Crash in js::jit::BaselineScript::nativeCodeForPC

NEW
Unassigned

Status

()

P3
normal
a year ago
a month ago

People

(Reporter: 6lobe, Unassigned)

Tracking

({nightly-community})

55 Branch
x86_64
Linux
nightly-community
Points:
---

Firefox Tracking Flags

(firefox60 fix-optional)

Details

(crash signature)

(Reporter)

Description

a year ago
This bug was filed from the Socorro interface and is 
report bp-ea445ca1-3f30-45b3-9ddd-be58b0170527.
=============================================================

I can consistently reproduce this crash on two different computers.

Go to https://www.reittiopas.fi/ and just use the site for a while and you'll get this crash sooner or later.


I also got the following crash signature once:

js::jit::BailoutIonToBaseline
https://crash-stats.mozilla.com/report/index/b974e836-cbf5-4268-ab10-75c130170527
(Reporter)

Updated

a year ago
Has STR: --- → yes
Keywords: nightly-community

Comment 1

a year ago
It doesn't crash for me on Win 7.
This should be actionable.  How urgent is this, and how high of a crasher is this?  It seems pretty low-frequency.  If we can get a repro for this it should be very actionable.
Flags: needinfo?(kvijayan)
(In reply to Kannan Vijayan [:djvj] from comment #3)
> This should be actionable.

Can you detail what would be the conditions needed to reach this assertion, and how fixing this issue should be approached?
Flags: needinfo?(kvijayan)

Comment 5

4 months ago
Could be related to running the gecko-profiler? 

https://crash-stats.mozilla.com/report/index/d0349b47-46f6-4a07-b6ef-c65470180627
(In reply to Paul Bone [:pbone] from comment #5)
> Could be related to running the gecko-profiler? 

Maybe! I just ran into this crash, just after capturing a profile with perf-html.io.  My crash report is bp-d16a027c-d908-49e8-a55f-1b5b10180913 .

The crash happened just after the profile was displayed -- possibly while resolving symbols.
The crash is almost definitely related to jit-coach, and native stack walking.  These crashes are basically impossible to reproduce and happen randomly.

The proper resolution is to implement bug 1426134 and get rid of the ad-hoc, extremely fragile stack-walking that explicitly keeps itself aware of the stack structure of every type of frame.  This is just extremely prone to bitrot, and crashes are very hard to diagnose.
Flags: needinfo?(kvijayan)
You need to log in before you can comment on or make changes to this bug.