1505853 - Unclear failure reason for js-bench-sm jobs

Status: NEW

(Firefox Build System :: General, defect)

Description

[Matthew Gaudet (he/him) [:mgaudet]]

(Guessing at component here, sorry if I missed) 

I have a try push with some patches [1] where all js-bench-sm jobs failed, but it's not clear at all why the jobs failed; 

- Some [taskcluster:error] exit status 245
- Some "UnboundLocalError: local variable 'bench_score' referenced before assignment

A push with no changes [2] the jobs succeed.

While its possible there's an error in patch [1], though I consider it unlikely, it's really hard to tell due to the lack of output. 

It seems like something might be improvable to help a failure in the job get diagnosed. 

[1]: https://treeherder.mozilla.org/#/jobs?repo=try&revision=66024b47ee0c858829600fc26e5e92005f91ef62
[2]: https://treeherder.mozilla.org/#/jobs?repo=try&revision=9ffab1d827c85bfb5a2f57883de9f940a90f05eb

Comment 1

[Dustin J. Mitchell [:dustin] (he/him)]

I'm guessing "sm" is spidermonkey?

It looks like this is a failure in the mach virtualenv bootstrapping process.  I'm not sure who would handle that.

Comment 2

[Gregory Szorc [:gps]]

From the log I looked at (https://taskcluster-artifacts.net/EiGz6xumSuWN0EYb0FObAg/0/public/logs/live_backing.log), the failure from the virtualenv should be non-fatal. But then nothing happens after it. It's almost as if the mach command no-ops after doing virtualenv things.

I'm not sure what's going on here :/

I would try to debug things by sprinkling print() statements in a try push inside the jsshell-bench mach command and see exactly where it is failing. You can also try an interactive loaner and setting Python breakpoints in the function. Add `import pdb; pdb.set_trace()` to a Python file to set a breakpoint and drop into a GDB-like debugger when that line is hit.