At the moment, the jit USE flag is masked in the x32 profile for dev-qt/qtwebkit, dev-qt/qtscript and dev-qt/qtdeclarative, citing that it does not work. That is true for qtwebkit, however, it's not true for qtscript and qtdeclarative, as I compiled them on my x32 machine without issues. Conversely, USE=jit does not work in spidermonkey (tested 17 and earlier), so that could be masked instead. Also, dev-lang/luajit does not work on x32 either, as per bug #498784.
are you talking about qt4 or qt5?
(In reply to Dainius Masiliūnas from comment #0) > Conversely, USE=jit does not work in spidermonkey (tested 17 and earlier), > so that could be masked instead. Also, dev-lang/luajit does not work on x32 > either, as per bug #498784. You should file separate bugs for these two packages.
(In reply to Davide Pesavento from comment #1) > are you talking about qt4 or qt5? I tested both, with the exception of qtwebkit:5 (testing that right now). Everything built correctly aside from qtwebkit:4 so far. (In reply to Davide Pesavento from comment #2) > You should file separate bugs for these two packages. For masking, or in general? There's already bug #498784, for one.
Just finished building qtwebkit:5, and it succeeded fine. So the only one that is still failing is in fact qtwebkit:4.
Ok, they build fine. Do they also run without crashing?
Well, I'm running lxqt, and it hasn't crashed once; I'm not sure how much it makes use of qtscript or qtdeclarative, though. I don't have a testcase for the qt4 versions yet. I can test qt5webkit on QupZilla, haven't tried yet.
I couldn't test it on QupZilla, since it very mysteriously crashed with SIGBUS when the qt5 version is used (but it's not related to jit, it does that even without jit anywhere). So instead I tried Otter, and it works perfectly well (in fact, it's incredibly smooth and responsive). So yes, qtwebkit:5 and qtscript:5 +jit works fine. I don't know good testcases for the qt4 versions of them, or for qtdeclarative.
Hold on, now I remember why the flag is masked... Those packages may even build and run on x32, but the JIT engine is *not* actually enabled there, even with USE="jit"! At least last time I checked there was some rather heavy #ifdef magic that enabled the JIT engine only on those CPU+OS combinations that were known to work, and x32 was not among them. The situation *may* have changed in Qt5 (it's possible that x32 JIT support was introduced later), but I doubt anything was done for Qt4. So, unless you can confirm that things have changed and the JIT engine is now enabled by the internal build logic, I'm not going to unmask the flag.
I checked qtdeclarative, 5.8 branch, and the JIT is not available on x32. qtscript is deprecated and qtwebkit development has stopped, so I don't expect any support there either. Closing.
