svn commit: r313614 - head

[Bryan Drewery]

On 3/8/2013 1:52 AM, Ruslan Makhmatkhanov wrote:
> Bryan Drewery wrote on 08.03.2013 09:35:
>> On 3/7/2013 11:32 PM, Ruslan Makhmatkhanov wrote:
>>> Bryan Drewery wrote on 08.03.2013 07:49:
>>>> Author: bdrewery
>>>> Date: Fri Mar  8 03:49:41 2013
>>>> New Revision: 313614
>>>> URL: http://svnweb.freebsd.org/changeset/ports/313614
>>>>
>>>> Log:
>>>>     - Must force reinstall py-distribute-0.6.35 as it is seen as a
>>>>       downgrade from py-setuptools-0.6c11_3
>>>
>>> Thanks. Would you provide more details on this? I still not using pkg,
>>> so I don't understand what the problem. Just used the same instructions
>>> as in previous UPDATING entries.
>>>
>>
>> The package py27-distribute that is replacing py27-setuptools is *older*
>> in version from the old package. So a force install is needed. Same as
>> the -f flag for portupgrade. This is due to them not wanting to
>> downgrade ports by default. Even when "replacing" or "moving" origins,
>> they will not downgrade without being forced.
> 
> Yes, but they are different applications. Isn't this a bug? I believe
> the version should only be checked withing the same application
> handling, not different ones... At least portmaster handled this without
> a problem.


Probably.

Portmaster can only handle 1 "-o" at a time. It re-executes itself in an
ugly hack to replace the package. So it simply just removes the old and
installs the new.

Portupgrade and pkg(8) act on jobs/transactions of a set of packages to
upgrade. It does seem reasonable that if the origin AND pkgname have
changed then a downgrade should not have to be forced. It's just more
complexity to put into the transaction calculation since (for both) it
is now mostly a "remove all versions older than the current".

I actually think I may have fixed this case in portupgrade already, but
am not sure.

For pkg(8) I've entered this at https://github.com/pkgng/pkgng/issues/464

> 
>>
>> This also happened in devel/pkgconf update in 20120726.
>>
>> An ugly way to avoid this would be to use a PORTEPOCH higher than the
>> old in the new port.
> 


-- 
Regards,
Bryan Drewery
bdrewery at freenode/EFNet

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 899 bytes
Desc: OpenPGP digital signature
URL: <http://lists.freebsd.org/pipermail/svn-ports-all/attachments/20130308/64493181/attachment.sig>