Intermittent tests\jit-test\jit-test\tests\basic\bug1296249.js | Timeout (code 1, args "--ion-eager --ion-offthread-compile=off --non-writable-jitcode --ion-check-range-analysis --ion-extra-checks --no-sse3 --no-threads")

RESOLVED FIXED in Firefox 51

Status

()

defect
P1
normal
RESOLVED FIXED
3 years ago
3 years ago

People

(Reporter: intermittent-bug-filer, Assigned: sandervv)

Tracking

({intermittent-failure})

unspecified
mozilla52
Points:
---
Dependency tree / graph

Firefox Tracking Flags

(firefox50 unaffected, firefox51 fixed, firefox52 fixed)

Details

Attachments

(1 attachment, 1 obsolete attachment)

Duplicate of this bug: 1308668
It seems that the testcase added in bug 1296249 takes a lot of time.
Can we make it quicker by decreasing the number of items it needs to allocate and still have the testcase triggering on the issue that was fixed by that bug?
Blocks: 1296249
Flags: needinfo?(sandervv)
Priority: -- → P1
Since the bug happens on windows8-64 only, and the link you gave suggests that it only a problem on the debug builds, I found this info about malloc on windows8 64-bit and debug builds: https://msdn.microsoft.com/en-us/library/974tc9t1.aspx

"The debug heap also fills new memory blocks with a known value." (Note the debug heap is used when _DEBUG is not defined.)

I think that memset()ing the huge amount of memory during debug mode is causing the timeout. What do you think?

I think that the fix is to disable memset()ing huge memory allocations by using _IGNORE_BLOCK during _malloc_dbg() while in debug mode on windows 8 64-bit: https://msdn.microsoft.com/en-us/library/974tc9t1.aspx#BKMK_Types_of_blocks_on_the_debug_heap
Flags: needinfo?(sandervv) → needinfo?(hv1989)
(In reply to Sander Mathijs van Veen from comment #4)
> I think that memset()ing the huge amount of memory during debug mode is
> causing the timeout. What do you think?

That totally makes sense. The extra things in debug builds probably takes a lot more time. Not sure we should disable those extra "checks". If we could decrease the runtime while still testing the "bug" this revision fixed, we would have both. Else we might just add it to the long-running list.
Flags: needinfo?(hv1989)
Added js function that we discussed on IRC.

Try job results are good: https://treeherder.mozilla.org/#/jobs?repo=try&revision=88a9035861e3e75ba35390ae40623569fff8b562
Although I'm not sure if I've selected the right try job options to execute the jit tests on windows 8.
Assignee: nobody → sandervv
Attachment #8800768 - Flags: review?(hv1989)
Comment on attachment 8800768 [details] [diff] [review]
bug1308647-disable-malloc-memset-on-windows-debug.patch

Review of attachment 8800768 [details] [diff] [review]:
-----------------------------------------------------------------

Nice!
Attachment #8800768 - Flags: review?(hv1989) → review+
Win8 debug on that Try push doesn't look amused.
Marked test case as slow, as discussed on IRC.
Attachment #8800768 - Attachment is obsolete: true
Attachment #8801661 - Flags: review?(hv1989)
Attachment #8801661 - Flags: review?(hv1989) → review+
Keywords: checkin-needed
https://hg.mozilla.org/mozilla-central/rev/48955fa87451
Status: NEW → RESOLVED
Closed: 3 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla52
Whiteboard: [checkin-needed-aurora]
You need to log in before you can comment on or make changes to this bug.