expected failures should not be seen as failures

[Vincent Ladeuil]

I need to use the following patch or using subunit2junitxml just doesn't work with hudson: the expected failures are seen as true failures which is not good (tm).

=== modified file 'junitxml/__init__.py'
--- junitxml/__init__.py 2010-01-12 10:26:29 +0000
+++ junitxml/__init__.py 2010-01-12 10:30:45 +0000
@@ -163,5 +163,5 @@
             pass
         self._test_case_string(test)
         self._results.append('>\n')
- self._results.append('<failure type="%s.%s">%s</failure>\n</testcase>\n'% (escape(error[0].__module__), escape(error[0].__name__), escape(self._exc_info_to_string(error, test))))
-
+ self._results.append('<xfail>%s</xfail>\n</testcase>\n'%
+ escape(self._exc_info_to_string(error, test)))

[Robert Collins (lifeless)]

Need to find out what junit's native XML contains for an expected failure.

[Robert Collins (lifeless)]

Martins fix makes this a success, which is ok even though a little imprecise.

[Martin Packman (gz)]

Right, I couldn't find much info about how accepting of format changes junit things are, so it seemed falling back to success was a safer first step than inventing a new tag. We could easily extend the schema to include more result types like xfail later if that turns out to be acceptable.