tech-pkg archive

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index][Old Index]

Re: pkgtools/pkgdiff: mkpatches, ignore time stamp for unchanged patches



Jonathan Perkin <jperkin%joyent.com@localhost> writes:

>> I think it would be good if mkpatches declined to update patchfiles if
>> they weren't actually meaningfully different, at least by default.  I
>> find it annoying when I run mkpatches and e.g. see 12 .orig files, when
>> examination shows that really there's just the new patch I added and 3
>> that were defuzzed, with 8 unchanged.
>> 
>> I guess this comes down to suggesting that mkpatches be "mkpatches;
>> patchdiff".  Does anyone ever want to not have this behavior?
>
> Me.  In my opinion patches should always be appropriate to the version
> currently in pkgsrc, and it should be up to the person who updates the
> package version to ensure that they are, rather than relying on fuzz.
>
> This helps a great deal when fixing packages without bumping the
> version.  It means that the diffs resulting from mkpatches are only
> those which are caused by your change, and that makes it much easier
> to verify that you aren't accidentally changing something else.

I think there are multiple cases, and we may be talking about different
things (due to me being unclear):

  a: patch is actually different

  b: defuzz only (with possible datestamp change)

  c: change of datestamps only

  d: no actual change (patch-foo and patch-foo.orig are identical)


I don't want mkpatches to ever change patch files for category d.  Right
now I get the original patchfile moved to patch-foo.orig and a
byte-for-byte-identical patch-foo witha new mod date.  That's just
noise/mess.

For category c, I also don't want a change.  I don't know why anyone
would.

For category a, a change is necessary.

So the only interesting case is b, and I can see reasons to do it either
way sometimes, as you say.

So if

  mkpatches --no-defuzz

changed category a only, and

  mkpatches

changed category a and b, and

  mkpatches --regenerate-timestamps

did a, b and c, would we all be happy?


Attachment: pgpFLUCaju9kW.pgp
Description: PGP signature



Home | Main Index | Thread Index | Old Index