817278 - ABORT: failed to construct LayersChild: file ... widget/xpwidgets/nsBaseWidget.cpp, line 877

Status: RESOLVED WORKSFORME

(Core :: Graphics: Layers, defect)

Description

[Myk Melez [:myk] [@mykmelez]]

Since November 22, Linux B2G Beta nightly builds die at startup on certain Linux installations running in VMware VMs with the following console message:

###!!! ABORT: failed to construct LayersChild: file ... widget/xpwidgets/nsBaseWidget.cpp, line 877


Working build from November 21:
  https://ftp.mozilla.org/pub/mozilla.org/b2g/nightly/2012-11-21-03-07-12-mozilla-beta/b2g-18.0.en-US.linux-i686.tar.bz2
  https://ftp.mozilla.org/pub/mozilla.org/b2g/nightly/2012-11-21-03-07-12-mozilla-beta/b2g-18.0.en-US.linux-i686.txt

Failing build from November 22:
  https://ftp.mozilla.org/pub/mozilla.org/b2g/nightly/2012-11-22-03-03-37-mozilla-beta/b2g-18.0.en-US.linux-i686.tar.bz2
  https://ftp.mozilla.org/pub/mozilla.org/b2g/nightly/2012-11-22-03-03-37-mozilla-beta/b2g-18.0.en-US.linux-i686.txt


The regression range (65e445cc6655:28a256a273f8) includes the fix for bug 799768, which seemed like a reasonable candidate; so I reverted that change and rebuilt, after which I no longer saw the problem.

I tested on CentOS 6 and Ubuntu 11.10 running in VMware Fusion 5.  Presumably the same problem exists on the other branches with that change and on other Linux distributions, although I haven't tested them.

In my one test on a physical machine (booting it into CentOS 6 via a LiveDVD), I couldn't reproduce the problem.  So it seems specific to VMs (at least so far in my testing; which reminds me of bug 776109).

Comment 1

[Karl Tomlinson (:karlt)]

Might be worth looking at "glxinfo | grep string".

Bug 816695 may cause similar consequences.

Comment 2

[Karl Tomlinson (:karlt)]

Does bug 799544 fit the regression range?

Comment 3

[Myk Melez [:myk] [@mykmelez]]

(In reply to Karl Tomlinson (:karlt) from comment #2)
> Does bug 799544 fit the regression range?

No, it's not in the range, as it landed afterward, and it appears to have landed only on Central, whereas this crash is on the Beta branch.

Comment 4

[Myk Melez [:myk] [@mykmelez]]

This is another of those development blockers that prevents all use of B2G Desktop on the Linux installations that are affected.  It isn't clear to me how many installations are affected by this bug (unlike related bug 816793, which clearly affects many installations), so I'm not sure this is a Basecamp blocker.

But it sure is hard to tell whether a given B2G Desktop build that RelEng or a developer produces will work for anyone else when it won't even work on the system that produced it (including on the CentOS 6 systems that RelEng uses to build Firefox and will use to build B2G Desktop in the future per bug 816793).

Thus I'm nominating this to block Basecamp.

Comment 5

[Jonas Sicking (:sicking) No longer reading bugmail consistently]

This *should* now be working on mozilla-central since we fixed bug 819714. I.e. we've basically disabled all the nice hardware acceleration stuff which should mean that we're not hitting these code paths.

So not blocking on this.

Not sure if it should even be marked as WFM until we start working on making hardware acceleration again

Comment 6

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

"Strategic retreat" is probably a better term than "fix" ;).

I used to prefer to leave this bugs open for someone to look at when we want to reenable later, but my experience has been that they just sit around and the underlying code tends to change often enough in the interim to make them irrelevant.  So closing.