Bug 190426 – [JSC] [Armv7] stress/sampling-profiler-richards.js crashes

• Status: REOPENED
• Component: JavaScriptCore

[Guillaume Emont]

The test crashes on linux/armv7. When running the test manually, it only seems to crash with the sampling-profiler enabled.

[Guillaume Emont]

Committed r237000: <https://trac.webkit.org/changeset/237000>

[Guillaume Emont]

Reopening as the commit was only to skip the test.

[Radar WebKit Bug Importer]

<rdar://problem/45157301>

[Radar WebKit Bug Importer]

<rdar://problem/45157300>

[Tadeu Zagallo]

FYI, I've put a patch up in https://bugs.webkit.org/show_bug.cgi?id=191184 to convert these to `skip if no $jitTests`, since they were also timing out on x86. I've removed the FIXME comments, since it should no longer be necessary to change the tests, only the runner. Let me know if you want me to keep the comments and I'll revert the changes.

[Guillaume Emont]

(In reply to Tadeu Zagallo from comment #5)
> FYI, I've put a patch up in https://bugs.webkit.org/show_bug.cgi?id=191184
> to convert these to `skip if no $jitTests`, since they were also timing out
> on x86. I've removed the FIXME comments, since it should no longer be
> necessary to change the tests, only the runner. Let me know if you want me
> to keep the comments and I'll revert the changes.

I think we still need an explicit '@skip if $architecture == "arm"' for this test since I expect this crash to still be there once we reenable the JIT on arm (unless the bytecode changes accidentally fixed it).

[Guillaume Emont]

(In reply to Guillaume Emont from comment #6)
> (In reply to Tadeu Zagallo from comment #5)
> > FYI, I've put a patch up in https://bugs.webkit.org/show_bug.cgi?id=191184
> > to convert these to `skip if no $jitTests`, since they were also timing out
> > on x86. I've removed the FIXME comments, since it should no longer be
> > necessary to change the tests, only the runner. Let me know if you want me
> > to keep the comments and I'll revert the changes.
> 
> I think we still need an explicit '@skip if $architecture == "arm"' for this
> test since I expect this crash to still be there once we reenable the JIT on
> arm (unless the bytecode changes accidentally fixed it).

FWIW your patch on Bug 191184 already does the right thing for this test.