Re: pkgquilt

Alistair Crooks <[email protected]> Sat, 13 Jun 2026 12:36:54 -0700
Newsgroups gmane.os.netbsd.devel.packages
Message-ID <CAN5gJXpt6t3hiY6FJvu3Ti0QjxztyUNwKt4eBv1rgWpdu2WhzQ@mail.gmail.com>
On Sat, 13 Jun 2026 at 12:19, David Holland <[email protected]>
wrote:

>
> You say "simply", but in practice it's a pain, and it's not an
> accident that sometimes patches just get thrown away when updating.
> (And then fixed problems come back to life and someone has to spend
> time re-debugging them, and believe me it's not fun to waste an
> afternoon fixing a problem with a package you depend on to discover
> that you'd already fixed the exact same problem with nearly the exact
> same patch two years ago, and somebody just blithely deleted the patch
> while updating.)
>
> The order issue is irrelevant, we've always had ordered application of
> patches.
>

This isn't true - in the very early days, we had patches that overlapped,
and we got into trouble with the order of application of patch-* in at
least one case that I can remember (this is when we had occasional multiple
source files patched in the one patch file). So we introduced a policy that
it was one patch file per source file. I'm glad to see the policy worked
well, though :)

Much later, we started the move from the patch-aa form of patch file to the
patch-<normalised-filename> style (and which I have come to hate, because
syntax-coloring fails drastically on diff files which some editors infer to
be standard source files. But I'll live)

All in all, though, I find myself much more sympathetic to Benny's and
Greg's argument - pkgquilt seems overkill to me, especially for the
standard pkgsrc case.

Best,
Al

PS. If you find yourself asking if people have read your mail, maybe the
problem isn't with other people - perhaps consider compressing the points
you're trying to make into a much smaller piece?