The Fuchsia test runner script, which calls qemu, is crashing hard (30% of the time?) in a manner that can't easily be caught with Python exceptions. It also renders the Python test runner's stdout and stderr inoperable (adding print debugging statements generates "[Errno 11] Resource temporarily unavailable)" error spew.

When the crashes occur, they tend to happen *around* test case #1320 (not necessarily the same entry). The test cases themselves don't appear to be the root cause of the error, since they can be run in isolation repeatedly without any issue.

I'm suspecting  that this is a POSIX pipe producer/consumer problem in the qemu configuration, or a bug in qemu itself. I have a local patch which routes the qemu terminal output out to a named pipe in the host OS. The issue ceases to repro under that configuration.


[00009.171] 02526.02752> [1314/2544] StringPiece16Test.CheckConversion (0 ms)
[00009.171] 02526.02752> [1315/2544] SplitStringIntoKeyValuePairsTest.EmptyString (0 ms)
[00009.198] 02526.02752> [1316/2544] SplitStringPieceUsingSubstrTest.TrimWhitespace (0 ms)
[00009.198] 02526.02752> [1317/2544] SplitStringPieceUsingSubstrTest.SplitWantAll (0 ms)
[00009.198] 02526.02752> [1318/2544] SplitStringPieceUsingSubstrTest.SplitWantNonEmpty (0 ms)
[00009.198] 02526.02752> [1319/2544] StringSplitTest.StringSplitKeepWhitespace (0 ms)
[00009.198] 02526.02752> [1320/2544] StringSplitTest.SplitStringAlongWhitespace (0 ms)
close failed in file object destructor:ringTokenizerTest.Simple (0 ms)
sys.excepthook is missing
lost sys.stderr

[00006.272] 02466.02599> [1320/2544] SplitStringIntoKeyValuePairsTest.EmptyValue (0 ms)
close failed in file object destructor:litStringIntoKeyValuePairsTest.UntrimmedWhitespace (0 ms)
sys.excepthook is missing
lost sys.stderr


[00005.439] 02389.02537> [1319/2544] SplitStringUsingSubstrTest.ConsecutiveDelimitersSkipped (0 ms)
close failed in file object destructor:litStringUsingSubstrTest.TrailingDelimitersSkipped (0 ms)
sys.excepthook is missing
lost sys.stderr