[Reportbug-maint] Bug#522010: reportbug: FTBFS with bin-NMU version

[Sandro Tosi]

On Tue, Mar 31, 2009 at 18:56, Daniel Schepler
<schepler at math.berkeley.edu> wrote:
> On Tuesday 31 March 2009 09:14:27 Sandro Tosi wrote:
>> And since you don't need to only schedule a simple rebuild, I don't
>> see the problem in editing also reportbug/__init__.py .
>>
>> Moreover, I can't see why you would re-build reportbug on your machine
>> and install it; if you do it to change something in the code (for
>> example because you're so kind and want to prepare & send us a patch
>>
>> :) ), you can also edit __init__.py ; if you do it to bring in other
>>
>> versioned depends, we define none, only some versioned Suggests, so I
>> still fail to understand the request itself.
>>
>> I hope I've clarified our situation, so if you want you can clarify
>> better your position :)
>
> OK, let me explain the context.  I'm running a home-brew "pbuildd" script,
> whose purpose is to rebuild the Debian archive (main) using pbuilder, starting
> from the base packages.  Essentially, it just repeatedly goes through the list
> of packages yet to be built, trying to build each one and moving the results
> into a local repository if it succeeds.
>
> I also like to install the locally built packages on the outside system as a
> bit of a sanity test, to check that they're not completely broken.  But if I
> also have the official archives in /etc/apt/sources.list, that tends to confuse
> apt -- for example, sometimes apt will repeatedly think a package is out of
> date and want to upgrade it, but end up installing the same version as before.
> So that's why I have the pbuilder hook script which adds an appropriate "+pbN"
> version to debian/changelog (and that's also useful to make sure an upgrade
> happens when I schedule a rebuild of a package locally for updating
> dependencies, etc.).
>
> I do have a "blacklist" for the hook script for packages where the maintainer
> has explicitly refused to allow this to work (for example, the kernel image
> packages).  But most of the time when something like this bug comes up, it
> works well to have the code extracting the debian/changelog version just chop
> off any +* from the end.

Well, I don't know if it's right or wrong to support this, but since
it's a simple sed expression, I made the change in our repo to strip
off all '+.*' at the end of version string in debian/changelog.

Cheers,
-- 
Sandro Tosi (aka morph, morpheus, matrixhasu)
My website: http://matrixhasu.altervista.org/
Me at Debian: http://wiki.debian.org/SandroTosi