Bug 156350 – stress/computed-function-names.js.dfg-maximal-flush-validate-no-cjit is failing

• Status: ASSIGNED
• Component: JavaScriptCore

[Jiewen Tan]

JSC test: stress/computed-function-names.js.dfg-maximal-flush-validate-no-cjit is failing on this build:
https://build.webkit.org/builders/Apple%20El%20Capitan%2032-bit%20JSC%20%28BuildAndTest%29/builds/2020

These are the JSC changes included in this build, I am not sure which one is directly related the failing.
https://trac.webkit.org/changeset/199140
https://trac.webkit.org/changeset/199144

Error log:
stress/computed-function-names.js.dfg-maximal-flush-validate-no-cjit: test_script_9910: line 2:  7270 Segmentation fault: 11  ( "$@" ../../.vm/JavaScriptCore.framework/Resources/jsc --useFTLJIT\=false --useFunctionDotArguments\=true --validateGraph\=true --useMaximalFlushInsertionPhase\=true --useConcurrentJIT\=false --thresholdForJITAfterWarmUp\=100 computed-function-names.js )
stress/computed-function-names.js.dfg-maximal-flush-validate-no-cjit: ERROR: Unexpected exit code: 139
FAIL: stress/computed-function-names.js.dfg-maximal-flush-validate-no-cjit

[Mark Lam]

I'll take a first stab at this.

[Mark Lam]

I've tried running the test manually on a local debug build of ToT (r199166), but have absolutely no luck in reproducing this.

[Ryan Haddad]

(In reply to comment #2)
> I've tried running the test manually on a local debug build of ToT
> (r199166), but have absolutely no luck in reproducing this.

If relevant, it only appears to be failing on 32-bit testers,

[Mark Lam]

(In reply to comment #3)
> If relevant, it only appears to be failing on 32-bit testers,

I am testing on a 32-bit build.  No repro.  I'm currently hijacking the bot to see if it's introducing any special sauce into the mix that makes this happen.