HELP! nvidia-driver patch: Weird REPLACE_CMD and command-line-sed behaviour. HELP!
John Hein
jhein at symmetricom.com
Tue Sep 10 16:11:14 UTC 2013
John Hein wrote at 10:07 -0600 on Sep 10, 2013:
> O. Hartmann wrote at 17:55 +0200 on Sep 10, 2013:
> > On Tue, 10 Sep 2013 09:50:05 -0600
> > John Hein <jhein at symmetricom.com> wrote:
> > > Actually I should have said you want:
> > >
> > > .if ${OSVERSION} > 1000055
> > >
> > > The API change to vm_map_find was made between shortly before 1000055.
> >
> >
> > I'm on FreeBSD 10.0-CURRENT #1 r255434: Tue Sep 10 02:01:02 CEST 2013
> > amd64 and when calling
> >
> > make -VOSVERSION in x11/nvidia-driver I receive
> >
> > 1000054
>
> You are in an unfortunate window between the moment the API change was
> committed (255426) and the OSVERSION was bumped (255449). That
> shouldn't happen ideally, but sometimes it does. The "backward" gap
> from 255426 to the last bump (255321) is larger than the "forward"
> gap, and in absence of better guidance, you should generally pick the
> OSVERSION larger than a particular API change to key off.
>
> Your false 1000054 is really 1000055 in spirit, but human committers
> just didn't catch on until a bit later. So you have to pick the
> lesser of two evils - which I am claiming is 1000055.
So if you buy that argument, I have a correction. What I meant was:
.if ${OSVERSION} > 1000054
(I had >= 1000055 in my head)
More information about the freebsd-ports
mailing list