[vtk-developers] The dashboard: tarvalon gerrit builds

[David Cole]

Gerrit builds with 0 build errors from tarvalon since Feb 20th:

http://open.cdash.org/index.php?project=VTK&filtercount=4&showfilters=1&filtercombine=and&field1=site/string&compare1=63&value1=tarvalon&field2=buildstarttime/date&compare2=83&value2=2013-02-20&field3=builderrors/number&compare3=41&value3=0&field4=groupname/string&compare4=61&value4=Gerrit

Sort the above results by build time, to look at things in date order.

Observe:
- Recurring persistent test failures were eliminated and reduced to 0 
at the VTK dash-a-thon on Feb. 27th (yay!)
- From Feb. 27th to Mar. 7th, there were predominantly 0 test failures 
on these builds. When it was non-0, it typically indicated a real 
problem with the topic being tested. Or with a VTKData change. Quite 
helpful.
- From Mar. 8th to Mar. 21st, there were no submissions from this 
machine. It was apparently down for that long.
- Beginning Mar. 22nd and continuing to the most recent submission, 
there are 10 recurring test failures again.

Conclusion:
- sometime between Mar. 8th and 22nd, changes were merged that caused 
these test failures (or perhaps something changed on the machine while 
it was on vacation?)
- tarvalon is now only good as a gerrit submitter for detecting newly 
introduced Windows compile/link errors or warnings, which is still 
somewhat useful. But... since tests persistently and recurringly fail, 
it's *not* very useful at detecting runtime behavior changes introduced 
by gerrit topics.

Questions:
- Can we, the VTK developer community, fix it again and keep it fixed? 
Or do we need a dash-a-thon every month? ;-)
- Can anybody else volunteer an additional Windows gerrit build?


I, for one, would appreciate a clean gerrit Windows build. I *did* 
appreciate it for the 9 brief days that we had it...


Thanks for considering,
David C.