770991 - NSS tests stuck because bash.exe crashed

Status: RESOLVED WORKSFORME

(NSS :: Test, defect)

Description

[Kai Engert [:KaiE:]]

Today I saw the following stuck windows tinderbox build slave:

On a windows test machine, bash.exe had crashed while executing the ssl tests.

The calling part of the scripts noticed that SSL testing was "done",
the logfile had:
  TIMESTAMP ssl.sh END

However, the parts that were responsible for killing the selfserv process were skipped.

The remaining tests succeeded until the end.
However, the script waited for the subprocess (selfserv) to terminate, unaware that it needs to be terminated.


I don't know how frequent this kind of crash is.
I think this scenario is different from the other stuck windows machines I had seen in the past.

Do you have a good idea how to detect this scenario and react accordingly?

My only idea is to add a call like "killall -9 selfserv" to the script, but I don't know how to do "killall" on the windows bash environment, apparently it's not implemented (command not found).

BTW, the scripts had created a bash.exe.stackdump - but the contents weren't helpful, just a few lines with hexadecimal values, no symbols.

Comment 1

[Tim Taubert [:ttaubert] (inactive)]

Windows tinderbox builders are gone.