i686-mingw32-RA-on-linux buildbot

I keep running into errors on the i686-mingw32-RA-on-linux bot and was curious if someone could test a patch against the bot or something with a similar setup. I believe it should now work properly, but would like to make sure that it’ll pass before possibly adding a third commit/revert cycle.

If the history isn’t an issue, let me know an I’ll just keep trying.

The differential is here: https://reviews.llvm.org/D30107.

FWIW, the errors have been from inconstancies with the ming available here:

Slightly different header version causing a missing function.

Mounted to a case insensitive volume.

Thanks.

It’s mine. Investigating.
You may ask me @chapuni in the IRC, oftc#llvm.

Thanks a lot.

Also the relevant test for the changes is

unittests/Support/DynamicLibrary/DynamicLibraryTests

Hoping it should easily pass if the compile errors have been irradicated.