Re: [udd] tracking cloned bug relationships
On 22/01/10 at 10:07 +0100, Stefano Zacchiroli wrote:
> On Fri, Jan 22, 2010 at 01:04:18PM +1300, Lucas Nussbaum wrote:
> > Well, that information is not available in the bugs metadata (.summary
> > file). So we would have to parse the full bug log. Which is not
> > necessarily something I was planning to do, but if you submit a patch,
> > why not. However, I don't know where we want to go there: do we want to
> > import the full bug logs into UDD?
>
> I've no idea whether it is feasible in term of resources or not, do you
> have any idea of how much will it take? Also, would it be possible to do
> that preserving the current approach of reimporting from scratch
> everything at each pulse?
I don't have any hard numbers, but I think so. Most of the time used by
the bugs gatherer is spent reading the files from disk (the "several
small files per bug" approach is incredibly unefficient) and, if I
remember correctly, we are already reading the full bug log. The
additional parsing + SQL inserts would of course slow things down, but
probably not more than by a factor of 2.
And we also could do meta-data only imports frequently, and full-log
imports only once a day or every two days.
> *If* is is feasible to do the bug log import, I've an interesting hack
> to test though, which is to extract the actual patches recognizing them
> heuristically either among attachments or as inlined diff. Once we have
> that, it would be quite useful to have a devscripts or the like that
> downloads patches from bug numbers.
Well, don't put too high hopes into this. For this to work, we would
likely need to create some meta-data when parsing the bug log, because
it's pretty raw currently. For example, we would need to find emails
where the patch tag was added (by parsing the mails and looking for "tag
nnn + patch"). This is very error-prone etc.
--
| Lucas Nussbaum
| lucas@lucas-nussbaum.net http://www.lucas-nussbaum.net/ |
| jabber: lucas@nussbaum.fr GPG: 1024D/023B3F4F |
Reply to: