ENH: use utf-8 via manifest on Windows

[codeling]

Elastix was not able to correctly handle special characters in some folder/file name parameter input on Windows; Elastix would interpret those names incorrectly, resulting in errors.

Test cases:

• Failing (1): With parameter -out C:\tmp\elöx\result-transform:

ERROR: the output directory "C:\tmp\el÷x\result-transform\" does not exist.
You are responsible for creating it.

(The folder C:\tmp\elöx\result-transform did exist before the call)

• Failing (2): With parameter -t0 C:\tmp\elx\input\initialTransförmParameter.txt:

itk::ExceptionObject (000000CF80DBF190)
Location: "unknown"
File: C:\Source\elastix\src\Core\ComponentBaseClasses\elxTransformBase.hxx
Line: 202
Description: ITK ERROR: SimilarityTransformElastix(000001D8B0FBAAD8): ERROR: the file C:\tmp\elx\input\initialTransf÷rmParameter.txt does not exist!
Errors occurred!

(the file C:\tmp\elx\input\initialTransförmParameter.txt did exist before the call)

• Working (tested with 796175c) - special characters already correctly interpreted previously:
  • in the moving/fixed volume file name (e.g. -m C:\tmp\elx\input\möving.mha / -f C:\tmp\elx\input\fäxed.mha)
  • The content of the file given via -t0 parameter is already read as UTF-8, and a .h5 file referenced from within that file (TransformFileName) is already correctly loaded when containing special characters

This PR fixes the 2 failing test cases specified above. It adds an application manifest to switch the executable's locale to utf-8 (available since Windows 10 1903, so for every still supported Windows version); this is same as e.g. applied in Slicer . Also the console output is switched to utf-8.

Note: A similar workaround (but setting locale to utf-8 systemwide instead of for the application alone) was already proposed in this issue way back in 2021 for fixing special character issues; the way this PR handles it does not require any action by the user and should just enable special characters working fine on any system >= 1903.

[N-Dekker]

Very interesting, thanks Bernhard! I see you did quite some research to get it right 👍 👍 👍

I think it would be good to have a test for this improvements. Similar to the existing test on spaces in output path names:

elastix/Testing/CMakeLists.txt

Lines 445 to 453 in 796175c

	file(MAKE_DIRECTORY "${TestOutputDir}/OutputDirectoryPathContainingSpacesTest")
	file(MAKE_DIRECTORY "${TestOutputDir}/OutputDirectoryPathContainingSpacesTest/path with spaces")
	file(MAKE_DIRECTORY "${TestOutputDir}/OutputDirectoryPathContainingSpacesTest/path with spaces/name with spaces")
	add_test(NAME OutputDirectoryPathContainingSpacesTest
	COMMAND ${EXECUTABLE_OUTPUT_PATH}/elastix
	-f "${TestDataDir}/2D_2x2_square_object_at_(1,3).mhd"
	-m "${TestDataDir}/2D_2x2_square_object_at_(2,1).mhd"
	-p "${TestDataDir}/parameters.2D.NC.translation.ASGD.txt"
	-out "${TestOutputDir}/OutputDirectoryPathContainingSpacesTest/path with spaces/name with spaces")

Do you still have time to add such a test? Otherwise, I'll see if I can add it myself.

[codeling]

Do you still have time to add such a test? Otherwise, I'll see if I can add it myself.

Done (see added commit).
I have added an additional test case - parameter files with special characters in them also caused failures.

Since I'm not really familiar with the intial transform file format, I just added the first (AI-generated) content that made elastix run successfully (it might not really contain a useful transform).

[N-Dekker]

Thanks @codeling, looks cool!

By the way, GitHub warns me that your commits are "Unverified", saying:

This commit is not signed, but one or more authors requires that any commit attributed to them is signed.
Learn about vigilant mode

Do you have a clue? Otherwise I think I'll just ignore the warning 😃

[codeling]

Do you have a clue? Otherwise I think I'll just ignore the warning

Ah yes, I typically sign my commits but hadn't set it up on that machine yet. I have resubmitted my changes now as single squashed signed commit!

[codeling]

In a ctest run, 3 tests are failing (elastix_run_example_COMPARE_IM, elastix_run_3DCT_lung.MI.bspline.ASGD.001_COMPARE_TP and TransformixTest), but at least on my machine, they also already fail without my fix, so I suppose this isn't something that this PR breaks.

[N-Dekker]

Thank you very much, Bernhard! Really cool!