-
Bug
-
Resolution: Fixed
-
Major
-
None
-
Platform: All, OS: All
hi! Hudson is great, thanks for writing it 
We're using it heavily
in the ASF these days: http://hudson.zones.apache.org/hudson/
Quick bug report. I'm writing a TAP-to-JUnit-XML conversion script for our
usage of Hudson in Apache SpamAssassin (which is written in perl, and therefore
uses TAP, the Test Anything Protocol, for test output interchange). the script
is here, if anyone's interested –
http://jmason.org/software/scripts/tap-to-junit-xml.txt
Anyway, there appears to be a small bug in how Hudson interprets the
single-file format (multiple <testsuite> blocks contained by one <testsuites>
container), compared to the multiple-file format (each file contains one
<testsuite> block, no <testsuites> container). With single-file format, the
<system-out> block is never displayed, whereas it is displayed fine with
multiple-file format.
I suspect the "it.parent" pointer in the jelly file is pointing at the
top-level parent node – the <testsuites> node – rather than the immediate
parent node, the <testsuite> node that the current <testcase> is contained by.
(that's just a guess, though.)
Here's single-file output:
And here's what multi-file output looks like on the same input:
if you need more info, let me know and I'll try to dig up a demo XML file.
[JENKINS-1472] single-file JUnit XML format hides <system-out>
redsolo
added a comment - Created an attachment (id=198)
File containing stdout and one test suite 
jmason
added a comment - it's a pity there's no well-defined standard or schema; even some kind of "best
practices" would be welcome.
Here's a sample of the single-file format. here's how it is displayed on our
Hudson:
here's an example of the missing stdout:
jmason
added a comment - it's a pity there's no well-defined standard or schema; even some kind of "best
practices" would be welcome.
Here's a sample of the single-file format. here's how it is displayed on our
Hudson:
http://hudson.zones.apache.org/hudson/view/SpamAssassin/job/SpamAssassin-tap-parser-test/39/testReport/make_test/
here's an example of the missing stdout:
http://hudson.zones.apache.org/hudson/view/SpamAssassin/job/SpamAssassin-tap-parser-test/39/testReport/make_test/t_basic_lint_t/test______1/ jmason
added a comment - Created an attachment (id=200)
singlefile.xml Code changed in hudson
User: kohsuke
Path:
/hudson/hudson/main/core/src/main/java/hudson/tasks/junit/SuiteResult.java (1.14)
/hudson/hudson/main/core/src/main/java/hudson/tasks/junit/TestResult.java (1.18)
/hudson/hudson/main/core/src/main/java/hudson/tasks/junit/CaseResult.java (1.15)
/hudson/hudson/main/core/src/test/java/hudson/tasks/junit/SuiteResultTest.java (1.2)
/hudson/hudson/main/core/src/test/resources/hudson/tasks/junit/junit-report-1472.xml (1.1)
http://fisheye5.cenqua.com/changelog/hudson/?cs=MAIN:kohsuke:20080403080351
Log:
fixed issue #1472.
For us to keep sanity in parsing JUnit XML, I think we have to break the earlier fix we made for JBoss guys. See http://www.nabble.com/difference-in-junit-publisher-and-ant-junitreport-td12265700.html#a12265700 for more about that.
The problem is that the fix for issue #1233 is at odds with this XML that JBoss uses, when it comes to whether <testsuite>/@name should win or <testcase>/@classname wins.
JUnit XML from JBoss was said to be written by a custom formatter, so I think we can be justified in breaking that in favor of #1233 (although junit XSLT apparently handles that correctly.)
jazzjack
added a comment - i've found a schema.
it doesn't look like official document
but may be useful anyway
http://jra1mw.cvs.cern.ch:8180/cgi-bin/jra1mw.cgi/org.glite.testing.unit/config/
JUnitXSchema.xsd?view=markup&content-type=text%2Fvnd.viewcvs-
markup&revision=HEAD
jazzjack
added a comment - i've found a schema.
it doesn't look like official document
but may be useful anyway
http://jra1mw.cvs.cern.ch:8180/cgi-bin/jra1mw.cgi/org.glite.testing.unit/config/
JUnitXSchema.xsd?view=markup&content-type=text%2Fvnd.viewcvs-
markup&revision=HEAD 
Yes, there is a difference on how the content in the XML files are parsed. There
is no XML schema for JUnit XML reports (or standard), so the logic for handling
multiple test suites in a file is a little hackish.
Could you please attach an example of the single-file XML report? I looked in
your workspace for the job, but could not find a single junit xml file.