I wrote:
> Cool, I'll go try changing all those conditions to use the msys test.
OK, that worked: all four relevant buildfarm members are now showing
the expected test failure. So I'll go fix the original bug.
Should we consider back-patching the CRLF filtering changes, ie
91bdf499b + ffb4cee43? It's not really necessary perhaps, but
I dislike situations where the "same" test on different branches is
testing different things. Seems like a recipe for future surprises.
regards, tom lane