cvs commit: ports/net/ekiga pkg-plist
Doug Barton
dougb at FreeBSD.org
Sat Jan 26 09:13:37 PST 2008
Joe Marcus Clarke wrote:
> On Tue, 2008-01-22 at 09:29 -0500, Mikhail Teterin wrote:
>> ????????? 21 ?????? 2008 09:21 ??, Edwin Groothuis ?? ????????:
>>>> The new package NEVER built because of the problem, that was just fixed.
>>> It's more to differentiate between working and non-working versions.
>> They are both working versions. If you need to make a distinction, you can
>> look at the $FreeBSD$-string.
>>
>> Bumping up the PORTREVISION would trigger an utterly useless rebuild on
>> thousands (millions?) of end-user computers and should not be performed
>> lightly.
While I agree with that in principle, I'm confused by your earlier
statement. Were users able to compile and install this new version of
the port previously, and only the package building was broken?
> It would trigger a rebuild which would provide a correct pkg-plist. For
> this reason alone, policy mandates any plist change requires a
> PORTREVISION bump.
Applying that policy without careful thought means that you're putting
the needs of the system ahead of what is best for our users. Stepping
away from this particular port for a second, if you have a port that
never compiled, installed, or packaged; then you fix it, bumping
PORTREVISION is meaningless either way. It doesn't affect the users or
the (non-existent) package. If only packaging is broken, bumping
PORTREVISION is not just pointless, it's actually harmful to users since
they have to rebuild something that won't actually change, and they can
just as easily pick up the plist fix on the next legitimate update.
(There is an edge case here where what is fixed in this plist update
won't be present in the next version, which would obviously require a bump.)
It's way too easy to say, "oh, this knob is shiny, let's turn it!"
without thinking of where the benefits lie for the users. I personally
like the way someone else stated the principle, "If a change affects the
package, PORTREVISION should be bumped." That makes sense, and balances
the needs of the system and the needs of the user.
Doug
--
This .signature sanitized for your protection
More information about the cvs-all
mailing list