AaronBallman wrote:
> > I think this is probably Clang 21 material
>
> Yeah, probably
First off, I really appreciate this patch because (to me) it's not really
defensible in 2025 that falling off the end of a function with a non-void
return type is anything but an error. So thank you!
I do think this is Clang 21 material though because I suspect the fallout from
this might be a challenge. Specifically, I wonder about `autoconf`; that tool
often plays fast-and-loose with the generated code and it relies on the
presence of diagnostics to say "nope, this is not supported". So I worry there
will be autoconf scripts out there which do `int test() { malloc(1); }` to test
for whether `malloc` exists and it now fails because it falls off the end of a
non-void function and so it gives the wrong answer. However, I don't have
evidence of this myself. Is that something you've looked into by any chance? If
not, it would be worthwhile to know what the fallout is. CC @thesamesam who
maybe has insights into what kind of fallout this will cause (or may be willing
to do a test-run with the patch?)
https://github.com/llvm/llvm-project/pull/123470
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
