On Mon, 2011-04-04 at 01:19 -0400, Christopher Faylor wrote: > There is a __INSIDE_CYGWIN_NET__ which I apparently added ten years ago > but my ideas about naming have changed. I also added > USE_SYS_TYPES_FD_SET which is closer to what I now prefer but it should > have had some leading underscores.
I could just manually define __NO_INLINE__; there's no net effect on the resulting object file. I suppose YA alternative would be to define these as cygwin_dev_* and then export them in cygwin.din as gnu_dev_*, but that does seem to be an overkill. > Maybe the functions were added to gcc before it had the ability to force > inlining. Not AFAICS; originally they were simple macros like we currently have, but for glibc-2.3.3 (Sept 2003), they were changed, adding both inline and exported functions simultaneously[1]. Of course, this is not unprecedented: note ntohl and friends, for which we also provide both inline and exported. But as I've mentioned before, that where a function is expected, providing a macro or inline function by itself is inadequate; an autoconf AC_CHECK_FUNC or cmake CheckFunctionExists call will result in a false negative if a real function is not present as well. (Although I do concede that is seems unlikely in this particular case.) Yaakov [1] http://sourceware.org/git/?p=glibc.git;a=commitdiff;h=9a276f8
