Update core tests: fixes for CRLF in multiline tests, executable-not-found issue

[cxw42]

See editorconfig/editorconfig-core-test#28

CC @xuhdev

[cxw42]

@xuhdev Have you seen this failure mode before? --- https://travis-ci.org/editorconfig/editorconfig-core-c/jobs/600209206#L299 . pcre2 static-link failure on Xenial with GCC. I checked, and libpcre2-dev on xenial does have a static library.

Other than that, it seems like my latest test change may have helped - most of the jobs in that build passed.

[xuhdev]

Awesome! This is really good work! 🎆

No I haven't seen that error before. I'm enabling verbose output in #64 so we can see what's going on.

[xuhdev]

But path_separator_backslash_in_cmd_line fails now on Windows, which did not fail before...

I'll try to fix the static build issue (which shouldn't be relevant here). Something may have changed on Travis...

[cxw42]

@xuhdev Thanks for looking into the static-build issue!

I am continuing to investigate the backslashes. The problem with the Windows builds seems to be twofold:

• CTest is checking its input strings for backslash escapes, and warning if they aren't valid, so we need \\.
• But CTest is also passing the strings through unchanged to the shell, so we need \!

I am checking whether I can move the $BackSlashes variable where CMake won't process it but CTest will. Edit https://cmake.org/cmake/help/v3.12/prop_dir/TEST_INCLUDE_FILE.html#prop_dir:TEST_INCLUDE_FILE

Edit The backslashes differ between editorconfig-core-c and editorconfig-vim because the former was using add_test(NAME ... COMMAND ...) and the latter was using add_test(...). The former processes the arguments in a way the latter does not.

[cxw42]

@xuhdev I am logging off for now and plan to work on this again tomorrow. o/

[xuhdev]

Thanks! #69 should have resolved the static build issue -- feel free to rebase

[cxw42]

@xuhdev Yep, Travis build passed :) . Thanks for the quick fix!

[xuhdev]

🎆 🎆 🎆