1285237 - Intermittent test_conformance__extensions__ext-shader-texture-lod.html | application crashed [@ js::LifoAlloc::getOrCreateChunk]

Status: RESOLVED FIXED

(Core :: JavaScript Engine, defect, P3)

Description

[Treeherder Bug Filer]

Filed by: tomcat

https://treeherder.mozilla.org/logviewer.html#?job_id=10345767&repo=fx-team

https://queue.taskcluster.net/v1/task/LScEx9t2TsKFhjZeDM7xuQ/runs/0/artifacts/public%2Flogs%2Flive_backing.log

Comment 1

[Nicolas B. Pierron [:nbp]]

Is this a debug-only thing?

This is likely an assertion that I added (Bug 1264948 part 2) to ensure that we do not blindly assume that LifoAlloc are infallible, but I did not expected to see that in the Parser, but only in IonMonkey's code.

Usually the stack trace is enough to investigate such issues, but this might be a little tricky as I did not expected that Parser to have a LifoAlloc with a flag set for using this assertion.

 11:28:39     INFO -   0  libxul.so!js::LifoAlloc::getOrCreateChunk [LifoAlloc.h:4a8d8898e507 : 71 + 0x0]
 11:28:39     INFO -   1  libxul.so!js::frontend::ParseNodeAllocator::allocNode [LifoAlloc.h:4a8d8898e507 : 225 + 0x7]
 11:28:39     INFO -   2  libxul.so!js::frontend::Parser<js::frontend::FullParseHandler>::newName [FullParseHandler.h:4a8d8898e507 : 33 + 0x3]
 11:28:39     INFO -   3  libxul.so!js::frontend::Parser<js::frontend::FullParseHandler>::identifierName [Parser.cpp:4a8d8898e507 : 8940 + 0x7]
 11:28:39     INFO -   4  libxul.so!js::frontend::Parser<js::frontend::FullParseHandler>::primaryExpr [Parser.cpp:4a8d8898e507 : 9571 + 0x7]
 11:28:39     INFO -   5  libxul.so!js::frontend::Parser<js::frontend::FullParseHandler>::memberExpr [Parser.cpp:4a8d8898e507 : 8753 + 0x11]
 11:28:39     INFO -   6  libxul.so!js::frontend::Parser<js::frontend::FullParseHandler>::unaryExpr [Parser.cpp:4a8d8898e507 : 8282 + 0x19]
11:28:39 INFO - 7 libxul.so!js::frontend::Parser<js::frontend::FullParseHandler>::orExpr1 [Parser.cpp:4a8d8898e507 : 7758 + 0xd] 

My blind guess would be that this is somewhat related to AsmJS.

Comment 2

[Nicolas B. Pierron [:nbp]]

Attachment: Consider ParseNodeAllocator::allocNode as fallible.

This patch add a debug-only AutoFallibleScope, what this does, is that it
set a flag on the LifoAlloc to mention that we are checking the return value
of it.  Thus, it disables the assertion in getOrCreateChunk added in Bug
1264948 part 2, and replaces it by the oomTest-able kind of allocation.

This mechanism got added as a way to ensure that IonMonkey does not rely too
much on the underlying ability to request more pages, without checking for
OOM, while also prevent oomTest-able failures on allocations which are
assumed to be infallible.

This issue likely comes from the Parser instances created by AsmJS
modules[1], which probably re-use the LifoAlloc create with the
TempAllocator[2], which has the fallibleScope_ flag[3] turned off by default.

[1] http://searchfox.org/mozilla-central/source/js/src/asmjs/AsmJS.h#33
[2] http://searchfox.org/mozilla-central/source/js/src/asmjs/WasmGenerator.h#120
[3] http://searchfox.org/mozilla-central/source/js/src/ds/LifoAlloc.h#166

Comment 3

[BMO Automation]

Bulk assigning P3 to all open intermittent bugs without a priority set in Firefox components per bug 1298978.

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/60843e550836
Consider ParseNodeAllocator::allocNode as fallible. r=jorendorff

Comment 5

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/60843e550836

Comment 6

[Ryan VanderMeulen [:RyanVM]]

Seems like a pretty trivial patch. Should we consider it for uplift?

Comment 7

[Nicolas B. Pierron [:nbp]]

Comment on attachment 8769688 [details] [diff] [review]
Consider ParseNodeAllocator::allocNode as fallible.

Approval Request Comment
[Feature/regressing bug #]: Bug 1264948
[User impact if declined]: Some intermittent remaining on treeherder
[Describe test coverage new/current, TreeHerder]: treeherder
[Risks and why]: None, because it is simple.
[String/UUID change made/needed]: N/A

Comment 8

[Ritu Kothari (:ritu) (Inactive, please n-i to RyanVM, jcristau, or pascal)]

Comment on attachment 8769688 [details] [diff] [review]
Consider ParseNodeAllocator::allocNode as fallible.

Fixes an intermittent, Aurora51+, Beta50+

Comment 9

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-aurora/rev/4631a67ac611

Comment 10

[Ryan VanderMeulen [:RyanVM]]

https://hg.mozilla.org/releases/mozilla-beta/rev/122bf25cc33c