[PATCH] D150966: [clang] Don't define predefined macros multiple times

[Aaron Ballman via Phabricator via cfe-commits]

aaron.ballman accepted this revision.
aaron.ballman added a comment.
This revision is now accepted and ready to land.

LGTM



================
Comment at: clang/test/Preprocessor/predefined-macros-no-warnings.c:5
+// warnings suppressed by default.
+// RUN: %clang_cc1 %s -E -o /dev/null -Wsystem-headers -Werror -triple arc
+// RUN: %clang_cc1 %s -E -o /dev/null -Wsystem-headers -Werror -triple xcore
----------------
john.brawn wrote:
> aaron.ballman wrote:
> > So the expectation is that any warnings that are emitted would be upgraded to an error and the test would be flagged as a failure because %clang_cc1 would return nonzero in that case?
> > 
> > (I was thrown for a loop by not using `-verify` and `// expected-no-diagnostics`)
> > 
> > Pretty sure `-Eonly` is equivalent (it runs the preprocessor without emitting output, so no extra overhead from piping to /dev/null).
> Yes the intent is for the test to fail on any warning. Using `-Werror` and not `-verify` means that on failure you get which macro caused the error:
> ```
> Exit Code: 1
> 
> Command Output (stderr):
> --
> <built-in>:351:9: error: redefining builtin macro [-Werror,-Wbuiltin-macro-redefined]
> #define __LITTLE_ENDIAN__ 1
>         ^
> 1 error generated.
> 
> --
> ```
> whereas `-verify` just outputs
> ```
> Exit Code: 1
> 
> Command Output (stderr):
> --
> error: 'error' diagnostics seen but not expected:
>   Line 351: redefining builtin macro
> 1 error generated.
> 
> --
> ```
> 
> Using `-Eonly` makes sense, I'll do that.
Oh, that's a handy failure mechanism, thank you for pointing that out!


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D150966/new/

https://reviews.llvm.org/D150966