[cfe-commits] r169778 - /cfe/trunk/test/CodeGen/code-coverage.c

Rafael EspĂ­ndola rafael.espindola at gmail.com
Mon Dec 17 20:14:09 PST 2012


On 17 December 2012 14:22, Bill Wendling <isanbard at gmail.com> wrote:
> Hell if I know. I didn't want to bother with it anymore.

Hi Bill,

I don't think that is an appropriate attitude. A clean testsuite is
one of llvm's strong points and a failing test should be investigated,
not ignored.

Coincidentally, this happens to be a perfect example. The test was
failing because of an uninitialized variable in llvm. So your patch to
clang was correct, but would only have the desired effect if you got
lucky with what was stored in the GCOVProfiler constructor.

I have fixed the underlying issue and added the test back. I hope you
will be more careful in the future.

> -bw

Cheers,
Rafael



More information about the cfe-commits mailing list