[libcxx-commits] [libcxx] [libcxx][lit] Fixing libcxx test failures on Windows (PR #194752)

[via libcxx-commits]

zeroomega wrote:

> @zeroomega , @boomanaiden154 ,
> 
> these changes broke these windows builders
> 
> * https://lab.llvm.org/buildbot/#/builders/38/builds/8698
> * https://lab.llvm.org/buildbot/#/builders/193/builds/15937
> 
> see details here: https://lab.llvm.org/buildbot/#/builders/38/builds/8698/steps/16/logs/FAIL__llvm-libc__-static_cfg_in__dsl_sh_py
> 
> ```
> # | ======================================================================
> # | FAIL: test_caching_is_not_too_aggressive (__main__.TestProgramOutput.test_caching_is_not_too_aggressive)
> # | ----------------------------------------------------------------------
> # | Traceback (most recent call last):
> # |   File "C:\buildbot\as-builder-1\x-armv7l\llvm-project\libcxx\test\selftest\dsl\dsl.sh.py", line 229, in test_caching_is_not_too_aggressive
> # |     self.assertEqual(output1, "MACRO=1%s" % os.linesep)
> # |     ~~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
> # | AssertionError: 'MACRO=1\n' != 'MACRO=1\r\n'
> # | - MACRO=1
> # | + MACRO=1
> # | 
> # | ?        +
> ```

I see what was going on, the dsl.sh.py uses the host line breaker while the test was running linux runtimes test on Windows..
How was it even possible to run Linux runtime test on Windows? Was it through WSL interface?

Reverting this PR will break all Windows bots in libcxx side. What we can do is probably temporarily disable the test while we find a better way to address this test on different line breakers.

https://github.com/llvm/llvm-project/pull/194752