[debpool] Re: Should Allow_Version distinguish different archs?

[Magnus Holmgren]

On Tuesday 24 April 2007 11:17, Free Ekanayaka wrote:
> |--==> Magnus Holmgren writes:
>   MH> The $VersionDBs have to use $arch in the keys, and Set_Versions(),
>   MH> Get_Version(), and Allow_Version() has to take it into consideration.
> I'm on MH> it, but I have to check out a new copy to be able to commit just
> that change MH> to Packages.pm, and also I don't know if it's worth it if
> the database is MH> going to change radically.
>
> Well, it's something that is probably needed to get true multi-arch
> support, unless we decided to "support" this feature through the
> rollback option, which is bit hacky.

There's an additional issue: DebPool can't handle binNMUs, AFAICT. Those have 
the property that the Version field in the .changes file (the "meta version") 
is greater than the source version, and typically equal to the binary version 
(although binaries can, as you know, have completely different version 
numbers) of that particular arch. The thing is that there can be an arbitrary 
number of binNMUs for each arch: for example, two rebuilds on i386, +b1 and 
+b2, then a rebuild on amd64, also +b1. This means that the "meta version" 
can go up and down, and we should *not* implement special handling for +bN 
version numbers - that's strictly dpkg-genchanges's business. What the heck 
is a "meta version" anyway? I say we get rid of it.

-- 
Magnus Holmgren        holmgren at lysator.liu.se
                       (No Cc of list mail needed, thanks)
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
Url : http://lists.alioth.debian.org/pipermail/debpool-devel/attachments/20070424/14775760/attachment.pgp