[10:42] asac: Hey [10:42] asac: How are you doing? [10:42] asac: So, let's move the mbf discussion here [10:42] asac: I think the SVG issue might be specific to 1.8 [10:43] So perhaps we should try moving to 1.9 before investigating [10:43] asac: So, I think you said we could either write a xul app or fix the security checks [10:43] asac: Did you hear from upstream about them? [10:44] asac: I think you traced the problem of xul 1.9's mbf to be some security nest clearing the javascript vars before being passed to the html [10:44] lool: we could do a xul application out of this mobile-basic-flash thing. [10:45] not yet sure how to best interface with the hildondesktop API though [10:45] asac: Ok [10:45] lool: how does mbf interact with hildon desktop? [10:45] does it just send "start X", start Y? [10:46] asac: So instead of setting up a gtk_mozembed thing, it would create a xul thing? [10:46] lool: basically yes. [10:46] lool: i have to look at the code [10:46] let me get that [10:47] asac: It basically receives events from being a HD plugin I think, and it triggers launching of app [10:47] Like "app foo starting" [10:47] (from a quick grep) [10:47] asac: http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/m/mobile-basic-flash/mobile-basic-flash_0.44-0ubuntu4.dsc [10:48] asac: Do you think we should start from the moblin (tip) version? [10:48] lool: what kind of events does it need to deal with? [10:49] asac: From hildon? not many I think [10:49] yes from hildon [10:49] Things like X events or home plugins change notifications [10:49] asac: But we could simply keep all of this in the current C code, no? [10:50] asac: Hmm perhaps we should think carefully at which version we port [10:50] is it possible to install ubuntu-mobile on a sony ericsson p1i? [10:50] asac: Currently there are three main trees, 1) moblin 2) UME 3) MSG [10:50] asac: MSG people had to bridge sound events triggerred by actions in their flash movie [10:51] asac: Do you have access to their tree? they changed it a whole lot [10:52] asac: So perhaps the safest is to take tip, and then port to downstreams? [10:53] And moblin's tree is at moblin.org:/home/repos/projects/mobile-basic-flash.git [10:53] lool: i have to think about it [10:53] cant tell whats really the best way to go on [10:54] asac: Ok; what about the SVG issue? Did SVG work fine in 1.8? [10:54] lool: afaik it works in firefox 2 [10:54] asac: Basically MBF calls gtk to get the icon pathname [10:54] so it should work in 1.8 [10:54] Then inserts the HTML snippet in the DOM with an pointing at it (I guess) [10:55] This works for pngs, but with svg it displays once, disappears, displays a second time, disappears and I'm in an infinite loop somewhere [10:55] lool: interesting [10:55] lool: opening that pic in firefox-2 works fine? [10:56] asac: Dunno [10:56] thats certainly a thing you could try [10:56] (Installing it) [10:57] lool: backup your production firefox profile to prevent eventual mess when switching back and forward [10:57] (2.0 <-> 3.0) [10:58] I'm installing it on my Q1 where I don't run firefox [11:00] asac: Works fine from firefox-2 [11:01] But then the HTML might cause the SVG to be resized or something [11:02] the image is loaded with a background-image: css property [11:03] hmm [11:03] might make a difference [11:21] asac: I tried reproducing with [11:22] To no luck yet [11:22] lool: didnt you say that mbf is using It was just a guess [11:23] Looking at the actual html, it uses [11:34] asac: All my test from firefox-2 work fine :-( [11:35] I even have the javascript working and all [11:44] asac: Oh, it doesn't work at all in my firefox 3 here [11:45] http://www.w3schools.com/svg/svg_examples.asp [11:45] those work for me [11:45] asac: SVG works [11:45] asac: I mean the javascript of mbf [11:45] yeah [11:45] It works fine with SVG loading and all in firefox-2 [11:46] thats why its broken with xulrunner 1.9 :) [11:46] (it hangs mbf) [11:46] asac: So in 1.9, it seems desktop_mgr.js is the culprit [11:46] asac: Did you pinpoint where? [11:46] lool: afaict its everly load_url("javascript:...") that is not working [11:47] in the .cpp code [11:47] what does desktop_mgr.js do? [11:47] The ! [11:49] This works [11:49] lool: cool does it fix it? [11:49] I have no idea, I don't have the xul 1.9 build [11:49] If you have a 1.9 xul mbf binary, I could try it out [11:50] asac: http://people.ubuntu.com/~lool/mobile-basic-flash/home.html [11:51] let me check master [11:51] i think the xul 1.9 changes are already committed there [11:51] They are [11:52] * lool grabs xul1.9-dev [11:53] lool: you need to patch configure.ac [11:53] http://paste.ubuntu.com/19856/ [11:53] asac: Ok; git log says they applied your patch [11:53] Right, that looks more like it [11:54] lool: oh ... no all is fine i think [11:54] Ah, I applied your patch [11:55] Ok: reverted; autogen [11:55] lool: good [11:55] git is so cool [11:55] hush [11:55] git clean -df; git reset --hard HEAD :) [11:56] i just used git checkout configure.ac :) [11:56] but resetting is nice ... especially that you can commit -c ORIG_HEAD :) [11:56] asac: It's empty, unfortunately [12:00] asac: Anyway, I think bspencer demonstrated how lower level calls were failing already [12:00] What I spotted was just another bug, later down, but it's not the xul 1.9 blocker :-/ [12:00] lool: right. its the load_url("javascript..." thing that doesnt work [12:01] Yeah [12:03] lool: i would instantly debug this if we could make a plain gtk window out of it and put the mbf widget in it [12:03] so i can drop the whole hildon stuff and reduce the testcase to a developer friendly complexity [12:04] asac: Ok [12:08] lool: maybe we can just make "home_plugin" a GtkWindow? [12:08] That's what I'm looking at [12:08] MobileBasicHomePlugin could be instanciated [12:09] (in _plugin_init i mean) [12:09] Good idea, perhaps you can try that [12:09] hmmm the MobileBasicHomePlugin is a custom widget from what i can tell [12:09] I'm looking at home_screen_changed(), it seems to do the init we want [12:10] ah no ... i am wrong ... i looked at wrong struct. its derived from HildonDesktopHomeItem [12:11] I think we could change home_screen_changed() to use a GtkWindow instead of a GdkWindow [12:12] lool: the mozembed widget is added to home_plugin in plugin_init [12:13] so i think just changing home_screen_changed is not enough [12:22] lool: http://paste.ubuntu.com/19864/ [12:22] thats ugly but it starts a window somehow [12:26] ok you would also require a hacky patch like http://paste.ubuntu.com/19865/ to test in-source html [12:26] asac: I'm doing the same thing ATM :) [12:26] they should honour ATM :) [12:26] ouch [12:26] honour DATADIR [12:26] or something [12:26] http://paste.ubuntu.com/19866/ is what I was doing [12:27] funny how we have basically the same thing [12:27] lool: did you switch to GtkWindow parent class too? [12:27] No, I was doing that [12:28] for me its all black now [12:29] is that the bug? [12:29] asac: i'll have to drop off for lunch now; since you're further than I have with a GtkWindow displayed, could you continue investigating? :) [12:29] asac: Yes that's the bug [12:29] asac: You get any stderr output? [12:29] lool: http://paste.ubuntu.com/19867/ [12:29] export NSPR_LOG_MODULES=all:5 [12:29] will show that [12:30] so its obvious. gtkmozembed doesnt support javascript: in load_url anymore [12:30] asac: Good stuff [12:30] asac: Are you sure it's not the function missing? [12:31] asac: Did you change the ? [12:31] let me check if there is some error before [12:31] Cause that was blocking my firefox (xul 1.9) to render the page I handed you [12:32] lool: it works ;) [12:32] it was the script thing [12:32] asac: Cool [12:32] lool: http://paste.ubuntu.com/19868/ [12:32] Yeah [12:33] but with that patch its not working for you? [12:33] i mean the real build? [12:33] asac: I was a bit hackish: I built from mbolin git and copied the .so over my installed plygin [12:33] Perhaps I misbuilt it [12:33] asac: Could you rebase the patch atop the .dsc I handed you? [12:34] http://ppa.launchpad.net/ubuntu-mobile/ubuntu/pool/main/m/mobile-basic-flash/mobile-basic-flash_0.44-0ubuntu4.dsc [12:34] lool: no ... that thing is not xul 1.9 ready [12:35] we need master from what i can tell [12:35] and the xul 1.9 patch cannot be easily rebased. i would have to redo it i think [12:35] better git merge master on the packaging tree i guess [12:36] let me see how far i get when trying to merge that thing on the hardy branch [12:36] Ok; we don't have a packaging git tree sadly, but I'll try to look into that [12:37] I really need to cook lunch for my wife now, she's complaining that she's hungry [12:40] lool: there is a hardy branch [12:40] why not put all the packaging files in there [12:40] heya [12:40] afaict its just a few files missing [12:40] for instance no debian/changelog ;) [12:40] but debian/ dir exists [12:40] and debian/rules [12:42] ok hardy branch building [12:43] lets see [12:43] lool: hmmm ... you probably cannot use the binaries from amd64? [12:44] lool: pleaes checkout hardy branch and apply http://people.ubuntu.com/~asac/mbf.hardy-to-latest.patch [12:44] that should be the latest [12:44] changelog version might be wrong [12:58] lool: let me know if this works [12:58] ill take a break now [12:59] lool: if that works, I'll just commit it to moblin i guess [13:11] asac: It doesn't apply unfortunately [13:11] asac: What do you mean with hardy branch? [13:11] origin/hardy oh [13:12] I don't have access to this [13:12] asac: it's not in sync with the ppa .dsc sadly [13:12] I guess they created it when they moved packaging out of master [13:13] But we can't ensure that uploaded packages and git branches are in sync unless we have commit access, and we didn't get commit access [13:13] asac: In the ppa, the source is still named .c for instance [13:17] asac: I'm kicking a build of hardy branch + your patch to at least see whether it works [13:18] lool: yes try that branch. i dont see that we can quickly fix 0.44 [13:18] if we know that it works we can do the work if we require a backport for hardy [13:18] but first lets check if it works in theory ;) [13:18] I came to the same conclusion [13:19] lool: i have commit access [13:19] you too? [13:19] checking for MOZEMBED... configure: error: Package requirements (mozilla-gtkmozembed-embedding >= 1.8.1) were not met: [13:19] lool: xulrunner-1.9-dev is required [13:19] for me: [13:19] pkg-config --cflags mozilla-gtkmozembed-embedding [13:19] -DXPCOM_GLUE -fshort-wchar -I/usr/include/xulrunner-1.9/unstable -I/usr/include/nspr [13:20] asac: I'm pbuilding my lpia binaries [13:20] The hardy branch was bdeping on libxul-dev [13:20] lool: ok ... yeah changed build-depends from libxul-dev to xulrunner-1.9-dev [13:20] lool: forgot to include that in the diff i guess [13:20] drwxrwsr-x 8 nobody dev-mbf 4096 Jul 26 2007 mobile-basic-flash.git [13:21] I'm only in group devel, so basically no commit access [13:21] You're in dev-brow dev-ma [13:21] asac: I don't think you have commit access anymore [13:22] oh ok. they have per project permissions [13:22] You have mobile-browser.git and moblin-applets.git [13:22] yeah. what is the "all" group? [13:22] no idea [13:22] ;) [13:22] ill ask bob if he wants me to commit the mobile-basic-flash fixes ;) [13:23] asac: I think it just WORKED ! [13:23] cool [13:23] I'm not sure I believe it [13:23] yay \o/ [13:24] Uh [13:24] libxul0d [13:24] lool: he? [13:24] asac: It's still using 1.8 [13:24] hmm [13:24] lool: you have the link line? [13:24] Oh wait [13:24] hardcoded in control [13:24] Depends: ${shlibs:Depends}, ${misc:Depends}, libxul0d, libwnck22 [13:25] lool: ldd src/mobilebasichome | grep libxu [13:25] nothing [13:25] lool: yeah ... thats just packaging ;) [13:25] * asac fixes that in his git clone [13:25] ldd /usr/lib/hildon-desktop/libmobilebasichome.so | grep xul => empty [13:26] Once I'll have removed the dep, I'll simply wipe libxul0d and see whether it starts :) [13:27] ok i have fixed it [13:27] here in git [13:27] ill ask bob about committing :) [13:27] asac: It's fixed in the ppa's packaging too [13:27] lool: ok cool [13:27] so all is fine? [13:27] (once all is committed upstream?) [13:28] asac: Sure, then Bob needs to release a tarball and not upload it to ppa [13:28] It's going to be very painful to sync the mobile-basic-flash stuff; it never used a patch system and was forked over and over [13:29] Only thing installed is xulrunner-1.9 and it comes up! [13:31] asac: So this is good to go [13:31] lool: what changes do we have for mobile-basic-flash? [13:31] asac: So the only change to complete support was that stupid in the *.html [13:42] asac: I can't change categories [13:42] asac: You have a Q1? [13:42] lool: yeah, but no good keyboard [13:42] so doing things is cumbersome [13:42] asac: You can ssh to it! [13:43] lool, simple enough, our html may not even have that issue [13:43] lool: hehe [13:43] yeah [13:43] pmcgowan: The rest of the required changes are in the moblin git [13:43] pmcgowan: autotools stuff, renaming, etc. [13:43] asac: If you prefer, you can build a VM [13:44] asac: That's easy to do nowadays [13:44] lool: with virtual box? [13:44] asac: https://wiki.ubuntu.com/Testing/Cases/UMEinstall [13:44] asac: Even if virtualbox if you like [13:44] s/if/with [13:44] lool: i wont be able to really work on more things today [13:45] lool: i have to do something and then prepare for a different battle of real life ;) [13:45] pmcgowan: This will probably enhance hildon desktop startup time (it did for me) and will certainly reduce midbrowser startup time [13:45] but I'll be here during weekend [13:45] asac: Ah, sad to read this [13:45] given that you are usually on irc as well, we can certainly continue then [13:45] asac: I'll try to grab bspencer [13:45] I'm not sure I'll have much time this WE [13:45] lool: yeah ... i send the mail, so you can jump in there i guess [13:46] maybe wait if he replies :) [13:46] I kind of neglected preparations for the coming baby, and have plenty of things to do in the last days [13:46] lool: i suggested Co-Maintainer from Ubuntu to commit to moblin branches [13:46] lool: oh. ok. better stay off on the weekend then [13:46] lool: baby is more important that Ubuntu-Baby ;) [13:46] I'll probably be around a little anyway [13:46] s/that/than/ [13:47] I knew that ;) ... how can't you be around the whole weekend? [13:47] Plus, the final is built already :-/ [13:47] Too bad we didn't find out before final [13:48] lool: yeah. lack of time. i asked bob to test in firefox [13:48] that would have given us the hint [13:48] i never received anything back [13:48] asac: Well it's how I found about it [13:49] I tested in firefox-2, in midbrowser; the console errors helped me catch the error [13:49] right ... thats why i am saying that [13:49] I then tried to alert("foo");, and it was obvious [13:49] bad luck ;) ... the ideas were there; just nobody did it :) [13:50] lool: but we decided that its ok to do the migration in a stable update [13:50] asac: Ideally, could you throw your hack for GtkWindow as a patch to mbf? It would allow to -DMAIN_WINDOW or something to get an executable [13:50] asac: Did we decide that? Hmm [13:50] Probably we did [13:50] lool: yes. two weeks ago or so in mobile meeting [13:50] the question came up: "is mbf with xul 1.8 a blocker" [13:50] Okay, I'm sorry my memory is leaky, but I certainly prefer xul 1.9 if it works [13:50] and i answered that i would maintain xul 1.8 security wise for a while and we could later upgrade to 1.9 [13:50] We'll have to iron out why categories don't work anymore [13:51] (but launching does) [13:51] So .cpp -> xul works and vice-versa [13:51] asac: Right, I recall now [13:53] yes [13:53] .cpp should also work with 1.8 [13:53] not sure why we dont have that in .0.44 [13:57] Dunno [13:58] Anyway, first lets get it upstream, fix it for other issues (categories), then merge, then backport :) [14:00] persia: .svg works with xul 1.9 if you were wondering :) [14:00] I'm not too interested in chasing 1.8 to fix svg rendering [14:05] lool, thats good, I will have Bill or Michael contact you later [14:16] pmcgowan: asac mailed bspencer; we need to test the upstream patch fully, it seems it has a small regression [14:17] lool, ok, what was regression? [14:23] pmcgowan: categories not working [14:23] pmcgowan: (marquee menu) [14:23] lool: categories like filtering on menu click? [14:24] lool: you can set [14:24] export NSPR_LOG_MODULES=all:5 [14:24] and NSPR_LOG_FILE=/tmp/output [14:24] and then see what happens if you click the menu [14:24] lool, may not effect our custom build [14:25] lool: or use bspencers instructions to get output from the hildon-desktop plugin stdout/err and just set NSPR_LOG_MODULES [14:25] but i thin the NSPR_LOG_FILE=... thing is easier to setup and test [14:35] asac: Yes, categories are filters for the displayed apps [14:36] asac: Sure, will try that [18:13] For UME, where in our scripts is the most appropriate place to set a global environment variable like $BROWSER? === emgent_ is now known as emgent [19:42] lool: I can never remember what I'm supposed to do after attaching a debdiff to an LP bug. This one's yours (LP #191064). If I hadn't poked you on IRC, would I need to set some further setting so it'd get your notice? [19:42] Launchpad bug 191064 in ume-config-common "Set midbrowser as the default URL handler for HTTPS" [Undecided,Confirmed] https://launchpad.net/bugs/191064 [22:27] \o [22:27] lo [22:28] is ubuntu-mobile x86 targeted or also other devices?