1503411 - Crash in ProfileBuffer::SetCapacityPow2 with crash reason being "MOZ_RELEASE_ASSERT(Length() <= aNewCapacity) (can't make the capacity smaller than the used size)"

Status: RESOLVED FIXED

(Core :: Gecko Profiler, defect)

Description

[Daniel Holbert [:dholbert]]

I've hit this crash several times today after leaving the profiler running.  Looks like this is a fatal assertion which was added recently in bug 1476757.


The failing assertion (as reported in the crash report) is:
MOZ_RELEASE_ASSERT(Length() <= aNewCapacity) (can't make the capacity smaller than the used size)

I had "buffer size" dialed all the way up in the profiler UI, for what it's worth (to 900MB I think). I've turned it down now to hopefully stop crashing.


This bug was filed from the Socorro interface and is
report bp-a880c0b6-c2c5-4fab-bdaa-fc0530181030.
=============================================================

Top 5 frames of crashing thread:

0 libxul.so ProfileBuffer::SetCapacityPow2 tools/profiler/core/ProfileBuffer.cpp:63
1 libxul.so SamplerThread::Run tools/profiler/core/platform.cpp:2320
2 libxul.so ThreadEntry tools/profiler/core/platform-linux-android.cpp:396
3 libpthread-2.28.so libpthread-2.28.so@0x8163 
4 libc-2.28.so libc-2.28.so@0x11adee 

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

Comment 1

[Daniel Holbert [:dholbert]]

Here are my crashes from this today:
bp-a880c0b6-c2c5-4fab-bdaa-fc0530181030 	10/30/18, 2:53 PM 	
bp-26758aae-4576-4c02-8f8e-add2e0181030 	10/30/18, 2:46 PM 	
bp-ebaa0eca-19dc-4a8f-bf2c-6e0f20181030 	10/30/18, 11:35 AM

Comment 2

[Nazım Can Altınova [:canova][:canaltinova on phabricator]]

We just decided to not go forward with that approach and backed out the relevant patches. It should be gone when it lands to central.

Comment 3

[Daniel Holbert [:dholbert]]

Great! I saw the backout -- wasn't sure if it was going to re-land (potentially with this same crash) after other issues had been addressed.  Sounds like it won't, though.

I'll just call this FIXED-by-backout then.

Comment 4

[Markus Stange [:mstange]]

This assertion failure was caused by integer overflow in this calculation:
uint32_t minDesiredCapacity = usedSize * 100 / 90;

(efaust told me about this bug last Friday and was going to submit a patch, but since the original change has been backed out, he doesn't need to anymore.)