Meeting on irc.gnome.org:#gtk-devel Meeting started January 31 2005 16:04 EST (21:04 UTC) In attendence: Owen Taylor (owen), Manish Singh (yosh), Federico Mena Quintero (federico), Matthias Clasen (mclasen), Billy Biggs (vektor), Robert Ögren (roboros), Tor Lillqvist (tml), Maciej Katafiasz (mathrick) lets begin with concrete things... I want to do 2.6.2 releases towards the end of this week, probably on Friday mclasen: I'll do a pango-1.8.1 this week as well Are there important bugs which need to be fixed before that ? mclasen: Might be worth mailing federico if he doesn't show up (I don't know if there are filechooser bugs, but if there are important bugs, they are there) bug 154007 perhaps? yes, I'll do that, last time I checked federico was trying to track down the file duplication issue roboros: oh, I hadn't noticed that you had a patch for that one. Nice, I'll check it out roboros: Not with that comment please in any case owen: you mean the comment about CLAMP in the actual patch? roboros: Yes. no problem, that was mostly a reminder for myself roboros: But looks like there is a deeper bug That change may be OK, but for the paned to get allocated less size than it's requisition is busted I don't think we can fix the treeview scrolling issues without jrb, so I'll remove 163214 from the 2.6.2 milestone The wholle treeview scrolling thing is really strange. vektor: I think there are at least 2 independent scrolling issues Agreed. one is that scrolling is affected by concurrent model changes the other one is the fixed height optimizations seem to not quite work Yeah, I've been investigating the first one and not the second. It's confusing. owen: yeah you're right, someone needs to do more debugging the next topic I would like to discuss is gail at the lsb a11y workshop in hawaii last week, jrb and billh discussed it and the idea came up to merge gail in gtk+ the motivation for having gail as a separate library was IIRC mainly fears of it not being ready in time for gtk 2.0 speaking of gail yes I'm looking into LDTP (it's the a11y-based GUI testing framework from the Bangalore folks) and while I can have the test data independent of gtk, it would be good to distribute both together ... but that assumes that gtk and its a11y stuff is self-contained does gail build on win32 ? does it do something useful there ? mclasen: hmm, i have compiled it once, but i doubt it does anything sensible. not that i have ever really tried what it does on x11... ok, I guess we should invite the a11y team to a joint meeting to discuss the idea more thoroughly federico: I was asking earlier about 2.6.2 bugs... I want to do 2.6.2 towards the end of this week are there important filechooser bugs we need to fix before that ? I remember I have promised to try to improve stuff in the file chooser wrt case insensitivity, but I don't know whether i will have time before 2.6.2. anyway, as the natives are getting restless, I guess iIdefinitely should distribute win32 binaries for 2.6.2 tml: can we try to finally get win32 binaries hosted on ftp.gtk.org ? we could do that, yes can you send instructions? presumably just upload files there with scp? yes, do you have an account there ? nope I think yosh hands them out ok will ask him ok, another topic I had in my mind is renaming of bookmarks in the file chooser and the question how we handle that, since the .gtk-bookmarks file format is not extensible anyway, have to leave now. Bye bye tml: I'm about ready to commit my win32 backend to Cairo, btw. The native font stuff isn't done yet, but I'll commit something initially that uses the fontconfig bcakend owen: good, will try to have a look tml: BTW, you say that gdb works OK with cygwin binaries for you? s/cygwin/mingw/ yes, no show-stopping problems also insight works tml: And no special tricks? well, you can't set breakpoints in dlls before they have been loaded, so usually i do "b main" and "run", then set breakpoints in the dlls i am interested in Jan 31 16:56:06 --- mathrick is now known as mathrick|sick and after control-c you aren in some weird location, "bt" doesn't give any sensible information, so you can't do that to figure out what's taking so long, for instance. but "cont" still works. * tml feels a fever coming * mathrick|sick joins tml tml: Hmm, I can't even break on main, and I've genreally had complete junk backtraces when I segfault * tml fetches some meds * mathrick|sick goes to bed the mingw gdb I was using previously didn't work well with the --image-base=0x10000000 link param that libtool puts there, I had to hack ltmain.sh to get rid of that and enable auto image base instead is that the mingw gdb or cygwin? roboros: i wonder wht auto-image-base isn't the default? have you ever had any problems with it? none that I'm aware of tml: mingw gdb owen: try the cygwin gdb (or was it so that you don't have cygwin at all installed?) tml: I don't have cygwin installed at the moment, but if it gives me a debugger that works, I may have to do that i haven't tried the mingw gdb (and can't do it right now, logged in as admin and running a full virus scan. (six hours and counting, sigh)) at least for me cygwin's gdb works, but might well be that also mingw's would, and that your problem is because of something peculiar in your machine.. tml: I'm doing some experiements now .... it looks like some particular library pulled in by cairo is breaking things ... gdb works on a "hello world" tml: Errr, what the heck, cairo.pc includes -Wl,-s (-s being strip!) heh nasty :) tml: OK, my skepticsm about gnuwin32 is much increased was that caused by something from them? tml: freetype-config output was *full* of junk. -Wl,-s -lgw32c -liberty -lintl -liconv -lwsock32.... hmm, dunno where my freetype-config is from liberty? there's library named "iberty"? yes, from FSF, included in various GNU packages IIRC, perhaps not intended to be installed separately though? mathrick|ick: Yep :-). It's been around in the GNU package forever owen: heh it does what? assures you can't link to non-GPL packages or something? :) s/packages/code/ mathrick|sick: It has random semi-useful crap, like bits of libbfd that arne't in libbfd for some reason ah I have *no* idea why a build of freetype, which builds natively just about everywhere would need it. or libiconv, or any of the other stuff. Plus, even if they do need it, the dll's will pull in themselves. * tml ponders other potential "fun" library names. oser, eprechaun, ard, ove, esbian, eet. hmm, there must be a list of them somewhere iconv is useful, no? mathrick|sick: iconv is useful in some case. Freetype should have no need of charset conversion, however. fonts can come in multiple encodings, AFAIK mathrick|sick: freetype doesn't do the conversion. owen: but it needs to extract the glyph from character table, no? owen: does the libglib-2.0.pc that gets generated for you on mingw include -lintl, -liconv, and -lwsock32btw? No -lwsock32 for me mathrick|sick: You hvae to give it character indices as the font is encoded. It can't covert euc-jp to unicode or anything like that but -lintl and -liconv? aren't these, too, wrong? shouldn't software that uses libintl link to it explicitly? tml: -lintl is probably right. -liconv shouldn't be needed since we have g_iconv() as a wrapper owen: ah, so you're responsible for parsing chartable into something useful yourself? mathrick|sick: More or less. Most real fonts have unicode chartables these days mhm owen: which reminds me, how are fontsets in pago created? Like, when I choose sans, I have no idea nor way to see what kanji font will be chosen, right? And also it seems to prefer some CJK fonts choosing them almost all the time All fontconfig which means... ? :) can fontchooser give that info? or influence the choice in any way? mathrick|sick: What is the fontchooser? owen: widget for choosing fonts in gtk+ oops, another seizure coming * mathrick|sick -> bed mathrick|sick: The fontchooser can't specify it but a program can say "arial,times roman,sans" - a list like that mathrick|sick: Talk to you some other time * tml leaves for bed, too Meeting ended January 31, 17:31 EST (22:31 UTC)