Open seanm opened 3 weeks ago
Some test interdependency probably needs to be defined. Indeed, ITK_USE_BRAINWEB_DATA
does not get a lot of testing. We could change that by turning it on in one of the nightly build configurations?
It's enabled on some of my nightly builds, like here: https://open.cdash.org/viewTest.php?onlyfailed&buildid=10165388
I was trying in yesterday's hackathon to green them up. :)
So since this has been set to 'backlog', I guess no one is going to work on it any time soon?
If so, I'll get rid of ITK_USE_BRAINWEB_DATA
on my nightly cdash submissions... (which I'd really like to get green)
I expect that no one will work on it any time soon. You are quite welcome to invalidate that expectation by working on it 😄
I expect that no one will work on it any time soon. You are quite welcome to invalidate that expectation by working on it 😄
Alas, I can't spend all my time on ITK. :)
I'll remove ITK_USE_BRAINWEB_DATA
from my builds then.
I suspect ITK_USE_BRAINWEB_DATA does not get a lot of testing...
If I enable that option, build, then do
ctest -V -R DeformableRegistration8Test3
, I get a failure like:Indeed that file is not there. If I then do
ctest -j22 -R .
, that seems to tickle something else to get/create the file. Now if I run justDeformableRegistration8Test3
it finds that file, and I get farther.It runs for minutes and consumes tonnes of RAM. I have 192 GiB of physical RAM and it uses it all and starts swapping.
If I then enable UBSan I find a use of a null pointer. I chased it a bit and if you take this
virtualImage
:https://github.com/InsightSoftwareConsortium/ITK/blob/1090d31e0360022b4414d0c3cbf27927a7d55901/Modules/Registration/Metricsv4/include/itkImageToImageMetricv4GetValueAndDerivativeThreader.hxx#L32
and add the line after
assert(virtualImage->GetBufferPointer());
that's the null, which I believe is unexpected.