[Buildbot-devel] Purpose of buildbot.status.web.tests?

[Brian Warner]

Marcus Lindblom wrote:
> and these are the WebStatus URLs:
> 
> /builders/$builder/builds/$buildnum/tests/$testname
> /builders/$builder/builds/$buildnum/tests
> 
> They look interesting and useful, but are they really used currently?


Probably not. The original idea was to have a well-defined notion of
"the set of tests that were run for this build", specifically a
dictionary mapping test name (perhaps a dot-separated hierarchy, like
the one that shows up in a lot of python unit test suites) to test
results (specifically an ITestResult -providing instance, where
.getResults() gives you SUCCESS or FAILURE, .getLogs() provides log
messages produced during that test, etc).

This was intended to support the "Problem Tracking" feature that I
always wanted but never got around to implementing. The idea would be to
record fine-grained machine-parseable test results, instead of just a
single full-suite SUCCESS/FAILURE and some unparseable logfiles. Then
other tools could ask questions like "how frequently does
buildbot.test.test_slaves.LatentSlave.testSequence fail?", or look for
transitions from passing to failing and correlate them with the Changes
that went into it.

I'd still love to see this sort of feature go in at some point, but I
gave up on actually doing it years ago. I think it'd be nice to reserve
some conceptual space for them, but I suspect that nobody is actually
using these interfaces, and as Dustin said, removing code is a good
thing. Yay negative code days :)

cheers,
 -Brian