1409904 - ERROR:audioipc_server: server poll error: Interrupted system call (os error 4)

Status: RESOLVED FIXED

(Core :: Audio/Video: cubeb, enhancement, P2)

Description

[Jed Davis [:jld] ⟨⏰|UTC-7⟩ ⟦he/him⟧]

I found this error in a Try run testing media.cubeb.sandbox with some future sandboxing features: https://treeherder.mozilla.org/logviewer.html#?job_id=137914894&repo=try&lineNumber=3012

It looks like audioipc maybe isn't handling EINTR correctly; for comparison, see https://searchfox.org/mozilla-central/search?q=HANDLE_EINTR&path=

Comment 1

[Matthew Gregan [:kinetik]]

The current code logs a message at error priority, then calls poll again in a loop.  So I think this is right, but we should probably tweak or silence the logging to avoid confusion.

Comment 2

[Matthew Gregan [:kinetik]]

https://github.com/djg/audioipc-2/pull/21

Comment 3

[Dave Justice]

Seeing this error starting Saturday. Unsure If it was happening before, but now when I launch my executable I get this error and the crash report dialog. I can't launch firefox at all.

https://gist.github.com/meandavejustice/a0b3d27320e957c8f7bfc5d095f5de13

Nightly
Linux 4.4.0-98-generic

Comment 4

[Matthew Gregan [:kinetik]]

(In reply to Dave Justice from comment #3)
> Seeing this error starting Saturday. Unsure If it was happening before, but
> now when I launch my executable I get this error and the crash report
> dialog. I can't launch firefox at all.
> 
> https://gist.github.com/meandavejustice/a0b3d27320e957c8f7bfc5d095f5de13

Please file a new bug because this is unrelated to this bug.  It looks like the child content process is crashing for some reason, the audioipc message is being generated by the parent chrome process when the child dies and may be unrelated to the real crash reason.  Please include the crash reports from about:crashes in the bug report, if they're being generated, and the exact Nightly build version/date.

Comment 5

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/c0e806b7a992
Convert error! logging to warn! to avoid spurious stderr spamming from audioipc.  r=kamidphish

Comment 6

[Matthew Gregan [:kinetik]]

(In reply to Matthew Gregan [:kinetik] from comment #2)
> https://github.com/djg/audioipc-2/pull/21

Landed in https://hg.mozilla.org/integration/mozilla-inbound/rev/c0e806b7a992dc9e8f608df3e6380bce574eb791 with review carried forward from the linked PR.

Comment 7

[Narcis Beleuzu [:NarcisB]]

https://hg.mozilla.org/mozilla-central/rev/c0e806b7a992

Comment 8

[Jeff Muizelaar [:jrmuizel]]

I still get this error every time I 'step next' a debug Firefox build on Linux. An interrupted system call is usual enough occurrence that we shouldn't be logging it all. Since there's nothing we can do to really control it from happening or do to handle it other than recalling poll() we should not warn! about it.

Comment 9

[Matthew Gregan [:kinetik]]

Right, that's because Stylo is setting the default log level to warning in debug builds: https://dxr.mozilla.org/mozilla-central/rev/fc194660762d1b92e1679d860a8bf41116d0f54f/servo/ports/geckolib/glue.rs#172

But you're right that it's not worth logging EINTR here, so I'll change that.

Comment 10

[Matthew Gregan [:kinetik]]

https://github.com/djg/audioipc-2/pull/23

Comment 11

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/f79a12dc2c55
Don't bother logging when audioipc server's poll() hits EINTR.  r=kamidphish

Comment 12

[Noemi Erli[:noemi_erli]]

https://hg.mozilla.org/mozilla-central/rev/f79a12dc2c55