failing tests on 32 vs 64 bits

[Cyrille Berger]

I ran the tests and got 4 failures out of 201 (that includes libs+krita):
KisConvolutionPainterTest
KisFilterTest
KisAsyncMerger
KisAutobrushTest

That is on a debian 64bits on intel hardware. Not sure if it makes a 
difference.

On Tuesday 13 April 2010, Boudewijn Rempt wrote:
> On Tuesday 13 April 2010, Cyrille Berger wrote:
> > I could not reproduce in a 64bits VM. I will try later to run more Krita
> > tests in that VM.
> 
> Could you also try to reproduce the invalid read by one issue I see when
> using memcheck on a 64 bit system?
I just have to run krita and paint ?
I will try that tomorrow.

> > And then extra nice would be if you could run tests
> > of OpenGTL (you would need to set OPENGTL_BUILD_TESTS to true in
> > opengtl's build dir, and then make test, I expect a lot of failures due
> > to the lack of the dataset, but I should be able to see if there are
> > unexpected failures).
> 
> The following tests FAILED:
> (...)
This is not good, most of those tests should have passed. But at least it 
proves that I could try to reproduce the problem with only OpenGTL test suite. 
I know that at some point you had a more recent opengtl, can you try to make 
sure that your installed library does not contains any of the opengtl library 
of so version 0.8 ? (aka libGTLCore.so.0.8@  libGTLGaishu.so.0.8@  
libGTLImageIO.so.0.8@  libOpenCTL.so.0.8@  libOpenRijn.so.0.8@  
libOpenShiva.so.0.8@ and libGTLCore.so.0.9.15*  libGTLGaishu.so.0.9.15*  
libGTLImageIO.so.0.9.15*  libOpenCTL.so.0.9.15*  libOpenRijn.so.0.9.15*  
libOpenShiva.so.0.9.15* and possibely also the 0.9.14 ones) maybe there is 
some mix up.

-- 
Cyrille Berger