cvs commit: src/sys/sys mount.h src/lib/libc/gen Makefile.inc
Jacques A. Vidrine
nectar at FreeBSD.org
Fri Apr 16 08:25:23 PDT 2004
On Fri, Apr 16, 2004 at 05:16:40PM +0200, Ollivier Robert wrote:
> According to Maxime Henrion:
> > and I would actually be surprised if there was any. I already bumped
> The external ones are the problem but I agree that there are probably none.
> > __FreeBSD_version for this change and it looks sufficient to me. Can
> > you explain why you think a major number bump of libc is in order here?
> Because that how you play the major number game in shared libraries world.
> Anytime you change (i.e. remove a function or change a function's argument
> in a way that's incompatible with the prevous one) a librabry's interface,
> you MUST (in the RFC-mode sense of the world) change the major.
> I agree with Warner that it is a major pain (and why we do this only once
> in a given branch) and that it this case, we can simply ignore the issue.
> </pedantic> :-)
More to the point is that we have it backwards: we really MUST not
make incompatible changes to the ABI in a given branch.
Of course, we have a little wiggle room before 5.x becomes -STABLE, but
it is *very little* room at this point. If removing getvfsent() would
be at all likely to affect software outside the tree, then I would
rather see a compatibility shim. However, Maxime has already insisted
that there are no other consumers, and I choose to believe him :-)
Jacques Vidrine / nectar at celabo.org / jvidrine at verio.net / nectar at freebsd.org
More information about the cvs-src