Summary: | www-client/chromium with gcc-6 - null pointer dereference in V8 | ||
---|---|---|---|
Product: | Gentoo Linux | Reporter: | Alexey Shvetsov <alexxy> |
Component: | Current packages | Assignee: | Chromium Project <chromium> |
Status: | RESOLVED FIXED | ||
Severity: | normal | CC: | chutzpah, const, erikdenstore+gbugs, qt, soap |
Priority: | Normal | ||
Version: | unspecified | ||
Hardware: | All | ||
OS: | Linux | ||
URL: | https://bugs.chromium.org/p/v8/issues/detail?id=3782 | ||
Whiteboard: | |||
Package list: | Runtime testing required: | --- | |
Bug Depends on: | |||
Bug Blocks: | 582084 |
Description
Alexey Shvetsov
2016-07-11 12:13:38 UTC
Could you put this build flag to the BUILD file, but just for V8 (that is, as localized as possible)? This (as you might expect) also affects qtwebengine. Qupzilla crashes at seemingly random times when processing javascript. I was unable to reproduce with sys-devel/gcc-6.2.0-r1 from portage (and chromium-56.0.2897.0). I made sure to gcc-config switch to it and verified with gcc -v. Is anyone else getting a different result? If so make sure to specify what page you navigate to. I've tried e.g. with Octane JS benchmark, which I'm pretty sure has some JS. I can not reproduce it either. Spoke too soon. It sometimes happens with some complex webpages. The practical workaround is to immediately switch tab after pressing enter in the location bar, and then switch back when page loads. Weird. Ctrl + PgDn is the shortcut for that. I've been able to reliably reproduce it, pretty much any website using JS would trigger a crash within about a second. Novadays, almost every page uses js. The pages that crash are rare, but they exist. Facebook is one of them. "temporary workaround is to build chromium with -fno-delete-null-pointer-checks in C{,XX}_FLAGS" You have to first enable custom-cflags. commit 382082c70b9339a463f32024f5d9101f951108f5 Author: Mike Gilbert <floppym@gentoo.org> Date: Sun Nov 13 12:45:35 2016 -0500 www-client/chromium: dev channel bump (56.0.2914.3) Addresses a couple bugs: - Null pointer deref in V8 (bug 588596). - Crash with old harfbuzz (bug 599098). Bug: https://bugs.gentoo.org/588596 Bug: https://bugs.gentoo.org/599098 Package-Manager: portage-2.3.2_p5 www-client/chromium/Manifest | 2 +- ...m-56.0.2897.0.ebuild => chromium-56.0.2914.3.ebuild} | 17 ++++++++++++----- 2 files changed, 13 insertions(+), 6 deletions(-) commit aa24ca1549e5f2212320e6801b2c297cd7b4c72c Author: Mike Gilbert <floppym@gentoo.org> Date: Sun Nov 13 15:59:04 2016 -0500 www-client/chromium: beta channel bump (55.0.2883.44) Addresses a couple bugs: - Null pointer deref in V8 (bug 588596). - Crash with old harfbuzz (bug 599098). Bug: https://bugs.gentoo.org/588596 Bug: https://bugs.gentoo.org/599098 Package-Manager: portage-2.3.2_p5 www-client/chromium/Manifest | 2 +- .../{chromium-55.0.2883.28.ebuild => chromium-55.0.2883.44.ebuild} | 6 +++++- 2 files changed, 6 insertions(+), 2 deletions(-) commit 78047b6f5c093fdfaa64c19f47798fcb1b5591d4 Author: Mike Gilbert <floppym@gentoo.org> Date: Sun Nov 13 16:06:17 2016 -0500 www-client/chromium: fix null pointer deref in V8 with gcc-6 Bug: https://bugs.gentoo.org/588596 Package-Manager: portage-2.3.2_p5 www-client/chromium/chromium-54.0.2840.100.ebuild | 3 +++ 1 file changed, 3 insertions(+) I am seeing this problem (or a very similar one) with Chromium 56 as well. This thread has more information: https://forums.gentoo.org/viewtopic-t-1057444.html If a new bug should be filed, I can take care of that. Cheers, Nathan Zachary (In reply to Nathan Zachary from comment #11) All versions of chromium in the tree are built with -fno-delete-null-pointer-checks. If you are encountering an error, it is likely different from the one reported here. |