Bug 206869

Summary:	Regression: JSC armv7 test failure stress/numberingSystemsForLocale-cached-strings-should-be-immortal-and-safe-for-concurrent-access.js.default
Product:	WebKit	Reporter:	Aakash Jain <aakash_jain>
Component:	Tools / Tests	Assignee:	Nobody <webkit-unassigned>
Status:	RESOLVED DUPLICATE
Severity:	Normal	CC:	ap, mark.lam, pmatos, ticaiolima, webkit-bot-watchers-bugzilla
Priority:	P2
Version:	Other
Hardware:	Unspecified
OS:	Unspecified

Description Aakash Jain 2020-01-28 05:36:25 PST

JSC test stress/numberingSystemsForLocale-cached-strings-should-be-immortal-and-safe-for-concurrent-access.js.default is failing on trunk for last few days.

e.g.: https://ews-build.webkit.org/#/builders/26/builds/7547

Comment 1 Aakash Jain 2020-01-28 05:37:14 PST

Most probably introduced by r255112 (https://bugs.webkit.org/show_bug.cgi?id=206779

Comment 2 Mark Lam 2020-01-28 09:05:56 PST

Already fixed in  r255120: <http://trac.webkit.org/r255120>. for https://webkit.org/b/206791.  Note that the jsc-armv7 bot of https://webkit.org/b/206791 is green.

*** This bug has been marked as a duplicate of bug 206791 ***

Comment 3 Aakash Jain 2020-01-28 09:24:53 PST

(In reply to Mark Lam from comment #2)
> Already fixed in  r255120: <http://trac.webkit.org/r255120>. for https://webkit.org/b/206791.

But the test is still failing. e.g.: https://ews-build.webkit.org/#/builders/26/builds/7542


> Note that the jsc-armv7 bot of https://webkit.org/b/206791 is green.
The green indicates that the patch did not introduce any new failure. If you click on the bubble https://ews-build.webkit.org/#/builders/26/builds/7407 (or look at any recent ARMV7 EWS build) it says: 'Found 1 pre-existing JSC test failure: stress/numberingSystemsForLocale-cached-strings-should-be-immortal-and-safe-for-concurrent-access.js.default'

Comment 4 Caio Lima 2020-01-28 09:29:47 PST

We are going to investigate this on https://bugs.webkit.org/show_bug.cgi?id=206864. We skipped the test for now.

Comment 5 Caio Lima 2020-01-28 09:30:19 PST


*** This bug has been marked as a duplicate of bug 206864 ***

Comment 6 Mark Lam 2020-01-28 09:41:47 PST

(In reply to Aakash Jain from comment #3)
> The green indicates that the patch did not introduce any new failure. If you
> click on the bubble https://ews-build.webkit.org/#/builders/26/builds/7407
> (or look at any recent ARMV7 EWS build) it says: 'Found 1 pre-existing JSC
> test failure:
> stress/numberingSystemsForLocale-cached-strings-should-be-immortal-and-safe-
> for-concurrent-access.js.default'

Thanks for the clarification.  I had forgotten about that.

(In reply to Caio Lima from comment #4)
> We are going to investigate this on
> https://bugs.webkit.org/show_bug.cgi?id=206864. We skipped the test for now.

Skipping is probably the correct remedy for this.  The test creates 100 threads all running concurrently.  This would most likely exhaust the resources on a armv7 machine, and would probably abort / fail a RELEASE_ASSERT when it fails to create the threads.