[Piglit] [PATCH 01/16] summary.py: replace multiple attributes with a single dict attribute

[Jose Fonseca]

Dylan,

I tried it with live results but unfortunately it fails:

$ ./piglit-summary-junit.py test-llvmpipe-piglit/results/custom.results/
Traceback (most recent call last):
  File "./piglit-summary-junit.py", line 54, in <module>
    main()
  File "./piglit-summary-junit.py", line 50, in main
    output.generateJUnit(args.output)
  File "/home/jfonseca/work/vmware/tests/piglit/framework/summary.py", line 645, in generateJUnit
    numberoftests=len(self.tests['all'])))
  File "/usr/lib/python2.7/dist-packages/mako/template.py", line 412, in render
    return runtime._render(self, self.callable_, args, data)
  File "/usr/lib/python2.7/dist-packages/mako/runtime.py", line 766, in _render
    **_kwargs_for_callable(callable_, data))
  File "/usr/lib/python2.7/dist-packages/mako/runtime.py", line 798, in _render_context
    _exec_template(inherit, lclcontext, args=args, kwargs=kwargs)
  File "/usr/lib/python2.7/dist-packages/mako/runtime.py", line 824, in _exec_template
    callable_(context, *args, **kwargs)
  File "/home/jfonseca/work/vmware/tests/piglit/.makotmp/templates/junit.mako.py", line 41, in render_body
    __M_writer(unicode(each['system-err']))
UnicodeDecodeError: 'ascii' codec can't decode byte 0xef in position 101: ordinal not in range(128)

This can be avoided doing

diff --git a/templates/junit.mako b/templates/junit.mako
index f123878..1a480dd 100644
--- a/templates/junit.mako
+++ b/templates/junit.mako
@@ -3,8 +3,8 @@
   <testsuite name="piglit" tests="${numberoftests}">
   % for each in tests:
     <testcase classname="${each['classname']}" name="${each['testname']}" time="${each['time']}">
-      <system-out>${each['system-out']}</system-out>
-      <system-err>${each['system-err']}</system-err>
+      <system-out>${each['system-out'].decode('ascii', 'ignore')}</system-out>
+      <system-err>${each['system-err'].decode('ascii', 'ignore')}</system-err>
     </testcase>
   % endfor
   </testsuite>

But still, the output is still very broken -- there are no failures, skips, etc. recorded on the XML.


I spent a considerable amount of time in fine tuning the existing junit.py script to work correctly and robustly, and given that nobody else is actually interested in jenkins integration, I rather not to spend time reimplementing this in a different clothing.

So I'd prefer to keep the non-mako piglit-summary-junit.py implemention.  Feel free to move any old code from framework inside piglit-summary-junit.py though.


If you want to verify, I've uploaded the results I used to test to http://people.freedesktop.org/~jrfonseca/test-llvmpipe-piglit-results.tar.bz2


Jose


----- Original Message -----
> Yes, it's at my github here: git://github.com/crymson/piglit.git
>  mako-complete
> 
> junit isn't too interesting to me or intel (whom I contract with), however
> I converted piglit's html summary to mako a while back, and wrote a new
> summary class for that. I'd like to get all of the summary output using
> this backend, so junit and the text output of piglit-summary.py are targets
> for that.
> 
> 
> On Wed, Jun 26, 2013 at 2:54 PM, Jose Fonseca <jfonseca at vmware.com> wrote:
> 
> > ----- Original Message -----
> > > I'm bumping this series hoping to get some review, since I feel it's too
> > > big to 'just push'
> > >
> > > Brian and Jose, I'd specifically appreciate if you would look at the
> > later
> > > set of patches, since they change/replace code that you originally wrote.
> >
> > Thanks for the heads up Dylan.  Sorry for not noticing earlier.  Nobody
> > else showed interest in junit/jenkins so I wasn't expecting it.
> >
> > I'm using the junit.py here in many other places (to conform results from
> > conform, apitrace, mesa demos, MS conformance test, etc) to jenkins/hudson.
> > So having piglit using something else doesn't really save me much work.
> >
> > But I don't object replacing my script with mako -- piglit has its test
> > framework anyway -- provided there are no regressions.
> >
> > I'd like to test first though.  Can I pull this from somewhere, or could
> > you do push this in two phases: 1) first add the new mechanism without
> > touching the old one, then 2) remove it after I have time to run more tests
> > here.
> >
> > Jose
> >
> >
> > > Thanks.
> > >
> > >
> > > On Thu, Jun 13, 2013 at 8:36 AM, Dylan Baker <baker.dylan.c at gmail.com
> > >wrote:
> > >
> > > > Replaces self.alltests, self.changes, self.problems, self.regressions,
> > > > self.fixes, and self.skipped with self.test, a dictionary containing
> > all
> > > > of the same information. This cleans the namespace for the NewSummary
> > > > object, as well as allowing for some other code simplification
> > > >
> > > > Signed-off-by: Dylan Baker <baker.dylan.c at gmail.com>