[ On Tuesday, May 11, 1999 at 08:37:57 (-0400), Todd Vierling wrote: ]
> Subject: Re: qpopper needs gdbm? 
>
> Actually, to prevent the need to regenerate the whole script (since such
> change will never go back to the maintainers), the appropriate block in
> "configure" should be "if false ... fi" bracketed.

The problem there is that if anyone ever does need to patch the
"configure.in" file for some other reason then it's rather hairy
sometimes to figure out just what the "configure" patch was doing, not
to mention somewhat messy to undo.

Never patch generated code if you can help it, even if it's human
readable stuff.

It's trivial to have the pkgsrc system "Do The Right Thing" all the time
and I think the 'cost' is low enough that it need not be avoided.  If
nothing else it serves as a good example of how things ought to be done
when more complex problems arise.

The only time I would consider patching the "configure" script directly
are in those unlikely cases where a truly ancient package is
incompatible with the current pkgsrc version of autoconf.  Even then
it's often easier to simply patch "configure.in" so that it is
compatible.

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>