Unit test failures

[Boudewijn Rempt]

On Sunday 14 October 2012 Oct, Sven Langkamp wrote:
> Hi,
> 
> 
> we currently have quite a few unit test failures. I have started going
> through them and already fixed some bugs. I think we reallly need to be
> more carefully with the test. There were some cases were the unit test
> broke and some serious problems slipped in unnoticed. We should try to get
> it to zero or at least turn off some test, so that we don't miss new
> problems.
> 

Yes, definitely. We really got sloppy there -- all of us!

> There are currently 15 Krita related unit tests are failing in Jenkins:
> 
> krita-ui-FreehandStrokeTest (Likely due to a slightly changed algorithm
> when using Vc, might need to update result images)
> krita-image-KisFilterSelectionsBenchmark (Timeout, works locally, Any idea
> how benchmarks are blacklisted?)

No -- but if it's a real benchmark, it should be in the benchmarks dir, those don't get executed with make test.

> krita-image-KisLowMemoryTests (also timeout)
> krita-image-KisHistogramTest (crash after it can't create histogram
> producer for YCBCR (no profiles available), disable YCBCR?)

Should be fixed in the test: this is a perfectly okay condition.

> krita-ui-KisZoomAndPanTest (fails on server, locally it seems to run in an
> infinite loop)

Runs here as well.

> krita-image-KisConvolutionPainterTest
> krita-image-KisProcessingsTest
> krita-ui-kis_node_model_test (I fixed that, so it runs locally. No idea why
> it still fails in Jenkins)
> krita-ui-kis_shape_controller_test
> krita-ui-KisSelectionManagerTest (fails as there is no selection node after
> deselect. Did we change that behavior?)

Dmitry?

> krita-plugins-formats-tiff_test
> krita-plugins-formats-png_test
> krita-fileformat-kis_exr_test

These are tricky, since they depend on the version of the tiff, png and exr libraries. I wonder whether this code is amenable to unittests after all?

> libs-widgets-KoResourceTagging_test (works locally, I think nepomuk might
> not working on the Jenkins server)

That one never worked for me!

> libs-pigment-TestKoRgbU8ColorSpaceTester (F32 doesn't compute exactly the
> expected values, might need fuzzy compare)

Yeah, and the values probably changed when moving to lcms 2.4.

> 
> Sometimes also the clone layer test fails (like every four or five time).
> 
> 
> Sven
> 


-- 
Boudewijn Rempt
http://www.valdyas.org, http://www.krita.org, http://www.boudewijnrempt.nl