[vtk-developers] Ninja-based CDash at home client?

[David Gobbi]

On Wed, Mar 27, 2013 at 10:45 AM, Marcus D. Hanwell
<marcus.hanwell at kitware.com> wrote:
> On Wed, Mar 27, 2013 at 11:16 AM, David Cole <dlrdave at aol.com> wrote:
>> Well it seemed to work for the past week until 2 days ago. If you look at
>> this build history in this filtered view, and sort by "Build Time":
>>
>>
>> http://open.cdash.org/index.php?project=VTK&filtercount=3&showfilters=1&filtercombine=and&field1=buildname/string&compare1=61&value1=Win32-ninja-Release&field2=buildstarttime/date&compare2=83&value2=2013-03-17&field3=buildstarttime/date&compare3=84&value3=now
>>
>> ...you can see that I turned python on the day it started having "222"
>> warnings and 462 more tests started running. It built just fine for 5 days
>> in a row, and then, for the last 2 days, has yielded this new error.
>>
> I would say you were probably lucky, like I said we tried it several
> times (on Linux - maybe the state was better on other platforms) and
> it was very unreliable. Chris mentioned that Utkarsh was able to build
> ParaView with wrapping though, so maybe it was precarious? It was
> certainly known to be unreliable, and I saw the exact error you are
> seeing when it failed. It looks like David Gobbi has an initial patch
> at least, hopefully that will work for you.

The rules for the hierarchy files are necessarily generator-specific.
They involve tricks to hide certain dependencies in order to allow
people to modify header files without triggering massive re-builds.
These "dependency tricks" are very robust for the VisualStudio
generator, are quite robust on the make/nmake generators, but it
looks like they'll have to be disabled for all other generators.

I had a plan to eliminate the Hierarchy files entirely (by adding certain
capabilities to vtkWrapPython.exe etc.) but it's been very hard to find
time to do the coding...

 - David