www-client/opera with USE="-qt-static -qt3" has severe issues with Qt 4.5.3 libraries. Users should stay with Qt 4.5.2 until these issues are resolved. This is Opera bug number is DSK-267319 Reproducible: Always Steps to Reproduce:
This bug is valid Users should use qt-static use flag as a workaround until this bug is fixed by upstream
=www-client/opera-10.10_pre4672 just went into the tree, which equals 10.10 beta1. Does it solve the performance issues seen in the snapshots? It does seem a lot more responsive to me, but sadly the ChangeLog[1] does not mention Qt-4.5.3. [1] http://www.opera.com/docs/changelogs/unix/1010b1/
The blog comments[1] seem to say it hasn't been solved. [1] http://my.opera.com/desktopteam/blog/show.dml/4424350#comment10436556 and onward.
Opera build 10.10_4685 fixes bug with 4.5.3 http://my.opera.com/desktopteam/blog/2009/10/29/post-seminar-unite-build Fixed Bug DSK-267319 (Sleepless main loop with Qt 4.5.3)
(In reply to comment #4) > Opera build 10.10_4685 fixes bug with 4.5.3 > http://my.opera.com/desktopteam/blog/2009/10/29/post-seminar-unite-build > Fixed Bug DSK-267319 (Sleepless main loop with Qt 4.5.3) Thanks. That's in the tree now as =www-client/opera-10.10_pre4685 . Please test and report back here.
The article was edited to strike through "Unix users will be happy to know that this build should finally work with Qt 4.5.3+." and indeed I see the same misbehaviour that was noted in previous builds.
(In reply to comment #6) > The article was edited to strike through "Unix users will be happy to know that > this build should finally work with Qt 4.5.3+." and indeed I see the same > misbehaviour that was noted in previous builds. > True. It doesnt work for me. I had to enable qt-static again in order to make opera work again :)
Should opera stable ebuilds be changed not to install shared opera builds if qt-4.5.3 is installed?
Seems the bug is fixed by last QT update which fixes eventloop regression (qt-core-4.5.3-r2, and 4.5.3-r1 for others)
I didn't want to rewrite major parts of the ebuilds so I settled for having IUSE=+qt-static, which should give everyone who hasn't set an explicit preference through USE / package.use the working build. (In reply to comment #9) > Seems the bug is fixed by last QT update which fixes eventloop regression > (qt-core-4.5.3-r2, and 4.5.3-r1 for others) Well, please report if it indeed does fix the issue.
Looks good to me (both 10.01 and 10.10_pre4685 with Qt 4.5.3). I am thinking about removing IUSE="+qt-static" when the patched x11-libs/qt-*-4.5.3 ebuilds are stable for all three stable-going arches.
(In reply to comment #11) > I am thinking > about removing IUSE="+qt-static" when the patched x11-libs/qt-*-4.5.3 ebuilds > are stable for all three stable-going arches. That would be now.
Thanks for the notice. For completeness' sake, Opera blogged thusly: "[I]t seems we got around the Qt 4.5.3+ 100% CPU issue. Also a fix for freezes during DST-to-normal time switch." http://my.opera.com/desktopteam/blog/2009/11/05/key-fixes-and-stability The wording suggests they worked around it and didn't really need to fix (all of) it, so I have removed the USE=qt-static preference from all the ebuilds rather than just from the 10.10_pre4694 build which purportedly "gets around" the issue.