Oddly the only site I have found it to crash with is http://bugs.gentoo.org but it consistently segfaults immediately after entering the URL Reproducible: Always Steps to Reproduce: 1.start chromium 2.enter bugs.gentoo.org in address bar 3. Actual Results: Segfault crash Expected Results: Open the site for viewing This happened after an upgrade of dev-libs/icu from 48 to 49.1. I had done a revdep-rebuild after updating last night and chromium came up to rebuild base on lib icu.
Created attachment 307767 [details] emerge info
Created attachment 307769 [details] gdb output
I've had similar experiences with www-client/chromium-18.0.1025.142 and dev-libs/icu-49.1. However for me chromium crashes immediately on startup. Reverting to dev-libs/icu-4.8.1.1-r1 fixes the issue.
Broken here, too.
I'm not sure if it would be better to put add <dev-libs/icu-49 dependency to chromium, or add dev-libs/icu-49 back to package.mask. Thoughts?
Unless other regressions pop up, I'd opt for adding a dependency constraint in the chromium ebuild.
04 Apr 2012; Mike Gilbert <floppym@gentoo.org> -chromium-17.0.963.83.ebuild, -chromium-18.0.1025.140.ebuild, chromium-18.0.1025.142.ebuild: Depend on <dev-libs/icu-49 for bug 410777. Remove old. I left chromium-19 alone for now. I'll test it later.
Thanks!
chromium-19.0.1084.1 is also affected. For me, the segfault is triggered immediately after logging into bugs.gentoo.org.
(In reply to comment #5) > I'm not sure if it would be better to put add <dev-libs/icu-49 dependency to > chromium, or add dev-libs/icu-49 back to package.mask. > > Thoughts? My vote would be to mask dev-libs/icu-49.1 for now as libreoffice-bin is built against .48 and breaks (I gave up after about 3 symlinks to the new .49 so's and just installed the compiled verion of libreoffice. Other packages probably need some time to be tested with 49.1 before unmasking as well. Thanks for jumping on this so quickly.
(In reply to comment #10) Can you file a bug for that libreoffice-bin issue? It should probably depend on =dev-libs/icu-4.8*.
Created attachment 307851 [details] Better backtrace
Oh, my backtrace is from 19.0.1084.1.
(In reply to comment #10) > (In reply to comment #5) > > I'm not sure if it would be better to put add <dev-libs/icu-49 dependency to > > chromium, or add dev-libs/icu-49 back to package.mask. > > > > Thoughts? > > My vote would be to mask dev-libs/icu-49.1 for now as libreoffice-bin is > built against .48 and breaks (I gave up after about 3 symlinks to the new > .49 so's and just installed the compiled verion of libreoffice. Other > packages probably need some time to be tested with 49.1 before unmasking as > well. > > Thanks for jumping on this so quickly. Stable version of libreoffice (3.4.5.2) breaks too, and refuses to compile with icu-49.1
(In reply to comment #14) > (In reply to comment #10) > > (In reply to comment #5) > > > I'm not sure if it would be better to put add <dev-libs/icu-49 dependency to > > > chromium, or add dev-libs/icu-49 back to package.mask. > > > > > > Thoughts? > > > > My vote would be to mask dev-libs/icu-49.1 for now as libreoffice-bin is > > built against .48 and breaks (I gave up after about 3 symlinks to the new > > .49 so's and just installed the compiled verion of libreoffice. Other > > packages probably need some time to be tested with 49.1 before unmasking as > > well. > > > > Thanks for jumping on this so quickly. > > Stable version of libreoffice (3.4.5.2) breaks too, and refuses to compile > with icu-49.1 i don't use both of chromium and libreoffice though, i have $ equery depends dev-libs/icu * These packages depend on dev-libs/icu: dev-db/sqlite-3.7.11 (icu ? dev-libs/icu) dev-libs/libxml2-2.7.8-r5 (icu ? dev-libs/icu) dev-util/dwdiff-2.0.2 (dev-libs/icu) w/ dev-libs/icu-49.1. masking bumped library in global package.mask to avoid some specific package segfaults? is icu-49.1 really broken?
(In reply to comment #14) > (In reply to comment #10) > > (In reply to comment #5) > > > I'm not sure if it would be better to put add <dev-libs/icu-49 dependency to > > > chromium, or add dev-libs/icu-49 back to package.mask. > > > > > > Thoughts? > > > > My vote would be to mask dev-libs/icu-49.1 for now as libreoffice-bin is > > built against .48 and breaks (I gave up after about 3 symlinks to the new > > .49 so's and just installed the compiled verion of libreoffice. Other > > packages probably need some time to be tested with 49.1 before unmasking as > > well. > > > > Thanks for jumping on this so quickly. > > Stable version of libreoffice (3.4.5.2) breaks too, and refuses to compile > with icu-49.1 testing app-office/libreoffice-3.5.2.2 builds fine with icu-49.1, not sure about runtime IIRC I've seen similar crashes with chromium after previous icu upgrade too, why is it so picky about lib like icu :/
libreoffice-3.5.2.2 builds and works fine with 49.1. libreoffice-3.4.5.2 refused to build *before* I updated to icu-49.1.
I just discovered that there's something going on with KDE. I've updated to 4.8.2, and during emerge, every single package produces a multiple copies of this warning: Unable to load library icui18n "Cannot load library icui18n: (libicui18n.so.48: cannot open shared object file: No such file or directory)" It does build though and run though (running 4.8.2 right now.)
(In reply to comment #18) > I just discovered that there's something going on with KDE. I've updated to > 4.8.2, and during emerge, every single package produces a multiple copies of > this warning: > > Unable to load library icui18n "Cannot load library icui18n: > (libicui18n.so.48: cannot open shared object file: No such file or > directory)" > > It does build though and run though (running 4.8.2 right now.) This would be caused from missing revdep-rebuild after the icu update.
(In reply to comment #19) > (In reply to comment #18) > > I just discovered that there's something going on with KDE. I've updated to > > 4.8.2, and during emerge, every single package produces a multiple copies of > > this warning: > > > > Unable to load library icui18n "Cannot load library icui18n: > > (libicui18n.so.48: cannot open shared object file: No such file or > > directory)" > > > > It does build though and run though (running 4.8.2 right now.) > > This would be caused from missing revdep-rebuild after the icu update. I already ran it when I first upgraded to icu 49.1. If I run it again (revdep-rebuild -i) it passes cleanly. So this can't be it. Maybe something is trying to dlopen the library?
(In reply to comment #15) > masking bumped library in global package.mask to avoid some specific package > segfaults? is icu-49.1 really broken? Sorry about that. Anyway, probably best to leave it masked at this point.
I reported the issue to Chromium upstream.
on a 6 day old amd64~ system; a world update was done before icu-49 was hard masked; a world update done after icu-49 was masked, partially failed because of icu-49* dependencies. revdep-rebuild could not correct this. gedit and epiphany failed. After a reboot gnome3 failed entirely displaying only background wallpaper. Unmerging chromium, unmasking icu, emerging icu (49.1.1), emerging -uND world restored normal functioning. emerge chromium caused icu-48* and chromium-18* to reinstall and killed chromium, gedit, epiphany, and gnome (at next startx) to fail. Unmasking chromium, emerge icu-49.1.1, emerge -uND world, emerge chromium restored all functions. Except: chromium-19* will open https://bugs.gentoo.org but will spontaneously close when login is attempted. login to other secure sites succeeds.
Just to sort things out, I'd like to refer to the upstream bug: http://code.google.com/p/chromium/issues/detail?id=122125 and the fix mentioned there: " Ok, digging further on why the failure code is returned on that particular regexp, it turned out to be an icu problem, specifically: http://bugs.icu-project.org/trac/ticket/9283 So using the corresponding "fix" http://bugs.icu-project.org/trac/changeset/31782/icu/trunk/source/i18n/regexcmp.cpp makes chromium work on autofill again. " Just writing from within chromium with icu-49 and that patch applied :)
Looks like there are two upstream bugs regarding that issue. One from April 5th, the only one the search returned to me earlier this day, and the one linked in this bug's description from April 6th. About libreoffice: At least librecalc works for me with icu-49.1.1 (even without the patch).
dev-libs/icu-49.1.1-r1 is now in gentoo-x86.
(In reply to comment #26) > dev-libs/icu-49.1.1-r1 is now in gentoo-x86. Let's give it a shot then; I plan to test each Chromium major version, starting with 20 and working back to 18. Help/additional reports would be appreciated.
I have verified that the following chromium versions do not segfault when built against dev-libs/icu-49.1.1-r1. 18.0.1025.168 19.0.1084.41 20.0.1123.2 I will give it a day for others to test, and then remove the dependency on <dev-libs/icu-49.
(In reply to comment #28) > I have verified that the following chromium versions do not segfault when > built against dev-libs/icu-49.1.1-r1. > > 18.0.1025.168 > 19.0.1084.41 > 20.0.1123.2 > > I will give it a day for others to test, and then remove the dependency on > <dev-libs/icu-49. Thanks for the work, tested here. I committed dependency changes.
