Merge lp:~lifeless/bzr/test-speed into lp:~bzr/bzr/trunk-old

Selftest was a bit tangled up. This makes it clearer without any
functional changes:

* Use startTestRun and stopTestRun, as upstream python does
* lift test listing code out of 'TestRunner' into the DWIM layer
* push all test result outputting down into ExtendedTestResult rather
  than being in the runner.

This means that the runner knows how to run, the DWIM layer knows when
to run, and Results know how to report on results.

Code-reviews will show the wrong diff though. Sorry - latency on trunk etc.

-Rob

=== modified file 'NEWS'
--- NEWS 2009-08-26 07:02:45 +0000
+++ NEWS 2009-08-27 03:51:02 +0000
@@ -79,6 +79,9 @@
   classes changed to manage lock lifetime of the trees they open in a way
   consistent with reader-exclusive locks. (Robert Collins, #305006)

+* ``bzrlib.tests`` now uses ``stopTestRun`` for its ``TestResult``
+ subclasses - the same as python's unittest module. (Robert Collins)
+
 Internals
 *********

=== modified file 'bzrlib/tests/__init__.py'
--- bzrlib/tests/__init__.py 2009-08-26 07:02:45 +0000
+++ bzrlib/tests/__init__.py 2009-08-27 03:51:02 +0000
@@ -175,17 +175,47 @@
         self._overall_start_time = time.time()
         self._strict = strict

- def done(self):
- # nb: called stopTestRun in the version of this that Python merged
- # upstream, according to lifeless 20090803
+ def stopTestRun(self):
+ run = self.testsRun
+ actionTaken = "Ran"
+ stopTime = time.time()
+ timeTaken = stopTime - self.startTime
+ self.printErrors()
+ self.stream.writeln(self.separator2)
+ self.stream.writeln("%s %d test%s in %.3fs" % (actionTaken,
+ run, run != 1 and "s" or "", timeTaken))
+ self.stream.writeln()
+ if not self.wasSuccessful():
+ self.stream.write("FAILED (")
+ failed, errored = map(len, (self.failures, self.errors))
+ if failed:
+ self.stream.write("failures=%d" % failed)
+ if errored:
+ if failed: self.stream.write(", ")
+ self.stream.write("errors=%d" % errored)
+ if self.known_failure_count:
+ if failed or errored: self.stream.write(", ")
+ self.stream.write("known_failure_count=%d" %
+ self.known_failure_count)
+ self.stream.writeln(")")
+ else:
+ if self.known_failure_count:
+ self.stream.writeln("OK (known_failures=%d)" %
+ self.known_failure_count)
+ else:
+ self.stream.writeln("OK")
+ if self.skip_count > 0:
+ skipped = self.skip_count
+ self.stream.writeln('%d test%s skipped' %
+ (skipped, skipped != 1 and "s" or ""))
+ if self.unsupported:
+ for feature, count in sorted(self.unsupported.items()):
+ self.stream.writeln("Missing feature '%s' skipped %d tests." %
+ (feature, count))
         if self._strict:
             ok = self.wasStrictlySuccessful()
         else:
             ok = self.wasSuccessful()
- if ok:
- self.stream.write('tests passed\n')
- else:
- self.stream.write('tests failed\n')
         if TestCase._first_thread_leaker_id:
             self.stream.write(
                 '%s is leaking threads among %d leaking tests.\n' % (
@@ -383,12 +413,12 @@
         else:
             raise errors.BzrError("Unknown whence %r" % whence)

- def finished(self):
- pass
-
     def report_cleaning_up(self):
         pass

+ def startTestRun(self):
+ self.startTime = time.time()
+
     def report_succe...

This far better that what we had.

You give us a far cleaner TestRunner, so whatever is left appears bigger now:
- the loop on result_decorators to propagate a boolean looks highly suspicious,
- the injection of BZRTransformingResult is even more ugly than before, couldn't
  that be a result decorator now ?
- I still don't like that self.verbosity handling here, again, can't that
  be turned into a decorator ?

Given that this patch is a clear improvement, please land, the above can come later
(if you agree on them).