>    mainly
>    as a way of using tar to extract a new pkgsrc.tar.gz over an existing
>    pkgsrc tree.
> 
> Hmm.  I thought it was mostly to catch people who weren't using "sup
> -o" or its moral equivalent to update pkgsrc.

Whatever method you care to leave old patch files lying around -
it doesn't bother me...

> I think exiting on error in this case is more likely to keep things
> reliable (and keep crud from building up in people's pkgsrc trees) in
> the long run.

I disagree - if a source build fails for a user, they now either
have to go and find out the file which caused the failure, delete
it, and then continue on.  They need to do this once for each of
the files that failed. Either that, or they need to run "make
makepatchsum", which is not meant as an end-user command. This all
exposes the internals of pkgsrc to a potentially naive user.

A cleaner method of implementing Hubert's modifications, and one
which I discounted as being nowehere near good enough, is to echo
the message should the patch(1) command fail.

Regards,
Alistair