[PATCH v2 0/7] increase wildmatch test coverage

This v2 addresses comments by Johannes Sixt in <8f4cdb23-8e2e-144a-1f70-99776b027166@kdbg.org> and there's osme other cleanups as noted below.

Ævar Arnfjörð Bjarmason (7):
  wildmatch test: indent with tabs, not spaces
  wildmatch test: use more standard shell style

No changes.

  wildmatch test: don't try to vertically align our output

NEW: Don't try to do whitespace alignment in the tests.

  wildmatch test: use a paranoia pattern from nul_match()

Explain in the commit message why we're not using say '...'; exit 1. I said I'd use this in <874logs7vi.fsf@evledraar.gmail.com>, but on further consideration it's a bad idea.

  wildmatch test: remove dead fnmatch() test code
  wildmatch test: perform all tests under all wildmatch() modes

Just changes to rebase them on the changes above.

  wildmatch test: create & test files on disk in addition to in-memory

Avoid some forking by using a case statement instead of if .. grep && return.

Add comments to the code to clarify what it's doing.

Factoro out the repetitive part of the tests into functions, making the patch shorter.

I didn't change the "rm -rf -- *" pattern Johannes was concerned about, because after looking at it it would be a pain to create some-test-subdir/ and only if some-test-subdir/$our_file gets created cd to it and then remove it afterwards, it's much easier not to change the directory. The test test_when_finished always runs in the directory the tests executed in, so I don't see how this is dangerous in practice.

I didn't move the "printf" pattern to here-docs as discussed in the thread.

 t/helper/test-wildmatch.c |   2 +
 t/t3070-wildmatch.sh      | 759 +++++++++++++++++++++++++++++++---------------
 2 files changed, 516 insertions(+), 245 deletions(-)