[PATCH 0 of 2] tests: hghave || exit 80

[Patrick Mézard]

Nicolas Dumazet a écrit :
> Hello!
> 
> We use in several tests "hghave $feature || exit 80" to check for avaibility
> of $feature before running a test: hghave exits with exitcode 0 only if
> $feature is availabe. If $feature is not available, hghave exits with 
> exitcode 1, which causes the test to exit with exitcode 80; and then run-tests 
> checks the test exitcode to show a "skipped" message.
> 
> If, however, "hghave $feature" crashes (Traceback, caused for example by
> a SyntaxError) it exits with exitcode 1, which in its turns exits the test with
> errorcode 80... and run-tests prints happily a "test-skipped" message instead
> of detecting a failure.
> 
> I have included two patches to avoid those issues:
> * The first one makes sure that all hghave targets are crash-free
> * The second one catches errors when hghave fails inside a test, to mark the
>   test as failed.
> 
> Those two patches might look overlapping, but both are necessary:
> 
> 1) when all hghave targets are tested for validity, a single test cannot 
>    indeed fail on a hghave call. However if one introduces an error in, 
>    say inotify, he will want test-inotify* tests to run (no skips), 
>    and will want them to fail: one should not have to run test-hghave to 
>    detect an error outside of hghave scope.
> 2) Simetrically, it is not because all individual tests calling hghave
>    run correctly that all hghave targets are correct.

In crew: http://hg.intevation.org/mercurial/crew/rev/41a2c5cbcb6a

Thanks!


--
Patrick Mézard