Make the new GetStyleWithEmptyFileName test case independent
of the file system used when running the test. Since the
test is supposed to use the fallback "Google" style we now
use a InMemoryFileSystem to make sure that we do not accidentaly
find a .clang-format file in the real file system. That could
for example happen when having the build directory inside the
llvm och clang repo (as there is a .clang-format file inside
the repos).
Details
Details
Diff Detail
Diff Detail
- Repository
- rC Clang
Event Timeline
Comment Actions
Thanks for fixing this. (Sorry for not getting to it, I was away traveling until today.)
I'll add a tidy-up in D43598 to ensure all tests which call getStyle() pass in an in-memory filesystem.
Comment Actions
Actually, looks like all the other tests explicitly want a non-empty in-memory FS, so there's no good cleanup I can do.
Comment Actions
Well, it was only this new test case that failed, so I think other tests are fine (some of them have been adjusted in a similar way in the past if I remember correctly).
And our downstream integration bots has started to pull in things from trunk again, after I landed this fix.