993715 - Windows-only test failure in jit-test\tests\basic\bug970643.js - Script runs for too long, terminating. (code 0, args "--baseline-eager --no-fpu")

Status: NEW

(Core :: JavaScript Engine, defect, P5)

Description

[Jason Orendorff [:jorendorff]]

https://tbpl.mozilla.org/php/getParsedLog.php?id=37462480&tree=Mozilla-Inbound

The message "Script runs for too long, terminating." comes from the shell itself, not the test-runner, and it's supposed to happen in this test.

This is a failure because the test-runner expects exit code 6, and observes exit code 0. Not sure how that could happen. Windows-only flakiness in the subprocess module is always a possibility, unfortunately...

Comment 1

[Jan de Mooij [:jandem]]

This looks like the problem we had with Windows exit codes sometimes being 0 instead of 3.

It's not a problem with the test harness, the process is really returning code 0. We worked around it in another bug by accepting both 0 and 3 on Windows if we expect an error, but we should probably do the same for other non-zero exit codes like 6 :(

The underlying issue is the one I described in bug 970063 comment 7; Windows can terminate your process with code 0 if weird thread races etc happen during shutdown.