1444238 - Crash in @0x407c84

Status: RESOLVED DUPLICATE of bug 1442718

(Core :: Graphics: WebRender, defect)

Description

[Andrew McCreight (out of office until 8/21) [:mccr8]]

This bug was filed from the Socorro interface and is
report bp-10c576f8-78be-4877-9509-aad310180307.
=============================================================

Top 10 frames of crashing thread:

0  @0x407c84 
1  @0x407c4d 
2  @0x7fe0b7172ad8 
3  @0x7fe0b7172467 
4  @0x7fe0b7039658 
5  @0x7fe0b70cf40e 
6  @0x7fe0b70db20f 
7  @0x7fe0b70b494b 
8  @0x7fe0b70b423b 
9  @0x7fe0b70b423b 

=============================================================

It seems bad that these stacks are all unsymbolized, but I did notice that they are almost all hitting an assertion that appears to be in WebRender code:
  assertion failed: descriptor.width > 0 && descriptor.height > 0

Specifically, this assertion: https://searchfox.org/mozilla-central/rev/588d8120aa11738657da93e09a03378bcd1ba8ec/gfx/webrender/src/texture_cache.rs#695

Comment 1

[Andrew McCreight (out of office until 8/21) [:mccr8]]

These crashes are also from a number of different install times, so it seems less likely that it is just one user with a broken machine.

Comment 2

[(not currently active) Ted Mielczarek]

The modules tab is empty, which likely means that something is causing us to not be able to read the module list from the minidump. We hit a similar issue in bug 1400952.

Comment 4

[Darkspirit]

https://crash-stats.mozilla.com/search/?moz_crash_reason=~assertion%20failed%3A%20descriptor.width%20%3E%200%20%26%26%20descriptor.height%20%3E%200&date=%3E%3D2018-03-02T00%3A26%3A46.000Z&date=%3C2018-03-09T00%3A26%3A46.000Z&_sort=-date&_facets=signature&_columns=date&_columns=signature&_columns=product&_columns=version&_columns=build_id&_columns=platform#facet-signature

Comment 5

[(not currently active) Ted Mielczarek]

I downloaded the dump from comment 0 and jeez, yeah:
2018-03-09 11:43:56: minidump.cc:2648: ERROR: MinidumpModuleList count 2969 exceeds maximum 2048

Comment 6

[(not currently active) Ted Mielczarek]

This actually looks very similar to the issue froydnj hit in bug 1400952. If I dump the /proc/self/maps stream from the minidump, it shows that there are 2293 entries containing /dev/shm.