this is based gaim-0.64-r2.ebuild. also, this ebuild contain patch from the im-ja package. Reproducible: Always Steps to Reproduce: 1. 2. 3.
Created attachment 14613 [details] ebuild.
Created attachment 14614 [details, diff] patch for cjk input.
Comment on attachment 14613 [details] ebuild. # Copyright 1999-2003 Gentoo Technologies, Inc. # Distributed under the terms of the GNU General Public License v2 # $Header: /home/cvsroot/gentoo-x86/net-im/gaim/gaim-0.64.ebuild,v 1.5 2003/07/09 14:11:56 lostlogic Exp $ IUSE="nls perl spell ssl nas cjk" DESCRIPTION="GTK Instant Messenger client" HOMEPAGE="http://gaim.sourceforge.net/" EV=2.00 SRC_URI="mirror://sourceforge/gaim/${P}.tar.bz2 ssl? ( mirror://sourceforge/gaim-encryption/encrypt-${EV}.tar.gz )" SLOT="0" LICENSE="GPL-2" KEYWORDS="~x86 ~ppc ~alpha ~sparc" DEPEND="=sys-libs/db-1* >=x11-libs/gtk+-2.0 >=dev-libs/glib-2.0 nas? ( >=media-libs/nas-1.4.1-r1 ) nls? ( sys-devel/gettext ) media-libs/libao >=media-libs/audiofile-0.2.0 perl? ( >=dev-lang/perl-5.6.1 ) ssl? ( dev-libs/nss ) spell? ( >=app-text/gtkspell-2.0.2 )" src_unpack() { unpack ${P}.tar.bz2 use ssl && { cd ${S}/plugins unpack encrypt-${EV}.tar.gz cd encrypt epatch patchfile.0.65 cd ../.. touch aclocal.m4 config.h.in Makefile.in */Makefile.in */*/Makefile.in */*/*/Makefile.in configure } use cjk && { cd ${S}/src epatch ${FILESDIR}/gaim_gtkimcontext_patch.diff } } src_compile() { local myconf use perl || myconf="${myconf} --disable-perl" use spell || myconf="${myconf} --disable-gtkspell" use nls || myconf="${myconf} --disable-nls" use nas && myconf="${myconf} --enable-nas" || myconf="${myconf} --disable-nas" econf ${myconf} || die "Configuration failed" emake || die "Make failed" } src_install() { einstall || die "Install failed" dodoc ABOUT-NLS AUTHORS HACKING INSTALL NEWS README TODO ChangeLog } pkg_postinst() { if [ `use ssl` ]; then ewarn ewarn "You have chosen (by selecting 'USE=ssl') to install" ewarn "the gaim-encryption plugin ( http://gaim-encryption.sf.net/ )" ewarn "this plugin is NOT supported by the Gaim project, and if you" ewarn "expierence problems related to it, contact the Gentoo project" ewarn "via http://bugs.gentoo.org/ or the gaim-encryption project." ewarn fi }
Created attachment 14615 [details] ebuild delete unneccesary line echo `pwd`. however, sorry that i'm not good at using bugzilla. please forgive me that i submit unnecessary comment ;)
in gaim-0.65-r1 soon.
Hi. I tested your patch and ebuild provided above with XIM and found that it makes it difficult to input messages via XIM. Running gaim, when I turn XIM (this time I tested skkinput and kinput2) on, I could input as many characters as I want, but once I press Ctrl or Return key, always it freezes :( If I remove the patch from the ebuild, this problem doesn't occur. Do you know any workaround with the problem?
Now I know what was the problem you described before, mithrandir. It will solve the cjk's input problem with im-ja, but it will cause many problems with other XIMs, skkinput, kinput (both are Japanese XIM) and chinput (Chinese Big5 XIM). Even though ami, Korean IM, doesn't have this problem as mithrandir reported on IRC, I think we should think better way to apply the patch.
The patch breaks stuff here as well, as I usually use kinput2 (I wanted to report it but found it is reported already). Since the big problem in gaim usually is that pressing Enter sends the message, a workaround would be to press Shift+Enter when you want to insert the converted text. This workaround should work with most input methods. Until the patch is officially pulled, I have to remember to USE=-cjk emerge gaim.
problem is simple. when we call gtk_im_context_filter_keypress() for one,same event, then key_press_event emitted infintely. Even if the return value of gtk_im_context_filter_keypress() is FALSE (this means im doesnt handle that keystroke(or event?). I test in XIM(skkinput, ami) , imhangul, im-ja. I cant find any problem that is not XIM. but skkinput has problem, and ami too. but the keystroke that occur problem is different. ami's problem-keystroke is backspace. and skkinput is shift. (in my case). I wonder it is bug of gtk2's XIM input module or XIM itself. This is not simple problem for me :( anyway I'm finding why this occur. If you want, please help me.
I suggest that we change use cjk to use immodule and add an entry for immodule local USE flag as: net-im/gaim:immodule - Enables GTK+2 immodule input method support or something like it. Since gaim-0.66-r2.ebuild was marked stable yesterday, not a few XIM users think gaim is broken, rush to file a bug report and find this bug ;-( gaim_gtkimcontext_patch.diff should be applied only if users know what to do because it changes default behaviour of gaim and not all cjk users want this patch (because immodule has known to be unstable for everyday use and is not widely used at least among Japanese). I'm not going to say we should remove the patch because some people have a benefit from it, but I want to emphasise that it should not do harm to existing users and we better stop nonsense of emerging gaim with USE="-cjk" even when we want to use cjk languages.
Ok, I was just debugging the 0.66-r2 gaim with USE=+cjk and I found that this is already a known problem... So, marking something as stable and applying a patch that break things (gaim locks up when presiing Shift or Ctrl or something else while kinput2 is active) is a BAD THING(tm). As long as we don't have separate useflags for chinese, japanese, etc. (and AFAIS, it is dependant on the input method, not language), I'd suggest omit the path by default, so people with +cjk can use gaim "the old way" (press Shift+Enter).
Created attachment 15428 [details, diff] patch updated to fix XIM event loop I have modified my original patch so that it works with gtkxim (tested kinput2) and im-ja as well. I hope it will please everybody now. (I just don't know why I keep spending my time on things mostly irrelevant to me such as gaim and gentoo...)
Created attachment 15435 [details, diff] modifed to gaim-0.66 modified boti's gaim patch for gaim-0.66 source. thanks to boti. ;)
Comment on attachment 15435 [details, diff] modifed to gaim-0.66 I test boti's new patch , and it works FINE at skkinput and ami. I think it works also with kinput. please change patch with. thanks.
OK testing out this new patch and will add a revision bump to alert users to the new patch. Will let you know when it's checked in.
OK gaim-0.66-r3 is in portage and is in STABLE, as the only change from the stable gaim-0.66-r2 is this patch. I know this is taboo but 0.66-r2 wasn't working for cjk so I see no harm in bumping this in right away. Non-cjk users will see absolutely no change. Please test out and verify.
Thanks for the patch and revision bump. I tested it with skkinput and kinput2 and saw no problem. I'm absolutely sure CJK users will appreciate Botond, Park and Don's work. I think it would be nice to add Botond's name to ChangeLog in respect to fixing the bug (this bug was so annoying for Japanese gaim users that some of us masked >=gaim-0.65-r1 completely rather than emerging every now and then with USE="-cjk", since there were so many releases after that...).
You are absolutely right. I've added a thanks to Botond Botyanszki and Park Ji-in in the ChangeLog, which should have been there from the start. I really deserve no credit as all I did was add files into CVS. :p Anyway as soon as I hear back from Park Ji-in I'll feel comfortable resolving this bug fixed.
By the way I take responsibility and apologize for marking 0.66-r2 stable without getting this bug taken care of. I realize that it is causing a lot of trouble for C/J/K users and can be incredibly frustrating to have to work around all of the time.
I have filed a bug against gtk, because I think the event-loop should be blocked in the input module, not in the signal handler as in the patch. See http://bugzilla.gnome.org/show_bug.cgi?id=119075
I just happy using gaim FINE. thanks to all of you :)
Alright I'm closing this bug as FIXED. Thanks again you guys.
Hi, I just noticed that after emerging gaim with this patch skkinput doesn't work well (it is alright with kinput2). When I use skkinput, at first hand nothing wrong but everytime immediately after I send a message the first character I type will not be passed through to skkinput (so I have to erase first character for each message :-/ ). A screenshot is available at http://www.usata.org/misc/gaim-skkinput.png (I tried to send the same message twice, but failed to input the second time) This only happens when I input hiragana/katakana with skkinput. If I input kanji (that means I press Shift key altogether) it will sucessfully be passed to skkinput. If I don't use the patch, this problem doesn't occur. Now I'm using kinput2 only for the sake of gaim :-(
Could you try emerging gaim-0.67 without the patch (it doesn't apply anyway). I think the problem may be fixed in gaim itself. I have no problem using kinput2 without the patch. I have a custom ebuild on my mirror if you want: rsync://gg3.net/gentoo-portage-chutz/net-im/gaim/
I have what will be the gaim-0.67.ebuild (no cjk support yet since the patch didn't take anyway. gaim really changed their headers) in Bug 26806. Wasn't it actually the case that this was a problem on gnome/gtk's side of the fence? I saw that they closed that bug as invalid. I don't use CJK so I'll need to lean heavily on you guys to come up with a fixed fix.
I just tried 0.67 (straight from sf) and it doesn't seem to be fixed. I looked at the source as well and it didn't change anything in this respect. Please note that you can still use gaim with CJK input with the unpatched version if you set your preferences so that gaim uses CTRL-Enter and disable Enter under Interface->shortcuts. (Maybe this is what Georgi did?!) See this thread also: http://sourceforge.net/mailarchive/forum.php?thread_id=2871169&forum_id=29602 Another dirty fix would be to remove the g_signal_connect(line 3454, and 3552) from gtkconv.c (with a patch) when CJK is set but I wouldn't recommend this. I think that with the patch applied, only skkinput users will see odd behaviour, this is still better than all CJK users. Unfortunately I don't have the time to look for the fix with skkinput. This bug is rather on gaim's side of the fence than GTK's. Though the bug I filed against GTK (lockup with kinput2) was invalid for some reason, I need to talk to Owen about this. Hope this clears thing up. I leave it up to the package maintainer what solution he choses.
My bad. This patch we're talking about was introducing so many problems I got used to pressing Shift+Enter and sort of forgot what the original problem was :(. I confirm the problem is not fixed.
I totally agree with Botond's opinion. I tried gaim-0.67 and found the original problem (I first didn't recognise it since I set CTRL-Enter to send a message in preference, as Botond suggested). gaim-0.67 works perfectly fine in terms of kinput2 and skkinput, because they use XIM to talk to gaim. This is not GTK+/GNOME problem but gaim's problem. I can use gedit with skkinput having no problem.
Created attachment 17201 [details, diff] Botond's patch which modified to gaim 0.68. same as 0.66's and so. I test this with im-ja, imhangul, skkinput. and works fine. (in my case) I suggest reinput this patch in portage, because there is someone who want this patch to use.
gaim 0.71 is latest
What a fast response :) Anyway, that patch also works for gaim-0.71. Tested by imhangul, nabi (these are korean), and im-ja, skkinput (these are japanese).
Will look into this further, has anyone seen any detrimental effects with this patch?
As I described in comment #23, the patch has a problem with skkinput.However, most people will have benefits from the patch (only skkinput'speople are affected) and they were having input problem atm. I agree with adding the patch by default or USE="cjk", but it would be nice to be able to turn it off by local USE flag such as "noimmodule" (as I describedin comment #10).
Hi! Welcome back, rizzo :) Is there any chance the patch committed again? I think applying it when USE="cjk" is fine (though I suggest local USE flag "noimmodules" might be useful). If you are not sure or don't have test environment about the patch, I'd like to add cjk herd to gaim and maintain the patch. What do you think of it?
Can you verify the patch works with gaim-0.72?
Attachment 17201 [details, diff] applies to gaim-0.72-r1.ebuild and works fine. (Still have the probmle described in Comment #23, though)
If you feel that it works despite the problems you mention in Comment 23, I'll go ahead and suggest to lostlogic that we add the patch back in.
Just one more thing as well: Have you guys brought this bug up to the Gaim developers?
I will probably commit this and have a postinstall warning about the possible issues in comment 23. lostlogic let me know if you have an issue with this.
Oh well, I should have done so. I submitted a bug to Gaim sf.net -> http://sourceforge.net/tracker/index.php?func=detail&aid=840904&group_id=235&atid=100235 I'm okay with the solution in Comment #39.
gaim-0.72-r2 checked into CVS.
Just an FYI that gaim-0.73 is out and the patch as is fails to take. I'll see if I can see what needs to be changed.
Looking at the sourceforge bug that usata filed, is the CJK patch posted here even necessary?
OK I redid the patch. Line numbers just got bumped down it seems. Releasing soon.
Looking again at usata's bug. It seems that at the very least I don't need the postinfo warning about CJK input now. Can anyone confirm? I don't use CJK, despite my being half-Korean. I don't need to tell you how disappointed my mom is in that.
I talked with mithrandir at IRC few days ago, and found freeze bugs of some input methods were fixed, while other bugs weren't. I tested only a couple of input methods and reported it was fixed but I was wrong about it. I was going to test as many input methods as possible and reopen that bug (but I haven't much time to do that until today). In short, we still need the patch.
Looking at the diff for gtkconv.c between 0.67 and 0.72 I don't see a single line of change that would fix this problem. Maybe it's just my bad eyes... And the comment "I believe it is fixed" on the SF bugtracker isn't very technical either.
Should I still be presenting the postinfo warning about Comment 23?
Sorry for late response. skkinput has trouble with the patch in gaim-0.74.ebuild, so we still need that message in postinst.