Broken upgrade paths in FC+FE 2006-07-31

[Paul Howarth]

Michael Schwendt wrote:
> On Mon, 31 Jul 2006 16:52:25 +0100, Paul Howarth wrote:
> 
>> buildsys at fedoraproject.org wrote:
>>> perl-String-CRC32: paul AT city-fan.org
>>>   4: 0:1.4-1.fc4 (FE4)
>>>   5: 0:1.4-1.FC5 (FC5-updates)
>>>   6: 0:1.4-1.FC6.1 (FC6)
>> Shouldn't the email address attached to the report be the one for the 
>> person that needs to do something to fix this (in this case, the 
>> maintainer of the package for FC5-updates and FC6) rather than the 
>> person that caused the breakage (i.e. me, maintainer for FE4 :-( )?
> 
> Even with a complete package database (which we could query on "package
> owner(s) per package _per dist_", it would not be bullet-proof either and
> would require additional logic in its implementation.
> 
> In above case:
> 
>   1.4-1.FC5 is lower than 1.4-1.fc4
> 
> Maybe the package was moved from Extras into Core, and the Core packager
> chose an incompatible dist tag. The script cannot know that.

That's quite close to what actually happened, though in fact the package 
was in FC5 before it was in FE4.

> Compare with the following scenario:
> 
>   4: 0:1.4-2.fc4 (FE4)
>   5: 0:1.4-1.fc5 (FC5-updates)
>   6: 0:1.4-1.fc6 (FC6)
> 
> Who is to blame now? The FE4 package owner for releasing something that's
> higher than FC5/FC6? Or the FC5/FC6 package owner for a missing update?
> 
> The script could only guess.

I don't think it's a "blame" thing, it's a "who needs to do something to 
fix it" thing. So that would be the maintainer for each version for 
which there is a version with a higher EVR for an older distro.

> The script could report to all package owners involved. Anyway, we
> cannot do much about it without a complete package database. Extracting
> package owners from Core bugzilla (Components list) returns also some
> mailing-lists.

Roll on the package database then.

Paul.