[vtk-developers] Looking into dashboard failures this morning

[Marcus D. Hanwell]

On Fri, Apr 5, 2013 at 11:02 AM, David Cole <dlrdave at aol.com> wrote:
> On Fri, Apr 5, 2013 at 10:37 AM, David Gobbi <david.gobbi at gmail.com> wrote:
>> On Fri, Apr 5, 2013 at 8:11 AM, Marcus D. Hanwell
>> <marcus.hanwell at kitware.com> wrote:
>>> The RogueResearch submissions all look like vtkgl headers are not in
>>> the include path, but looking at our changes I am not sure what would
>>> cause that.
>>
>> There was probably a previously undetected problem with the dependency
>> chain for vtkgl.h that was never noticed until yesterday's change caused
>> a change to the order in which certain things are built.
>>
> When all wrapping was turned off this still worked though, and so it
> seems like any changes to wrapping/hierarchy code would not reveal
> unknown implicit dependencies as pure C++ builds never even invoke the
> wrapping logic. Sean - are there local changes on those machines you
> made recently? I don't see this failure mode on anything but Rogue
> Research systems.
>
> All those machines are using CMake 2.8.11-rc1 or cmake ‘master’...
>
> Is that a clue?
>
Possibly, I see hythloth fails in the same way and also used 2.8.11
rc1. The londinium machine used CMake master from a few days ago and
passed, hythloth is Brad's machine so he may be able to look into that
(thanks for pointing that out).

Marcus