[vtk-developers] valgrind still missing

[David E DeMarle]

I think the VTK valgrind submission should be settling out in the next
couple of days.

>From a preliminary look at the errors I think that:
TestGenericDataSetTessellator and the GeoVis and InfoVis tests are the
worst offenders.

My guess is that these particular tests had not been run through
valgrind until after VTK modularization. With that change these kits
became less optional, and once we started activating the tests that
were inactive before or de-activated during the big modularization
push they were exposed. It is hard to prove that though because CDash
doesn't go back far enough to show the pre-modularization valgrind
submitter's cmake configuration.

David E DeMarle
Kitware, Inc.
R&D Engineer
21 Corporate Drive
Clifton Park, NY 12065-8662
Phone: 518-881-4909


On Sat, Sep 8, 2012 at 2:16 PM, Bill Lorensen <bill.lorensen at gmail.com> wrote:
> Unfortunately, it always takes a while to get rid of those bogus errors.
>
>
> On Sat, Sep 8, 2012 at 10:34 AM, David E DeMarle <dave.demarle at kitware.com>
> wrote:
>>
>> Darn, I thought we had zero defects! :)
>>
>> I fixed that right away, but now I have the same problem on amber8 as
>> I did on karego-at. Valgrind is catching leaks in the system libraries
>> on this machine that don't matter from our perspective. As a result
>> the valgrind result file is several GB in size, which cdash wouldn't
>> be able to digest. I'll take another shot at writing a valgrind
>> suppression file which cuts out the junk and leaves the actual
>> problems later this coming week.
>>
>> David E DeMarle
>> Kitware, Inc.
>> R&D Engineer
>> 21 Corporate Drive
>> Clifton Park, NY 12065-8662
>> Phone: 518-881-4909
>>
>>
>> On Sat, Sep 8, 2012 at 10:26 AM, Bill Lorensen <bill.lorensen at gmail.com>
>> wrote:
>> > David,
>> >
>> > Looks like you're getting close.
>> >
>> > However, each test is failing with:
>> >
>> > valgrind: Bad option: --num-callers=100
>> > valgrind: --num-callers argument must be between 1 and 50
>> > valgrind: Bad option: --num-callers=100
>> >
>> >
>> >
>> > On Thu, Sep 6, 2012 at 10:26 PM, David E DeMarle
>> > <dave.demarle at kitware.com>
>> > wrote:
>> >>
>> >> Note though that the bigger priority right now is restoring the
>> >> coverage
>> >> we had before and resolving enough of the comp failures, warnings and
>> >> test
>> >> failure noise to make the dashboards useable again. No sense running
>> >> valgrind really when tests are missing and failing and VTK_DEBUG_LEAKS
>> >> is
>> >> complaining.
>> >>
>> >> On Thursday, September 6, 2012, David E DeMarle wrote:
>> >>>
>> >>> The valgrind supressions need work to separate the wheat from the
>> >>> chaff.
>> >>> Especially since it is a Mac, not Linux this time. I may give up and
>> >>> find a
>> >>> Linux box instead, but we will have a valgrind submission soon and
>> >>> hopefully
>> >>> we can all chip in and plug the leaks like we did last time around.
>> >>>
>> >>> On Thursday, September 6, 2012, Bill Lorensen wrote:
>> >>>>
>> >>>> I did see that,  but  no valgrind summary on cdash.
>> >>>>
>> >>>> On Thu, Sep 6, 2012 at 8:27 PM, David E DeMarle
>> >>>> <dave.demarle at kitware.com> wrote:
>> >>>>>
>> >>>>> I'm working on it. See karego-at on the experimental section.
>> >>>>>
>> >>>>>
>> >>>>> On Thursday, September 6, 2012, Bill Lorensen wrote:
>> >>>>>>
>> >>>>>> Folks,
>> >>>>>>
>> >>>>>> I notice a flurry of activity to clean up the dashboard. I assume
>> >>>>>> this
>> >>>>>> is motivated by the desire to release VTK6.
>> >>>>>>
>> >>>>>> A valgrind build has been missing for some time now. I assume we
>> >>>>>> will
>> >>>>>> not release VTK6 until we have a valgrind build that has zero
>> >>>>>> defects.
>> >>>>>>
>> >>>>>> I spent a bit of time last Fall getting the defects to zero:
>> >>>>>> http://vtk.org/Wiki/VTK/SoftwareQuality/Valgrind/Fall2011
>> >>>>>>
>> >>>>>> The last valgrind report was August 14 with over 2400 defects.
>> >>>>>>
>> >>>>>> Bill
>> >>>>>>
>> >>>>>
>> >>>>>
>> >>>>> --
>> >>>>> David E DeMarle
>> >>>>> Kitware, Inc.
>> >>>>> R&D Engineer
>> >>>>> 21 Corporate Drive
>> >>>>> Clifton Park, NY 12065-8662
>> >>>>> Phone: 518-881-4909
>> >>>>
>> >>>>
>> >>>>
>> >>>>
>> >>>> --
>> >>>> Unpaid intern in BillsBasement at noware dot com
>> >>>>
>> >>>
>> >>>
>> >>> --
>> >>> David E DeMarle
>> >>> Kitware, Inc.
>> >>> R&D Engineer
>> >>> 21 Corporate Drive
>> >>> Clifton Park, NY 12065-8662
>> >>> Phone: 518-881-4909
>> >>
>> >>
>> >>
>> >> --
>> >> David E DeMarle
>> >> Kitware, Inc.
>> >> R&D Engineer
>> >> 21 Corporate Drive
>> >> Clifton Park, NY 12065-8662
>> >> Phone: 518-881-4909
>> >
>> >
>> >
>> >
>> > --
>> > Unpaid intern in BillsBasement at noware dot com
>> >
>
>
>
>
> --
> Unpaid intern in BillsBasement at noware dot com
>