Port update needs commit

[Bryan Blackburn]

On Fri, Nov 21, 2008 at 03:56:15PM -0500, Luis O'Shea said:
> > On Thu, Nov 20, 2008 at 08:25:00PM -0500, Luis O'Shea said:
> >> Could someone commit my update to the asymptote port?  See Ticket
> >> #13249 (http://trac.macports.org/ticket/13249).  Note that this is a
> >> closed ticket, but I have been using it to attach updates.
> >
> > That's probably why it wasn't noticed, just create a new ticket for port
> > updates.
> 
> Fair enough.  In previous updates to this port I have attached diffs
> to the Portfile to the original ticket as requested by another
> committer.  See
> 
>   http://trac.macports.org/ticket/15936#comment:3

Perhaps Simon was referring to just the bit in your comment about "It might
be useful to factor out the version of the gc lib into its own variable (in
a separate commit)."?  At least I hope so, since the "Port Updates"
milestone in trac is meant for exactly what your new patch does.

> 
> However this time I attached the diff to the wrong ticket.  In future
> I'll follow your suggestion.
> 
> > However, trying 1.51, I get errors on 10.5.5/Intel Xcode 3.1.1:
> >
> > /usr/bin/g++-4.0 -Wall -ansi -DHAVE_CONFIG_H -D_FILE_OFFSET_BITS=64 -DUSEGC
> > -I/opt/local/include -pipe -O2  -I . -Igc-7.1/include -o drawsurface.o -c
> > drawsurface.cc
> > drawsurface.cc: In member function 'virtual void
> > camp::drawSurface::render(GLUnurbs*, double, const camp::triple&, const
> > camp::triple&, double, bool)':
> > drawsurface.cc:260: error: invalid conversion from 'GLvoid (*)(...)' to
> > 'GLvoid (*)()'
> > drawsurface.cc:260: error:   initializing argument 3 of 'void
> > gluNurbsCallback(GLUnurbs*, GLenum, GLvoid (*)())'
> > make: *** [drawsurface.o] Error 1
> 
> I am (still) on 10.4 (and don't have access to a 10.5 machine).  I
> will consider upgrading.  In the meantime, what is macports policy in
> this situation (i.e., we have a port which builds on one version of
> the OS but not another)?  Is to wait until it builds on all versions
> before committing?

If something is definitely known to not work on a given platform, and would
require extensive work to fix, then you can error out as necessary.  For
example, some ports work on 10.5 only (Lingon, Smultron, and others).

Note that the current version, 1.46, builds fine here, and my personal
policy is that I won't commit something if I can't at least verify it succeeds
through the destroot phase.  It would appear this issue has been fixed at
least in svn though:

<https://sourceforge.net/tracker2/?func=detail&aid=2262140&group_id=120000&atid=685683>

If the patch can be found, we could always apply it to 1.51, or wait until
1.52.

Bryan


> 
> Thanks,
> 
> Luis