[csw-maintainers] [policy] Re: feature patching, and naming

[Dagobert Michelsen]

Hi Peter,

Am 08.02.2011 um 09:38 schrieb Peter FELECAN:
> Dagobert Michelsen <dam at opencsw.org> writes:
>> Am 08.02.2011 um 02:45 schrieb Ben Walton:
>>> Excerpts from Peter Bonivart's message of Sun Feb 06 00:06:37 -0500 2011:
>>> 
>>>>> 1.2.3,pi,REV=YYYY.MM.DD
>>>> 
>>>> This is the best in my opinion. Let's treat it as flags and only
>>>> allow to pick from a fixed list to keep it from getting carried away
>>>> and be easily checked.
>>> 
>>> Agreed.  Does anyone _not_ like this choice?
>>> 
>>>> p = patched
>>>> i = i386 only
>>>> s = sparc only
>>> 
>>> Works for me.  Anything else that would be useful to define at the
>>> outset here?
>> 
>> Two things:
>> - Lets remove i/s as it is good to always release a bundle. The i386 only
>>  was used in the past if someone made a manual mistake during packaging and
>>  wanted to respin i386 only. This is IMHO generally bad. If an error occurs
>>  all packages should be rebuild. Introducing extra complexity to allow for
>>  manual patching is not a good idea. We should focus on full automation and
>>  the flag is useless for this.
> 
> I generally agree with you but I have the feeling that if there is a
> packaged project which have sense only on a given platform it should be
> possible to deliver only that; but this doesn't require the usage of a
> flag, it only requires the relaxation of the "bundle" rule.

What you say is already the case, see adobereader. But it is not what the
flag was used in the past: it indicated if a package was re-released
with the same version but only "fixed" on either sparc or i386. This
is IMHO unnecessary and confusing.

>> - If the "p" flag is present there should be checkpkg-check that
>>  /opt/csw/share/doc/<catalogname>/README.CSW is present.
> 
> Absolutely right.

:-)


Best regards

  -- Dago