non-zero wait status: 9

[kentfredric]

Need to dig into this, its happened a few times, but the test output is useless.

http://www.cpantesters.org/cpan/report/c8a3a120-ad4c-11e5-b4f6-84cfe0bfc7aa

t/00-report-prereqs.t ........................................... ok
t/01-piecewise-xml.t ............................................ 
No subtests run 
t/02-piecewise-subclass-xml.t ................................... 
No subtests run 
t/03-piecewise-saxclass-xml.t ................................... 
No subtests run 

Test Summary Report
-------------------
t/01-piecewise-xml.t                                          (Wstat: 9 Tests: 0 Failed: 0)
  Non-zero wait status: 9
  Parse errors: No plan found in TAP output
t/02-piecewise-subclass-xml.t                                 (Wstat: 9 Tests: 0 Failed: 0)
  Non-zero wait status: 9
  Parse errors: No plan found in TAP output
t/03-piecewise-saxclass-xml.t                                 (Wstat: 9 Tests: 0 Failed: 0)
  Non-zero wait status: 9
  Parse errors: No plan found in TAP output

http://www.cpantesters.org/cpan/report/1e083e66-a5b2-11e5-8407-9788e0bfc7aa

[kentfredric]

Could be a dependency issue, but stats are vague:

http://analysis.cpantesters.org/solved?distv=IO-Async-XMLStream-SAXReader-0.001000#mod%3ACarp

[eserte]

Actually, wstat = 9 means that a "kill -9" was done. This is what I am usually doing if the tests are hanging for a long time.

[kentfredric]

I try to send SIGALARM to tests in that case when I spot them. Mostly beacuse the hope is that if somebody decodes that stat, they'll get "oh, so it took too long to run", as opposed to "maybe it got oom-killed or something"

[eserte]

... but a SIGALRM can be caught and ignored by the application, a SIGKILL not. However, maybe a possible workflow is to send SIGALRM first, and if nothing happens a SIGKILL later.

Or just send a mail to the report sender...

[kentfredric]

Sure. :) . But to an extent, that's on purpose.

If the code in question has SIGALRM handlers, it might do the right thing and only fail a part of the test, then continue the rest, maybe generating a diagnostic FAIL instead of a confusing "bail"

Kill gives no opportunity for graceful handling.

But I certainly will reach for kill if the test in question doesn't play nice with ALRM :)