Re: Selectively suppress -Wcast-function-type (issue 357770043 by address@hidden)

[dak]

On 2018/08/15 21:51:46, Dan Eble wrote:
> On 2018/08/15 16:43:27, dak wrote:
> > This is more a sort of logic nitpick than of actual relevance given
how GCC is
> > wired, but if we want to be using -Wno-cast-function-type , that's
what we
> > should announce and check for rather than -Wcast-function-type .

> I couldn't figure out any other way to check for
-Wno-cast-function-type
> reliably.  Attempts involving -Werror were defeated by the fact that
some
> preprocessor variable is redefined somewhere, leading to an error in
all cases.
> After spending much time to figure that much out, I was not interested
in
> spending more to figure out how to resolve it (if that is even
possible).

> I could reword the message, something like "whether the compiler might
warn
> about casting functions to incompatible types."  Would that be better?

I was "Huh?" until I tried out a few things.  g++ -Wno-blubber produces
no diagnostic while g++ -Wblubber bombs out.  Just add a comment like
# g++ -Wno-whatever compiles so we need to check for the positive option

because otherwise nobody will be able to guess what is up here.  In
general, what looks like a mistake or oversight and/or what just became
obvious through testing warrants a comment.  Don't keep things secret
between you and the compiler.

https://codereview.appspot.com/357770043/