Open Bug 1474567 Opened 6 years ago Updated 2 years ago

Console ipc_channel_win.cc error thrown on Windows

Categories

(Core :: IPC, defect, P3)

Unspecified
Windows
defect

Tracking

()

People

(Reporter: aflorinescu, Unassigned)

References

Details

(Keywords: good-first-bug)

[Affected versions]:
Buidl ID: 20180710001201 version 63.0a1

[Affected platforms]:
Windows 10, 8.1 x64 and probably win 7 as well - cannot test on win 7 due to bug 1474559

[Steps to reproduce]:
1. Download target.zip from https://tools.taskcluster.net/index/gecko.v2.mozilla-central.latest.firefox/win64-asan-opt
2. Open the build with a new profile
3. Observe the console output.

[Expected result]:
-

[Actual result]:
[GPU 7428, Chrome_ChildThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[GPU 7428, Chrome_ChildThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[6132, Gecko_IOThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[6132, Gecko_IOThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[6132, Gecko_IOThread] WARNING: pipe error: 232: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 513
[6132, Gecko_IOThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[GPU 7428, Chrome_ChildThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[Parent 4656, Gecko_IOThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[GPU 4700, Chrome_ChildThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[GPU 4700, Chrome_ChildThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[Parent 4656, Gecko_IOThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[GPU 4700, Chrome_ChildThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346
[GPU 4700, Chrome_ChildThread] WARNING: pipe error: 109: file z:/build/build/src/ipc/chromium/src/chrome/common/ipc_channel_win.cc, line 346

[Note:]
Most likely this issue should belong to Core:IPC, but I think it would be useful to investigate and exclude first that the cause above set of errors is clang-cl
I've seen this for a long time on my own asan builds and it has never caused a problem with the actual functionality of the browser.

Official builds don't have a console so maybe they're encountering this too and just not reporting it.
Environment:
Windows 10 Pro x64
Fx 63.0a1 20180705220111
Fx 63.0a1 20180711100118

Thanks David!
Using the -console arg., I can confirm that the ipc_channel_win.cc console error is not something specific to the ASAN builds and not something specific to the clang-cl either.

Based on the above, updating the Summary, removing the blocking and moving to core:general.
Leaving myself a NI for additional investigation.
Flags: needinfo?(adrian.florinescu)
Summary: Asan opt build throws ipc_channel_win.cc console error → Console ipc_channel_win.cc error thrown on Windows
Flags: needinfo?(adrian.florinescu)
Product: Firefox Build System → Core
No longer blocks: 1474546
Flags: needinfo?(adrian.florinescu)
Component: General → IPC
Flags: needinfo?(adrian.florinescu)
We sometimes see this on Unix, when one process crashes and the other process gets hung up on, or during shutdown when ordering is maybe a little confused.  But there seems to be a lot more of this on Windows, and I'm not sure why.

One thing about Unix sockets is that, if the remote end is closed, reading returns an end-of-file indication and writing returns a broken pipe error; if Windows named pipes are saying “broken pipe” in both directions that might explain this.  (We have different code implementing channels for Windows vs. Unix, but they both have similar “pipe error” log messages.)

I'm not so sure about the `good-first-bug` tag, though, because there's some subtlety: we still want to report unexpected errors, and sometimes the spray of pipe errors is the only thing in the log that tells us a process crashed.
See Also: → 1428509
Severity: normal → S3
You need to log in before you can comment on or make changes to this bug.