BUG: fixed memory leaks in MINC IO

[issakomi]

MINC IO does not free allocations before throwing an exception sometimes,
s. discussion #3529

https://open.cdash.org/viewDynamicAnalysis.php?buildid=8097352

(reviewed writing an image).

[jhlegarreta]

Thanks for doing this @issakomi. Have not tested but looks good.

Maybe, following for comment in the related issue:

I have done trivial test, without using itkIOTestHelper.h, just read MHA and write MNC to wrong path, the leaks are there. If the path is correct -- there are no problems,

Adding a test such as the one you did would be convenient.

[gdevenyi]

@vfonov

[issakomi]

I have done trivial test, without using itkIOTestHelper.h, just read MHA and write MNC to wrong path, the leaks are there. If the path is correct -- there are no problems,

Adding a test such as the one you did would be convenient.

it is really very simple test, to try valgrind --leak-check=full ./test without ITK's testing subsystem, already merged new ITK tests with intentional wrong path (which discovered leaks) are doing the same. Maybe later more tests could be done to trigger exceptions and increase coverage, e.g. 4D image, unsupported data type, etc. I have tested all exit points manually, they work.

[vfonov]

@vfonov

Looks ok