[vtk-developers] Request: turn off BTX/ETX in some dashboad builds

[David Gobbi]

Getting back to the BTX/ETX issue, it would like to request permission
to turn off BTX/ETX detection in the wrapper code for at least one
night so that I can see how all the dashboard machines react.

Last week me and Marcus ran experimental dashboards on Linux, Win32,
and OS X and only two tests failed as a result of the change:
TestTkRenderWidgetPython (which I've fixed) and TestSetGet (which
fails because removing BTX/ETX causes many addition Set/Get methods to
be wrapped).  ParaView also continued to compile and test with no
errors, which was expected because the "IGNORE_BTX" option is
currently limited to VTK's own classes and does not apply to external
VTK classes.

So what do people think?  Should I wait, or can I pull the switch this
afternoon?

  David


On Fri, Jul 9, 2010 at 3:27 PM, David Gobbi <david.gobbi at gmail.com> wrote:
> Okay, my VTK_IGNORE_BTX Win32 test finished (just nmake, not devenv).
> Exactly the same expected failures: TestSetGet and
> TestTkRenderWidgetPython
>
>  http://www.cdash.org/CDash/viewTest.php?onlyfailed&buildid=660432
>
> The TestTkRenderWidgetPython test failure really does seem to be due
> to the fact that this is an old crufty test, since the
> vtkTkRenderWidget actually does work in Python.
>
>  David
>
>
> On Fri, Jul 9, 2010 at 12:01 PM, David Gobbi <david.gobbi at gmail.com> wrote:
>> I just built ParaView on linux64 with VTK_IGNORE_BTX and it looks good:
>>
>> The following tests FAILED:
>>         27 - PrintSelf-PVServer-Common (Failed)
>>         28 - PrintSelf-PVServer-Filters (Failed)
>>         29 - PrintSelf-PVServer-ServerManager (Failed)
>>         97 - pvcs.PropertyLink (Failed)
>>        137 - pvcrs.PropertyLink (Failed)
>>
>> The PrintSelf tests always fail on my machine, and the PropertyLink
>> tests are also failing on the continuous dashboard, so no worries
>> about them either.
>>
>> Here are the results for an OSX 10.6 build of just VTK by itself:
>>
>> The following tests FAILED:
>>         65 - TestSetGet (SEGFAULT)
>>        751 - TestGaussianBlurPass (Failed)
>>        768 - TestOpacity (Failed)
>>        783 - TestTranslucentLUTDepthPeeling (Failed)
>>        784 - TestTranslucentLUTDepthPeelingPass (Failed)
>>        786 - TestTranslucentLUTTextureDepthPeeling (Failed)
>>        843 - TestTkRenderWidgetPython-image (Failed)
>>
>> All those GPU tests always fail on my system, so only the TestSetGet
>> and TkRenderWidgetPython tests concern me, and both are expected.
>>
>> So I'd have to do nmake and devenv builds to be sure, but it looks
>> like VTK_IGNORE_BTX is a pretty safe option now.
>>
>>   David
>>
>>
>>
>> On Fri, Jul 9, 2010 at 10:10 AM, David Gobbi <david.gobbi at gmail.com> wrote:
>>> Thanks.  I've seen TestTkRenderWidget failures, too, and I suspect
>>> that the test itself is broken.
>>>
>>> I've just pushed my safer BTX/ETX ignore code.  This should allow
>>> VTK_IGNORE_BTX to be used even with ParaView and Titan.
>>>
>>>   David
>>>
>>>
>>> On Fri, Jul 9, 2010 at 9:45 AM, Marcus D. Hanwell
>>> <marcus.hanwell at kitware.com> wrote:
>>>> On Fri, Jul 9, 2010 at 11:38 AM, David Cole <david.cole at kitware.com> wrote:
>>>>>
>>>>> On Fri, Jul 9, 2010 at 11:16 AM, David Gobbi <david.gobbi at gmail.com>
>>>>> wrote:
>>>>>>
>>>>>> The one dashboard test that is guaranteed to break is TestSetGet.
>>>>>> Some of the BTX'd methods are Get methods that segfault when they are
>>>>>> called, and if BTX is ignored, then TestSetGet will call them. These
>>>>>> methods will either have to be fixed or excluded from TestSetGet.
>>>>>
>>>>> The main point of that test is to ensure that any Get method may be called
>>>>> directly after instantiating any VTK object. I would argue for fixing, not
>>>>> excluding. VTK should not crash if somebody "accidentally" calls a Get
>>>>> method before the object is ready for it. It should just return 0, or false,
>>>>> or uninitialized...
>>>>> It might be ok to add error or warning macro output, but crashing is not
>>>>> cool.
>>>>
>>>> So, here is my experimental dashboard. I can add this to my crontab until it
>>>> is more globally enabled. You can see three tests failing there.
>>>>
>>>> http://www.cdash.org/CDash/viewTest.php?onlyfailed&buildid=660272
>>>> Marcus
>>>> --
>>>> Marcus D. Hanwell, Ph.D.
>>>> R&D Engineer, Kitware Inc.
>>>> (518) 881-4937
>>>>
>>>
>>
>