Support for building www-client/firefox-36 without JIT support was temporarily disabled as of bug 541534. If possible, please re-add USE="jit" to allow building without JIT.
As soon as upstream supports disabling jit, we will re-add the ability to toggle it via USE.
Is there an upstream bug report for this?
OK I have a couple of patches and have re-added USE="jit" on the firefox-36.0 ebuild on mozilla overlay. It builds now but I can't confirm if it's going to work properly at run-time, especially on hardened and/or without the same pax-marks that are necessary for USE="jit" in the first place. Please test.
The ebuild from the "mozilla" overlay emerges and runs just fine as far as I can tell. It installed and ran with the following PAX markings: # paxctl-ng -v `which firefox` /usr/bin/firefox: open(O_RDWR) failed: cannot change PT_PAX flags PT_PAX : -e--- XATTR_PAX : not found
Upstream bugs are https://bugzilla.mozilla.org/show_bug.cgi?id=1138187 and https://bugzilla.mozilla.org/show_bug.cgi?id=1138177
(In reply to Jaak Ristioja from comment #4) > The ebuild from the "mozilla" overlay emerges and runs just fine as far as I > can tell. It installed and ran with the following PAX markings: > > # paxctl-ng -v `which firefox` > /usr/bin/firefox: > open(O_RDWR) failed: cannot change PT_PAX flags > PT_PAX : -e--- > XATTR_PAX : not found Thank you. If you can test for a couple more days just to make sure it's not unstable/crashy, I'll push these patches to the tree and re-enable USE="jit".
Created attachment 397826 [details] firefoxcrashes.tar.xz (In reply to Ian Stakenvicius from comment #6) > Thank you. If you can test for a couple more days just to make sure it's > not unstable/crashy, I'll push these patches to the tree and re-enable > USE="jit". It's been rather stable so far. And version 36 seems to have resolved at least four issues I've had for the couple of previous releases: * Passwords not being filled on forms (not even a master password prompt); * Multiple add-ons failing to install / upgrade / run properly; * Middle-click on the "Back" and "Forward" buttons not working; * Ctrl+UP/DOWN not working in search box. So overall my experience is better than with version 35. But because you asked, I also ran a few JavaScript benchmarks and tests and noticed two crashes in X11 code. The crash with XCreatePixmap can be reproduced by using an empty profile and just opening a new tab. I've attached some backtraces I collected from these crashes. Note that firefox36btfull.txt is probably the same bug as firefox36btfull_3*. Another bug I noticed is that running the http://dromaeo.com/ JavaScript tests and moving that tab to a new window during some DOM test caused the testing to stop for some reason. But I guess that might be due to the DOM changing due to the move?
More related upstream bugs: * https://bugzilla.mozilla.org/show_bug.cgi?id=1138180 * https://bugzilla.mozilla.org/show_bug.cgi?id=1108834 While I'm a it: https://bugzilla.mozilla.org/show_bug.cgi?id=1024038 suggests there's no more YARR in Firefox 36.0, so --disable-yarr-jit is useless in the mozconfig-v5.36 eclass.
Firefox-36.0-r1 committed to the tree with IUSE="jit" re-added. I'll clean up --disable-yarr-jit in the mozilla-overlay for launch with ff37 et. al.