[PATCH/RFC] t0000-*.sh: Fix the GIT_SKIP_TESTS sub-tests

Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
---

The test suite has been failing for me on the pu branch for a while now. I finally found a few minutes to take a look.

This failure is specific to the dash shell (/bin/sh) on my system (ie it may well affect other shells, but I haven't tested them all ... :). This does not affect bash (or bash as /bin/sh aka bash --posix).

The GIT_SKIP_TESTS, 13-15, all fail with verbose output similar to:

  $ ./t0000-basic.sh -i -v
  ...
  --- expect	2014-05-20 20:55:54.138342361 +0000
  +++ out	2014-05-20 20:55:54.134342341 +0000
  @@ -1,5 +1,5 @@
   ok 1 - passing test #1
  -ok 2 # skip passing test #2 (GIT_SKIP_TESTS)
  +ok 2 - passing test #2
   ok 3 - passing test #3
   # passed all 3 test(s)
   1..3
  not ok 13 - GIT_SKIP_TESTS
  #	
  #		GIT_SKIP_TESTS='git.2' 		run_sub_test_lib_test git-skip-tests-basic 		'GIT_SKIP_TESTS' <<-\EOF &&
  #		for i in 1 2 3
  #		do
  #			test_expect_success "passing test #$i" 'true'
  #		done
  #		test_done
  #		EOF
  #		check_sub_test_lib_test git-skip-tests-basic <<-\EOF
  #		> ok 1 - passing test #1
  #		> ok 2 # skip passing test #2 (GIT_SKIP_TESTS)
  #		> ok 3 - passing test #3
  #		> # passed all 3 test(s)
  #		> 1..3
  #		EOF
  #	
  $ 

... which looks like the sub-test does not see the GIT_SKIP_TESTS variable at all. Indeed, if I put the entire test in a sub-shell and replace the use of that variable on the run_sub_test_lib_test invocation with a separate explicit assignment and export, then the tests start working. (ie if I do the opposite of some other recent commits!)

This patch is an RFC, because I take a different approach to the above solution, only because the diff is much smaller and easier to read! Is it a better solution?

ATB, Ramsay Jones

 t/t0000-basic.sh | 15 +++++++++------
 1 file changed, 9 insertions(+), 6 deletions(-)