[Rpm-metadata] createrepo, --update and the sqlite database

Oliver Falk oliver at linux-kernel.at
Mon Oct 22 07:50:29 UTC 2007


On 10/15/2007 08:55 PM, seth vidal wrote:
> On Mon, 2007-10-15 at 11:55 -0400, Jesse Keating wrote:
>> On Mon, 15 Oct 2007 11:28:55 -0400
>> Mike McLean <mikem at redhat.com> wrote:
>>
>>> Since I wrote the --update code, I guess I should be the one. I'll
>>> look at this this week. I suspect this will be nontrivial (i.e. at
>>> least as much work as the original --update patch).
>> I have to wonder just how much time is really saved here though.  I
>> thought that once createrepo is to the point of dumping to an sqlite
>> file that it has all the information from the xml files in question and
>> it doesn't have to go touching every package in the tree which is where
>> the real time delays come in.  Also the code that is generating the
>> sqlite content is compiled C and quite fast IIRC.  Are we really
>> expecting a significant speedup here?
>>
> 
> Jesse,
>  Your in large part correct. However, if we want to create database
> 'diffs' then something along these lines will need to happen anyway...
> -sv


Well. Pre-generating the sqlite database might not be the best way to
heavily speed up the builders, but it takes some time on slower archs
(in my case Alpha) and it was the easiest way to gain some speed.

(Seth... yum's arch.py is still missing ev67 :-) ).

Best,
 Oliver



More information about the Rpm-metadata mailing list