[cfe-commits] [patch] Function literal to bool conversion fixit notes

[Douglas Gregor]

On Dec 9, 2011, at 12:15 AM, David Blaikie wrote:

> 
> 
> On Thu, Dec 8, 2011 at 8:01 PM, Matt Beaumont-Gay <matthewbg at google.com> wrote:
> On Thu, Dec 8, 2011 at 19:43, David Blaikie <dblaikie at gmail.com> wrote:
> > Just adding some finishing touches to Lang's recent excursion into the front
> > end by adding notes to describe how to suppress (and in simple cases, fix)
> > the warning
> >
> > /tmp/webcompile/_7230_0.cc:3:10: warning: address of function 'f1' will
> > always evaluate to 'true' [-Wbool-conversions]
> >   return f1;
> >          ^~
> > /tmp/webcompile/_7230_0.cc:3:10: note: prefix with the address-of operator
> > to silence this warning
> >   return f1;
> >          ^
> >          &
> > /tmp/webcompile/_7230_0.cc:3:10: note: suffix with parentheses to turn this
> > into a function call
> >   return f1;
> >          ^
> >            ()
> >
> > The latter note is only included if the function has a bool return and no
> > arguments (I couldn't figure out how to account for default or variadic
> > arguments - suggestions welcome. Also for bool conversion) so this is fairly
> > simple.
> 
> You might be interested in Sema::isExprCallable and Sema::tryToRecoverWithCall.
> 
> Hey - thanks for the tip. That is really handy. It doesn't /seem/ to be doing much better than my first attempt (it doesn't seem to ever trigger the interesting case that uses FunctionDecl::getMinRequiredArguments - I'm still poking around with that. Perhaps this function can be generally improved for all callers) - and, curiously, the simple test cases I've tried ("struct foo { void bar(); }; int main() { foo f; f.bar; }") seem to have interesting behavior as you change whether the function is overloaded, callable, etc (it seems that the diagnostic caret location changes from f to bar, depending on that, among other things))

This function could/should be improved, especially to make it cope with member functions better. 

> [updated with a version using isExprCallable - and also keeping the CC location in the warning that I'd previously removed

LGTM!

> (but fixing its location by giving BinaryOperator a getExprLoc that provides the operator location which makes things read better - is that OK?)]

Yes, that's a useful improvement.

	- Doug
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-commits/attachments/20111209/b341f947/attachment.html>