, the kernel uses some kind of pre DT devicetree feature, you can re-route any HW you like in a textfile :) [16:23] geotec: Yeah, see above about "poor life decisions". :P [16:23] geotec: Seriously, why would one need to do such a thing? [16:24] cauz the arm binary is worth one year work? [16:24] geotec: (But sure, if you do need to, and there's no way around it, recompiling a kernel with OABI flipped on and running an ancient Debian chroot seems "sane") [16:24] sadly the only documentation for this file format i can find is a chinese PDF :) [16:24] though i already found out how to switch displays etc [16:24] infinity: what i am doing actually [16:24] ;) [16:24] geotec: It's a binary you have no source to, and no one knows where it came from? :/ [16:25] (which doesnt help if the kernel doesnt make any use of fbcon) [16:25] i need a kernel with old eabi support to run a etch chroot :D inside an arm device [16:26] geotec: It's not "old eabi", just "old abi". EABI was what replaced OABI. [16:26] (Just nitpicking) [16:26] otherwize i am doomed, what other choice do i have? Kill who came before me for choosing closed binaries from the heck, out of the world? [16:26] old eabi is probably the ganddad of both of them :) [16:26] "never touch a running system"? what kept netware alive? [16:26] geotec: Or find something else that does the job. Or find the people who produced the original binary? [16:27] yeah right. Kinda stole the binary ;s [16:27] not really stole, but almost [16:27] * infinity covers his ears. [16:27] LA LA LA CAN'T HEAR YOU. [16:27] GOING AWAY NOW. [16:28] lol [16:28] * LetoThe2nd hands infinity some iron maiden cds [16:29] anyways, time to leave ;) [16:29] * geotec wonders what was that thing about 'poor life decisions'.. :S [16:34] you know, steal is something very strange to define on the web. Is getting something sincerly offered, you should not take a steal? Is like leaving donuts to indians saying: keep it there for me, i'll be back in a min [16:35] whell i am morally way below indians... i admit [16:37] for loggers: dont worry is just an update to an embedded device, which included this binary. They do not let the package without their hardware. I just grabbed the update and took what i needed [16:39] * geotec wonders where is everyone.. o_O [16:47] oha [16:48] stgraber, ... [16:48] booting off USB key works flawless [16:48] i even have a mouse pinter now [16:49] using my panda USB key [16:49] oh, I never tried usb boot ;) [16:50] me neither, but it was clear the the kernel comes up fine [16:50] For which platform, zatab? [16:50] so it could only be a panic or issues mounting the rootfs [16:50] yes [16:50] ok, so the problem is that booting from emmc is that the kernel can't find the rootfs... [16:50] apparently, i now have a runningg system so i can start inspecing that part :) [16:51] sweet. [16:51] nice! [16:51] * ogra_ goes to find an usb hub so i can attach a kbd [16:52] hmm, and i better create an .xsession file so that it doesnt try to boot into unity all the time :) [16:53] ogra_: once you figure out the right kernel config to get emmc boot, can you try and add that to the script (maybe create a zatab branch on LP with the needed kernel/uboot/script.bin/...)? [16:54] stgraber, sure [16:55] But unity works everywhere, right? [16:55] :P [16:56] bah [16:56] adding a hub doesnt work :/( [16:57] You probably need a usb gadget host cable. [16:57] yes, the device only has micro USB ports [16:57] (read: i'm already using one) [16:59] Kernel probably not configured for it. I know it kept going out of cycle on omap/omap4 images. [17:03] https://plus.google.com/107109423598372241322/posts [17:04] dunno if i have shared it the right way [17:04] * ogra_ isnt such a big g+ user [17:05] ogra_: "Oliver hasn't shared anything with you." [17:05] heck, how do i just publish something on g+ ? [17:05] unrestircted etc [17:06] ogra_: you need to make sure it's shared with "public" [17:06] if i click share i can only select groups or individuals [17:06] ah, found it [17:06] intuitive is something else though [17:06] stgraber, try again [17:07] yay, login screen on a tablet! [17:07] highvoltage: ^ [17:07] right [17:07] no touch yet [17:07] and i cant attach a uhub, i think that draws to much power [17:07] so no mouse/kbd yet [17:08] bah and now DPMS kicked in [17:08] getting it to connect to wifi shouldn't be too difficult [17:08] oh, indeed, i was just wondering why it didnt [17:08] i should probably copy the modules onto the panda image :) [17:08] ajmitch managed to get wifi working pretty easily [17:08] well, i bet he had modules ;) [17:09] yeah :) [17:18] hmm, intresting, doesnt finish booting with the modules in place [17:19] * ogra_ removes them again [17:21] heh [17:21] boots fine [17:22] that's odd [17:23] well, some module or some firmware being evil here [17:25] ah, awesome, my USB 1.1 hub seems to work [17:25] hmm, or not [17:25] seems it goes into constant reboots now [17:26] Maybe it needs some QA? [17:28] no, power [17:28] i found a powered hub, lets see [17:30] yeah, mouse and kbd [17:32] and you can actually use ctrl-alt+t even on non GL systems to get a terminal up ... great [17:33] aha ! [17:33] init: failed to open system console [17:33] all over dmesg [17:38] stgraber, insmodding the wlan driver gives me a hard hang [17:38] :/ [17:38] so that device might have changed between pre-prod and today [17:39] next i'll try the fbcon module [17:42] ugh, these modules are also out of date [17:42] sigh [17:42] no woder wlan hangs it [17:43] stgraber: omg [17:43] yeah, i know, i need to clean my desk [17:43] :) [17:46] heh [17:46] oh cooool ! [17:47] pressing the power button actually gets me a shutdown dialog, nice [17:47] stgraber: I suddenly find it weird that we never considered trying USB [17:47] highvoltage: I don't have usb sticks [17:47] well, i dont think its actually USB vs MMC [17:47] i guess there is really something weird with the modules [17:47] stgraber: yeah but they're $4 across the street :) [17:49] ok,. freshly built modules ... [17:49] * ogra_ starts over [17:52] hmm, hangs again ... [17:52] even though i should have the matching symbols now [17:52] thats wried [17:53] hmm, unbelivable [18:06] ok, so its definitely fbcon that kills it [18:26] arm-linux-ld: unrecognized option '-Wl,-Bsymbolic-functions' .. this is tricky :S [18:36] tried removing the comma , I don't think its suppose to be there [18:38] i think is a bug cauz ld shouldnt be passed -Wl [18:38] gcc is [18:42] hmmm, based off of some googling it should be -Wl,-B,symbolic-functions so needs another comma ;) [19:35] hah, it helps to also disable android network crap isf you want network :P [19:36] silly stuff [19:49] grumble [19:50] so my /e/n/i network config doesnt get picked up and NM doesnt work in lightdm [19:52] still on the zatab ogra_? [19:53] ogra@zatab:~$ uname -a [19:53] Linux zatab 3.0.38+ #29 PREEMPT Thu Sep 20 21:11:35 CEST 2012 armv7l armv7l armv7l GNU/Linux [19:53] ogra@zatab:~$ [19:54] ;) [19:54] got ssh :) [20:06] ogra_: so you're making progress with yours? I'd love to see what's different & work out why mine booted so easily :) [20:12] stgraber, hmm, i thought the a10 was a dual core [20:12] i dont see any second core here [20:14] oh, and i only have 400M [20:14] hmm, no, 325M [20:15] and a constant system load of 3.00 [20:15] uh [20:15] bad plymouth [20:16] ogra_: nope, a10 isn't dual-core, it's a single core Cortex A8 [20:16] wow, its pretty speedy for that [20:16] yeah, it's surprisingly good for an a8 [20:27] Is it remarkably speedier than, say, an mx53 (also a single-core A8)? [20:27] dunno [20:28] i have never touched an mx53 [20:28] Ahh. Well, they're pretty speedy for what they are. :P [20:28] it feels a lot speedier than an XM [20:28] And my guess is that the A10 is just a stock design from ARM, I somehow doubt that, at that pricepoint, they're employing brilliant engineers to tweak it. [20:30] (Or it could be a relicensed part from another ARM licensee, like Freescale, just to confuse matters more) [20:30] infinity, it's not speedier than an MX53.. except it's easy to find an A10 1.2GHz, you have to special order 1.2GHz MX53 [20:30] well, no idea, it comes from deepeds china [20:31] NekoXP: Indeed, getting parts out of Freescale is vaguely like pulling teeth. [20:31] might have stolen designs at fsl or developed something themselves [20:31] it's also definitely not even close to i.MX or OMAP. An ARM reference design with some IP cores they generated themselves seems like the best bet, but a really fast turnaround and very little "optimization" [20:32] as far as the docs I've seen it's about as far from i.MX as you can get, either they randomly changed every register or the IP is just different [20:32] Heh. [20:32] Yeah, I've not looked at one at all. [20:32] ogra@zatab:~$ cat /proc/meminfo |grep MemTotal [20:32] MemTotal: 331520 kB [20:32] thats definitely not right [20:32] Though, I did say "relicensed from some licensee 'like Freescale'", I didn't imply it would have been Freescale. Tons of people design A8-alike parts. [20:33] a lot of IP comes out of China though. There's probably some dirty little video codec supplier around there that gave them a video decoder.. the display unit is a very reasonable and boring design. [20:33] Just seems unlikely that, if pushing for volume and low cost, they'd employ any in-house engineering to reinvent any wheels. [20:33] infinity, its china, why would they relicense, they just take it :) [20:34] ogra_: That only works as long as they don't want to sell into IP-whiney markets (like Europe and North America). [20:34] lol [20:34] ogra_: If their goal was world domination, they have to play by our bizarre rules. [20:34] well, go to fairs in europe and find all the gucci and rolex made in china [20:35] Yeahp, and Gucci tried to shut them down as they pop up. [20:35] didnt they even copy a complete BMW once ? [20:35] Gucci probably makes all their stuff in china anyway [20:35] But most of them are much harder to find than AllWinner, who's made themselves a large target. [20:35] So, here's hoping they're playing nicely. [20:36] yeah, i'm indeed not serious here :) [20:36] it's no wonder all that stuff gets Shanzhai treatment [20:36] anyway, enough tabletting for a day ... [20:36] * ogra_ wonders off to the TV [20:36] Any day with tabletting is a bad day, IMO. [20:36] thanks for doing it ogra_ [20:36] I'm still so not the target market for those things. [20:36] ogra_: so does booting from USB make a difference? [20:37] highvoltage, well, now that i have it, i want to use it :) [20:37] * infinity goes back to arguing with clang. [20:37] it could be that the IP in use is some standard FPGA library stuff. Some of it looks a lot like the things you'd find on opensource websites for Altera or Xilinx FPGAs [20:37] highvoltage, yeah, still cant boot from MMC [20:37] I swear, I develop a new brain bleed every time I dive into this codebase. [20:37] And it boggled the mind every time someone else claims it's "clean and well-written" or, my favourite "has great ARM support". [20:37] ogra_: I didn't think mine was too different (got it from zareason at UDS), as it boots from MMC just fine [20:37] s/boggled/boggles/ [20:38] infinity: their definition of 'great' might be that it compiles sometimes [20:38] ajmitch: yeah there's definitely some difference between the two [20:38] ajmitch, well, might be pre-production [20:38] ajmitch: Yeah, well. There are some things about the llvm architecture that are pretty neat, and if we could go back in time a few decades, GCC could learn some lessons. [20:38] ajmitch: But overall, the project is painful to work with, especially with their haphazard support for multiple architectures. [20:39] And every small mistake llvm makes is magnified about four thousand times by clang's clunky interface on top. [20:39] BUT I'M NOT BITTER. [20:40] llvm as a backend to gcc might actually be easier to support than clang is. [20:41] gcc's architecture has been driven by the fsf philosophy, not always in a good way [20:41] Nope, GCC has a lot of warts, and a lot of things that are almost literally unfixable at this point. [20:41] But llvm's hardly a utopia of software design either. [20:43] you should know by now open source devs just jump from cool eTLA to cool eTLA :-D [20:45] Whatever. Perl 4 lyfe, yo. [20:46] lol [20:46] Anyhow. Let's see if this hacked-up clang actually (A) builds, and (B) produces binaries that look kinda like armhf ones. [20:46] * infinity crosses his fingers. [21:42] infinity: abiword keeps aborting on armel [21:44] micahg: Yes, this is known. Stop giving it back. :P [21:46] infinity: I figured I wasn't running into the wall fast enough :P [21:47] any idea when things might return to normalish? [21:47] Not sure. It's a catch-22 where hacking around this breaks other things. [21:47] And fixing it properly is real effort. [23:39] Has anyone here built the kernel.org kernel using Ubuntu arm-linux-gnueabihf-gcc?