1253202 - EXCEPTION_ILLEGAL_INSTRUCTION crashes in _ZN6thread7Builder3new20h78612e3b213f1cf4IvbE (Rust MP4)

Status: RESOLVED FIXED

(Core :: Audio/Video: Playback, defect)

Description

[David Baron :dbaron: (⌚️UTC-5, no longer working on Mozilla)]

This bug was filed from the Socorro interface and is 
report bp-4ef38e6e-2f20-4275-8ec8-e15c32160302.
=============================================================

Following the fix for bug 1252041, it appears to have been replaced by a different topcrash with similar characteristics.

See the full set of such crashes at:
https://crash-stats.mozilla.com/signature/?product=Firefox&release_channel=nightly&date=%3E%3D2016-03-01&signature=_ZN6thread7Builder3new20h78612e3b213f1cf4IvbE

CPUs observed so far:
Rank 	Cpu info 	  	  	  	  	Count 	%
1 	AuthenticAMD family 6 model 8 stepping 1 | 1 	29 	80.56 %
2 	AuthenticAMD family 6 model 10 stepping 0 | 1 	7 	19.44 %

Comment 1

[(not currently active) Ted Mielczarek]

via c++filt, this is thread::Builder::new::h78612e3b213f1cf4Ivb.

I'm guessing this the same bug but in the pre-built libstd we're using?

Comment 2

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

Tracked since it's a top crasher.

Comment 3

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

Hi Xidorn, since you fixed bug 1252014, I thought I'd start with you. Could you please help investigate since this is a top crash? Thanks!

Comment 4

[Xidorn Quan [:xidorn] UTC+10]

That was fixed by rillian, not me :)

Comment 5

[Ralph Giles (:rillian)]

Hmph. I was just wondering about that. Sounds like we need to do a custom build of the standard library with sse2 generation turned off. :(

Comment 6

[Ralph Giles (:rillian)]

Alex, what do you think about adding an i585-pc-window-msvc (or i686+nosse2) rust-std target? This is another thing moving us away from being able to use rust-lang builds.

Comment 7

[Alex Crichton [:acrichto]]

Seems reasonable to me, we recently added an i586-unknown-linux-gnu target (https://github.com/rust-lang/rust/pull/31629) with the expectation that we'd add more. I'll work on getting some nightlies produced for this.

Comment 8

[Ralph Giles (:rillian)]

Excellent, thanks!

Comment 9

[Ralph Giles (:rillian)]

Alex's change is https://github.com/rust-lang/rust/pull/32034

Comment 10

[Ralph Giles (:rillian)]

Alex's change is now landed, but no builds yet and I'm away next week. Exploring runtime switching based on cpuid as a workaround in the meantime.

Comment 11

[Ralph Giles (:rillian)]

No crashes with a build id more recent than March 2 since re-enabling rust, so it looks like the runtime check in bug 1255869 is working. Leaving this open until we have the more robust i586-pc-windows-msvc standard library in place.

Comment 12

[Ralph Giles (:rillian)]

No return of this signature, and I don't see anything related in the top-crashers list, so confirming this is properly fixed now.

Comment 13

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

Updating status-fx47 to fixed based on comment 12.