UDD and bugs assigned to two or more packages

2008-09-26 Thread Frank S. Thomas
Hi all,

I've just toyed a bit with the UDD and probably discovered a bug:
#445875 is assigned to two binary packages that belong to the same source 
package but UDD lists as 'source' the same as the bug's 'package':

udd=> select package, source from bugs where id = '445875';
  package   |   source
+
 boinc-client,boinc-manager | boinc-client,boinc-manager

So queries with "where source = 'boinc'" do not include #445875. On the other 
hand Debbugs knows that this bug belongs to the boinc source package:

$ bts select source:boinc | grep 445875
445875

So UDD should also know that this bug belongs to boinc and should not list 
binary packages in the 'source' field, right?

Cheers,
-- 
Frank S. Thomas <[EMAIL PROTECTED]>   PGP public key ID: 0xDC426429
Debian Developerfinger fst/[EMAIL PROTECTED]


signature.asc
Description: This is a digitally signed message part.


Re: UDD and bugs assigned to two or more packages

2008-09-26 Thread Lucas Nussbaum
On 26/09/08 at 12:52 +0200, Frank S. Thomas wrote:
> Hi all,
> 
> I've just toyed a bit with the UDD and probably discovered a bug:
> #445875 is assigned to two binary packages that belong to the same source 
> package but UDD lists as 'source' the same as the bug's 'package':
> 
> udd=> select package, source from bugs where id = '445875';
>   package   |   source
> +
>  boinc-client,boinc-manager | boinc-client,boinc-manager
> 
> So queries with "where source = 'boinc'" do not include #445875. On the other 
> hand Debbugs knows that this bug belongs to the boinc source package:
> 
> $ bts select source:boinc | grep 445875
> 445875
> 
> So UDD should also know that this bug belongs to boinc and should not list 
> binary packages in the 'source' field, right?

There are two issues:
(1) a known problem with the way UDD identifies the source package for a
binary package, for bugs. You can't just look at the packages table for
that, because lots of bugs are filed against packages that are no longer
in any suite. We just need to copy the way bugreport.cgi does this (help
wanted!)

(2) the way we deal with bugs affecting two (or more) packages.
That's a tricky problem. On one hand, the correct way to solve that
would be to have a separate "affected_packages" table, with (id,
package) columns, and also an "affected_sources" table.

On the other hand, that's makes all queries more complex, because you
have to join this additional table (and it's not going to be a small
table). Currently, there are only 78 unarchived bugs (out of 69982)
which are affected to 2 or more packages (select count(*) from bugs
where package like '%,%').

One possibility would be to keep the package/source columns in bugs, but
add tables with (id, package) and (id, source). That way, people not
interested in this case can join the additional table, while others can
just ignore it. It would only increase a bit the time needed to import
the bugs (about half an hour currently, only I/O bound).

What do people think?
-- 
| Lucas Nussbaum
| [EMAIL PROTECTED]   http://www.lucas-nussbaum.net/ |
| jabber: [EMAIL PROTECTED] GPG: 1024D/023B3F4F |


signature.asc
Description: Digital signature


Re: UDD and bugs assigned to two or more packages

2008-09-26 Thread Kurt Roeckx
On Fri, Sep 26, 2008 at 01:10:52PM +0200, Lucas Nussbaum wrote:
> 
> (2) the way we deal with bugs affecting two (or more) packages.
> That's a tricky problem. On one hand, the correct way to solve that
> would be to have a separate "affected_packages" table, with (id,
> package) columns, and also an "affected_sources" table.
> 
> On the other hand, that's makes all queries more complex, because you
> have to join this additional table (and it's not going to be a small
> table). Currently, there are only 78 unarchived bugs (out of 69982)
> which are affected to 2 or more packages (select count(*) from bugs
> where package like '%,%').
> 
> One possibility would be to keep the package/source columns in bugs, but
> add tables with (id, package) and (id, source). That way, people not
> interested in this case can join the additional table, while others can
> just ignore it. It would only increase a bit the time needed to import
> the bugs (about half an hour currently, only I/O bound).

I was thinking about adding the bug two times in the bugs table.


Kurt


-- 
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]