Closed Bug 867856 Opened 11 years ago Closed 11 years ago

Intermittent auto-regress/bug743094.js | --ion-eager

Categories

(Core :: JavaScript Engine, defect)

Product:
Core
Component:
JavaScript Engine
Platform:
x86_64
macOS
Type:
defect
Priority:
Not set
Severity:
normal

Tracking

()

Status:
RESOLVED FIXED
Milestone:
mozilla25
Tracking Flags:
Tracking Status
firefox24 --- fixed
firefox25 --- fixed

People

(Reporter: philor, Assigned: nbp)

Details

(Keywords: intermittent-failure)

Attachments

(1 file)

Fix intermittent timeout of bug743094.js test case.
902 bytes, patch
h4writer
: review+
Details | Diff | Splinter Review 
https://tbpl.mozilla.org/php/getParsedLog.php?id=22478229&tree=Mozilla-Inbound
OS X 10.7 64-bit mozilla-inbound leak test build on 2013-05-01 14:08:34 PDT for push 47ad2545d5f5
slave: bld-lion-r5-083

FAIL - /builds/slave/m-in-osx64-d-00000000000000000/build/js/src/jit-test/tests/auto-regress/bug743094.js
TEST-UNEXPECTED-FAIL | /builds/slave/m-in-osx64-d-00000000000000000/build/js/src/jit-test/tests/auto-regress/bug743094.js | --ion-eager

(In the unlikely event that anyone ever was going to investigate any of these intermittent failures... how? With that utterly expressive and detailed logging, how?)
Assignee: general → nicolas.b.pierron
Status: NEW → ASSIGNED
The test case was used to make sure that we are correctly marking everything.

As the test case is running for a long time I limited the number of iterations such as in the worst case we still do a GC of a  compiled instance which inlined it-self.

Checking with 10000, show that we recompile this function ~100 times before the end of the test case.
Attachment #775964 - Flags: review?(hv1989)
Attachment #775964 - Flags: review?(hv1989) → review+
https://hg.mozilla.org/integration/mozilla-inbound/rev/b45bdd149e34

This time, I removed the expected failure from the test case, as we are no longer supposed to do a stack overflow.
(In reply to Ryan VanderMeulen [:RyanVM UTC-4] from comment #30)
> Backed out again.

Ok, this problem seems to be isolated to the fact that the baseline compiler is consuming way more stack space than ion monkey.  So if we are stuck in a recursive baseline-compiled function, then we waste a lot of stack space quickly.

I will reduce the counter to 5000 to make the test pass with optimized builds.

The problem does not appear in debug builds, because we are increasing the stack space to make the test pass, but apparently we do not enlarge it enough for optimized builds.
(In reply to Nicolas B. Pierron [:nbp] from comment #31)
> (In reply to Ryan VanderMeulen [:RyanVM UTC-4] from comment #30)
> > Backed out again.
> 
> I will reduce the counter to 5000 to make the test pass with optimized
> builds.

This time, I'll send it to try …

https://tbpl.mozilla.org/?tree=Try&rev=36a75f6dea80
Ok, lowering number of recursion again to get the test pass on 32 bits builds.

https://tbpl.mozilla.org/?tree=Try&rev=2676e625c63b

Strangely, this push to try fails the B2G builds on fedora, which does not make any sense knowing that this is just a test modification on top of the last push to try, where it is working …

So, I am sending it to inbound again, and pray for it to pass.

https://hg.mozilla.org/integration/mozilla-inbound/rev/d2aab866cf34
https://hg.mozilla.org/mozilla-central/rev/d2aab866cf34
Status: ASSIGNED → RESOLVED
Closed: 11 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla25
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: