Re: suggested fix

[Samuel Thibault <sthibault@debian.org>]

Nicholas Bamber, le Thu 21 Jun 2012 13:04:17 +0100, a écrit :
> ++/* As per #678358, Hurd defines AF_LINK but does not
> ++ * provide the functionality. When that bug is fixed
> ++ * These three lines could be removed and #678375 closed.
> ++ */
> ++#ifdef __GNU__
> ++#undef AF_LINK
> ++#endif

I'm generally against such kind of hacks, because we then have to
remember where they had been introduced once the actual issue is
resolved, and get yet another upload there instead of a mere buildd
giveback.  I understand that getting "undefined AF_LINK" error message
is clearer to people, but here the maintainers of the package already
understand what the matter is, so they don't actually need the clearer
message.

Btw, the Hurd itself doesn't actually define AF_LINK, it's just that
AF_LINK was added to the generic glibc header instead of system-specific
header. Same for mremap flags which were discussed earlier in another
thread.

Samuel