Closed
Bug 1136040
Opened 10 years ago
Closed 10 years ago
Latest Chromium sandboxing code update (bug 1102195) uses C++ features not available in gcc 4.6.
Categories
(Core :: Security: Process Sandboxing, defect)
Core
Security: Process Sandboxing
Tracking
()
RESOLVED
WONTFIX
People
(Reporter: bobowen, Assigned: bobowen)
References
Details
We still officially support gcc 4.6, which Chromium has already dropped.
This means that, for example, they now directly use "override" which is not supported in gcc 4.6.
Comment 1•10 years ago
|
||
I suggest that you propose dropping GCC 4.6 support on dev-platform. I will support you in that. The only (AFAICT) reason we support GCC 4.6 is so that a few people who build Firefox for Linux distros don't have to install a non-default compiler. But, that's not a good enough reason. Also, it's unreasonable to expect us to maintain GCC 4.6 support when GCC 4.6 isn't used in any treeherder builds.
Assignee | ||
Comment 2•10 years ago
|
||
(In reply to Brian Smith (:briansmith, :bsmith, use NEEDINFO?) from comment #1)
> I suggest that you propose dropping GCC 4.6 support on dev-platform. I will
> support you in that. The only (AFAICT) reason we support GCC 4.6 is so that
> a few people who build Firefox for Linux distros don't have to install a
> non-default compiler. But, that's not a good enough reason. Also, it's
> unreasonable to expect us to maintain GCC 4.6 support when GCC 4.6 isn't
> used in any treeherder builds.
Thanks Brian
Yeah, I think I'm going to have to go down that route.
The code also uses template aliases, which are more difficult to fix than the simple override and final #defines.
There was some support in #build for doing this after Fx38 (as it's an ESR), which would still mean backing out and re-landing, but I can live with that.
Comment 3•10 years ago
|
||
It is better to do it *before* Fx38 because it will make it easier to backport changes to ESR 38.
Note that when we created ESR, one of the things we agreed on was that we wouldn't delay changes due to ESR. Lots of people objected to creating ESR in the first place and that was one of the compromises that was made to make the ESR concept tolerable.
Assignee | ||
Comment 4•10 years ago
|
||
Just to bring this bug up to date, I'm trying to get some idea from some Linux distro Firefox packagers, over how painful dropping gcc-4.6 for Fx38+ would be.
We would still keep a min version of 4.7.
Thanks to glandium for talking me through this and for the contacts.
The story so far:
* Debian - glandium said wheezy has 4.7 available and so he would be OK.
* openSUSE - Wolfgang Rosenauer has confirmed that he doesn't see any issues. 4.7 is available on all supported versions.
* SLES - Petr Cerny has confirmed that on older systems they already maintain a special stack of packages for building Firefox, which contains gcc-4.7 (since we dropped support for 4.4).
* Ubuntu - waiting to hear from Benjamin Kerensa and Chris Coulson. glandium spotted that there is a gcc-snapshot package available for Ubuntu 12.04, which appeared to be 4.8 when I installed it.
* RedHat - waiting to hear from Martin Stransky.
If we decide we can't drop 4.6 for Fx38, I've got m-c compiling on Unbuntu 12.04 using gcc-4.6 with:
* bug 1134487 - the patch already posted.
* bug 1134511 - '-pedantic-errors' removed for now (might have to look at #ifdefs around this).
* this bug - one base file and the Linux files rolled back to the last stable release for Chromium 39.
I think we would then want to look at dropping for Fx39 and we could undo some of the above.
I'll wait and see what I hear from Ubuntu and RedHat before getting wider views.
Assignee | ||
Comment 5•10 years ago
|
||
Heard from Chris Coulson over Ubuntu today.
He doesn't see any major issues, with us dropping 4.6 as long as there are no runtime problems when compiled with 4.7 or later.
The latest Nightly seems to run fine on 12.04, so this doesn't seem to be an issue.
He also mentioned that this might be a good time to look into backporting gcc-4.8 to 12.04.
He's going to have an initial look at how much work that might be next week.
CCing glandium and sylvestre, so you know where I am up to with this.
Comment 6•10 years ago
|
||
(Note: if we do bump our required GCC version, we'll probably need to update mfbt/Compiler.h in the same way that we did in bug 1120620.)
Assignee | ||
Comment 7•10 years ago
|
||
Heard from Martin Stransky yesterday over Red Hat.
gcc on both RHEL5 and RHEL6 are 4.4, so he didn't see that moving to 4.7 was going to be any more painful than moving to 4.6.
I've also mailed the mozilla-linux-taskforce@kuix.de list, which was suggested by Petr Cerny.
Assignee | ||
Comment 8•10 years ago
|
||
With bug 1144155 landed and uplifted, as of Gecko 38, gcc-4.7 is now officially our minimum supported version.
Status: ASSIGNED → RESOLVED
Closed: 10 years ago
Resolution: --- → WONTFIX
Assignee | ||
Updated•9 years ago
|
You need to log in
before you can comment on or make changes to this bug.
Description
•