Jakub Jelinek <ja...@redhat.com> writes: | Hi! | | This is a patch I had written two months ago, but concluded it to be too | risky for 4.8 at that point. | | The issue this patch attempts to solve is that compiler behaves differently | in some cases depending on if we've just warned about something or if | -Werror has been in effect and that warning has been promoted to error. | Of course the difference should be the exit code from the compiler and | the wording of the warning resp. error, but various bits of code e.g. in the | frontends check whether any errors have been reported already (or reported | while parsing some construct etc.), and if they have been, act differently | from the case when there haven't been any errors. | | The patch just counts the warnings promoted to errors in a different | counter, werrorcount, and thus each place can actually decide if it wants to | treat warnings promoted to errors as warnings, or as errors, or something | else for its purposes. | | Bootstrapped/regtested on x86_64-linux and i686-linux, ok for trunk?
It makes sense; patch OK. Thanks! -- Gaby
