On Sat, Aug 13, 2005 at 11:22:22AM -0400, Todd Vierling wrote:
> On Wed, 10 Aug 2005, Richard Rauch wrote:
> 
> > Of course, now I have to convert the patches all back to mixed-case,
> > including the diffed-against-/dev/null Makefile.
> 
> You can put brand-new files into a "files" subdirectory of the package,
> rather than creating a new-file patch, and then use the following to put it
> in the source tree:
> 
> post-extract:
> 	${CP} ${FILESDIR}/Makefile ${WRKSRC}/
> 
> (...or similar.)

Hm.  A little less obvious...  Well, I diff'ed against /dev/null
since that's what I've seen other packages do, and it's done now.
(^&

But thanks for the tip.


> > Should I still see this as a pkgsrc bug(let), or should I consider
> > this an acceptable solution to the problem?
> 
> It's quite appropriate for the default unzip args to remove CRs and
> case-fold names.  This is because a vast quantity of zipped sourcefiles are

Yes, but then shouldn't other work against the same archive also
case-fold names?  In particular, while extraction does case-fold
the names, patching seems to go back to the archive and not
case-fold.  (That's what I remember, anyway.  That was in short-term
memory and all happened more than a day ago.  (^& Anyway, the
"bug(let)" that I was talking about was the inconsistency, not
the policy of case-folding.)

I.e., shouldn't pkgsrc be consistent about this?

-- 
  "I probably don't know what I'm talking about."  http://www.olib.org/~rkr/