To do: Review library code and remove any uninitialized memory access

[cookpa]

Relating to #579. Further fixes may be pending on the ANTs side (ANTsX/ANTs#1701)

[cookpa]

Additional ANTs fixes ready for update here ANTsX/ANTs#1702

[cookpa]

Going through the C++ code replacing Allocate with AllocateInitialized - just wanted to note this

https://github.com/ANTsX/ANTsPy/blame/1e363730d033500a91264e4ab6509f093dd609ae/ants/lib/LOCAL_antsImageToImageMetric.h#L357-L363

a prior call to FillBuffer is commented out for some reason. Hopefully just a performance thing

[ntustison]

Im guessing I just missed that one.

[cookpa]

@ntustison this is in the ANTsPy code, I think you got everything in ANTsX/ANTs

I'm trying to figure out if images with Vector pixel type require special handling. Elsewhere we have

    VectorType zeroVector( 0.0 );
    ITKFieldPointerType rasterizedField = ITKFieldType::New();
    rasterizedField->SetOrigin( bsplineFilter->GetBSplineDomainOrigin() );
    rasterizedField->SetSpacing( bsplineFilter->GetBSplineDomainSpacing() );
    rasterizedField->SetDirection( bsplineFilter->GetBSplineDomainDirection() );
    rasterizedField->SetRegions( bsplineFilter->GetBSplineDomainSize() );
    rasterizedField->Allocate();
    rasterizedField->FillBuffer( zeroVector );

I am hoping we can just call rasterizedField->AllocateInitialized. I think this is the case because itkImage must be aware of the total buffer size it needs to initialize

[ntustison]

Okay, was on my phone. There's quite a few code blocks like this where we're switching back and forth from an image of vectors to a vector image. I would assume there's nothing complicated going on and whatever changes you want to make should be fine.