1667507 - Relax a failure-prone layout assertion "MOZ_ASSERT(kidOverflowBEnd >= kidBEnd)" to be non-fatal

Status: RESOLVED FIXED

(Core :: Layout: Positioned, task)

Description

[Daniel Holbert [:dholbert]]

We have assertion that's known to fail on real sites & which isn't actually a sign of any imminent crash or other catastrophe, as far as I can tell:

        MOZ_ASSERT(kidOverflowBEnd >= kidBEnd);

https://searchfox.org/mozilla-central/rev/9fab6c0021ed9b103a70bf5c296b45949a2950bc/layout/generic/nsAbsoluteContainingBlock.cpp#202

See bug 1373898 and bug 1666888 for cases where the assertion fails in the wild. (We have reduced testcases on the former case.)

Ideally we should investigate and see why it's failing & fix the logic to prevent the failure; but until someone has cycles to do that, let's just make it non-fatal, because there's not really any reason for it to be fatal and having it be fatal blocks people from investigating more-severe issues on the affected sites. (& per reasoning at https://robert.ocallahan.org/2011/12/case-for-non-fatal-assertions.html )

Comment 1

[Daniel Holbert [:dholbert]]

Attachment: Bug 1667507: Downgrade a positioned-layout assertion from fatal to non-fatal, since it doesn't lead to crashes/catastrophe and we know of web content that triggers it. r?mats

Comment 2

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/e8e82eb1b9f0
Downgrade a positioned-layout assertion from fatal to non-fatal, since it doesn't lead to crashes/catastrophe and we know of web content that triggers it. r=TYLin

Comment 3

[Cosmin Sabou [:CosminS]]

https://hg.mozilla.org/mozilla-central/rev/e8e82eb1b9f0