On Mon, 22 Dec 2008 21:54:37 +0000
"Adam D. Barratt" <[email protected]> wrote:

> The central issue is the fact that the bug is (well, was) filed against
> the binary package "soci" which does not exist. In the past, the BTS
> attempted to guess what you meant in such cases, which wasn't always
> correct; there have therefore been some recent(ish) changes which mean
> that such bugs don't appear in a query for bugs assigned to the source
> package.
> 
> The solution is to assign the bug to "src:soci", which I've done for the
> bug you used as an example. 

That then breaks other things:

Debian Bug report logs - #504907
FTBFS with GCC 4.4: missing #include

Package: src:soci; Maintainer for src:soci is (unknown);

and still doesn't populate the source field in bts status:

$ bts status 504907
bug_num 504907
source  unknown
found_versions  2.2.0-4
date    1226092516
found   2.2.0-4 => 
keywords        patch
msgid   <[email protected]>
id      504907
severity        normal
location        db-h
log_modified    1230017765
originator      Martin Michlmayr <[email protected]>
subject FTBFS with GCC 4.4: missing #include
pending pending
package src:soci
tags    patch

Also, bug #509453 has now disappeared from the list of bugs affecting
the source package soci, only 504907 now shows up.

However, dch -a --closes 504907 does now work. (--closes 509453,
naturally, does not.)

> A naive "fix" would be to make debchange look at binary package bugs in
> the absence of any filed against the source package. Doing so would open
> a can of worms, however; for example, the archive currently contains a
> source package named "sm" which produces an "r-cran-sm" binary package
> and a "screen-message" source package which produces an "sm" binary
> package. For that reason, I'm not sure there's much that debchange can /
> should do here.

All it can do is hint in the error output that this might be a source
package. Currently, that hint is a little obscure and confusing,
especially when $bugpkg == $PACKAGE and $bugsrcpkg is undefined or '?':

dch warning: bug #509453 belongs to package soci (src ?),
  not to soci: disabling closing changelog entry
dch: Did you see that warning?  Press RETURN to continue...

That could be extended a bit to say:
dch warning: bug #509453 belongs to package soci,
 which might be the source package. disabling closing changelog entry
 To reassign to soci source use: 'bts reassign 509453 src:soci'
dch: Did you see that warning?  Press RETURN to continue...

Maybe only if $bugpkg == $PACKAGE ?

> [For reference, I believe that #509498, filed a short while ago against
> bugs.debian.org is another instance of the same underlying issue].

This does seem to be a bug in b.d.o and the same issue as 509498.

I'm thinking we could change this bug report into a wishlist for
src: to be documented in the bts manpage and maybe hinted in the error
line from debchange and mark it blocked by 509498.

To close 509498, I believe it would be necessary for 'bts status' to
receive the relevant data to populate the source field *and* not list
the package as "src:soci" in this case, just as "soci". 

-- 


Neil Williams
=============
http://www.data-freedom.org/
http://www.nosoftwarepatents.com/
http://www.linux.codehelp.co.uk/

Attachment: pgpv3bf97vqrb.pgp
Description: PGP signature

Reply via email to