On Fri, Jul 19, 2013 at 7:09 AM, Jonathan Schleifer <[email protected]> wrote: > Am 19.07.2013 um 03:02 schrieb Eli Friedman: > >> Ah, I figured out what's happening. The other invocation is the one >> that's failing. Try changing your "STRET" prefix to "HASSTRET". :) > > Interesting. At one point, I tried WITHSTRET and WITHOUTSTRET, but that > wouldn't help. Now it indeed works, though I wonder why it worked when I > manually invoked it and didn't when it was invoked using llvm-lit. > > Anyway, this means the last missing piece is the documentation. I looked at > LanguageExtensions.rst, but this does not seem like the right place to > document it, considering it is such a small change. What would be a good > place to document __has_feature(objc_msg_lookup_stret)? (Btw, I thought some > more about a better name and came to the conclusion that anything containing > forwaring, like objc_stret_forwarding, is completely wrong, as that is a > feature the runtime / framework has to support. The objc_msg_lookup_stret > call just makes it possible for the runtime / framework to correctly > implement stret forwarding).
That's the place where every single other __has_feature is documented. I mean, if it's important enough to add a __has_feature name for it, it's surely important enough to document the feature in question. -Eli _______________________________________________ cfe-commits mailing list [email protected] http://lists.cs.uiuc.edu/mailman/listinfo/cfe-commits
