BUG: Fixes crashes in unit tests when argv[0] == nullptr

[N-Dekker]

Many unit tests could crash (or have undefined behavior), when doing something
like the following with argv[0] == nullptr:

std::cerr << "Usage: " << argv[0] << std::endl;

This commit prevents such crashes by using a new macro,
itkNameOfTestExecutableMacro(argv), instead of argv[0].

[N-Dekker]

This pull request fixes crashes (undefined behavior) from many ITK test drivers, when the user manually selects a test from the command-line interface, for example:

Run ITKBiasCorrectionTestDriver. The following output appears:

Available tests:
  0. itkCompositeValleyFunctionTest
  1. itkMRIBiasFieldCorrectionFilterTest
  2. itkN4BiasFieldCorrectionImageFilterTest
To run a test, enter the test number:

Then type 2 and press Enter

Without this fix, it would crash, trying to print argv[0] at

ITK/Modules/Filtering/BiasCorrection/test/itkN4BiasFieldCorrectionImageFilterTest.cxx

Line 332 in f963831

std::cerr << "Usage: " << argv[0] << " imageDimension inputImage "

When this fix is applied, it won't crash anymore. Instead, it will say:

Usage: <itkN4BiasFieldCorrectionImageFilterTest executable> imageDimension inputImage outputLogControlPointLattice [shrinkFactor,default=1] [numberOfIterations,default=100x50x50]  [maskImageWithLabelEqualTo1] [splineDistance,default=200]

What do you think?

[dzenanz]

I was bothered by that behavior, but never enough to spend the time to fix it. Thanks Niels!

[hjmjohnson]

+1

[jhlegarreta]

@N-Dekker I have just bumped into this PR. Very nice 👍 !

I've got two comments whose scope is a little bit broader but related to this topic:

• About the test driver message:

Run ITKBiasCorrectionTestDriver. The following output appears:
Available tests:

0. itkCompositeValleyFunctionTest
1. itkMRIBiasFieldCorrectionFilterTest
2. itkN4BiasFieldCorrectionImageFilterTest

To run a test, enter the test number:
Then type 2 and press Enter
Without this fix, it would crash, trying to print argv[0] at

Then the test driver message is also/still misleading right? Even if I usually call the tests by their name, I remember having experienced this, and dot not specifically remember the crash conditions, but I would dare to say that hitting the right number does execute the test (?). If this was not the case, and even if the usage is now printed and this prevents the crash, then test instruction does still not honor the behavior, right? Sorry if I'm missing some piece of information.

• Has this change been incorporated into the ITK SW Guide? This is a broader concern, but also applies to this topic: I feel that there have been quite a few enhancements concerning class declarations (C++11, performance improvements, etc.), in the last few months that have not been incorporated into the SW Guide. An effort has been made to update the Release Notes, but I my feeling is that some have not made their way into the SW Guide. After a few weeks, it is all the way harder to remember them, so as to add them to the SW Guide.

Thanks.

[jhlegarreta]

I realize that this missed all av[0] cases. In some cases, where I was editing a given test file to increase the code coverage, I changed them.

A PR to change the remaining ones: #2939