1674203 - Assertion PR_GetPageSize() == alignof(ReadOnlyPage) on DTKs

Status: RESOLVED WORKSFORME

(Core :: XPConnect, task, P2)

Description

[Andrew Halberstadt [:ahal]]

+++ This bug was initially created as a clone of Bug #1659902 +++

We started hitting this assertion again when trying to run tests on the dtk machines:
https://firefoxci.taskcluster-artifacts.net/Qf8zMCupRQm36hePz-bGWg/0/public/logs/live_backing.log

The assertion led me to:
https://searchfox.org/mozilla-central/source/js/xpconnect/src/xpcpublic.h#673

and bug 1659902.

Comment 1

[Andrew Halberstadt [:ahal]]

Mike, is this something you could help out with? Looks like you've dealt with this before and are likely able to fix this much quicker than I can at least.

Comment 2

[Andrew Halberstadt [:ahal]]

Maybe we just need to expand the defines?
e.g, https://searchfox.org/mozilla-central/source/js/src/jit/AtomicOperations.h#379

Comment 3

[Mike Hommey [:glandium]]

You're running an x86-64 xpcshell aren't you?

Comment 4

[[e-mail] (:bhearsum)]

(In reply to Mike Hommey [:glandium] from comment #3)

You're running an x86-64 xpcshell aren't you?

I assume so, we haven't applied any changes to the build, eg: https://treeherder.mozilla.org/jobs?repo=try&revision=3b8b55a36abcc912b94c5e59a64c1853e3c965da

(ahal is out, responding in his stead)

Comment 5

[Andrew Halberstadt [:ahal]]

Thanks!

I was able to work around this with this patch:
https://hg.mozilla.org/try/rev/41aab6b5006e47864fecce0aa5146b168459217a

Obviously that can't land. Are we planning to use ARM64 based xpcshell builds in the proper pool? If so should we just dupe this to that bug? Is there a preprocessor if statement that could allow us to push to try without carrying around the above patch (while also not breaking anything else)?

Comment 6

[Mike Hommey [:glandium]]

With bug 1675740 now landed, xpcshell should be a universal executable too, so running it on the DTK should work now.