On Mon, Nov 05, 2012 at 02:26:49PM +0100, David Coppa wrote: > On Mon, Nov 5, 2012 at 2:07 PM, Landry Breuil <lan...@rhaalovely.net> wrote: > > On Mon, Nov 05, 2012 at 02:00:55PM +0100, David Coppa wrote: > >> On Mon, Nov 5, 2012 at 1:49 PM, Landry Breuil <lan...@rhaalovely.net> > >> wrote: > >> > Hi, > >> > > >> > here's an update to mnemosyne 2.1, now using py-qt4. Crashes python > >> > @startup here on amd64, somewhere in qtwebkit : > >> > > >> > #0 0x00000727266bc3d7 in WTF::OSAllocator::reserveAndCommit () from > >> > /usr/local/lib/libQtWebKit.so.2.0 > >> > > >> > But it might be a local issue. I dont use it, but i'd like to see less > >> > ports using ancient toolkits. Anyone can really test it ? > >> > > >> > >> There's a linux-only fix for this crasher: > >> > >> https://bugs.launchpad.net/ubuntu/+source/ubiquity/+bug/710582 > >> > >> https://bugs.webkit.org/show_bug.cgi?id=42756 > >> > >> http://trac.webkit.org/changeset/86957 > >> > >> But we lack the MAP_NORESERVE flag for mmap(2), so dunno if a > >> workaround exists for OpenBSD... > > > > Oh, not that fucking crap again... so id say maybe our qt4 lacks some > > patches from www/webkit. www/arora segfaults right away at the same spot > > on amd64, so qtwebkit looks broken. > > Are you referring to > www/webkit/patches/patch-Source_JavaScriptCore_jit_ExecutableAllocatorFixedVMPool_cpp > ?
Yes. I'll backport it to x11/qt4, as of now anything using qtwebkit with default ulimit -d segfaults at startup. Bumping ulimit -d >150000 allows me to start arora and mnemosyne fine, and they eat 1Gb of memory directly at startup, as webkit used to do before being whacked on the neck. Landry