Re: sgml-base related conffile prompt
On 2013-04-08 10:58, Helmut Grohne wrote:
> Dear release team,
>
> TL;DR: This issue is fixable with 18 binNMUs of which 14 are arch:all.
>
> On Sun, Apr 07, 2013 at 05:46:57PM +0100, Adam D. Barratt wrote:
>> We can move discussion of the conffile issue to a new bug / thread if
>> needed.
>
> Let me give you some data on this. The basic issue was first found by
> Andreas Beckmann as #681194. Packages built with squeeze debhelper would
> create package catalogs using packaging scripts and only remove them on
> purge. With the wheezy version package catalogs were turned into
> conffiles, but I forgot to properly handle the removed-but-not-purged
> case. Upon installation of a rebuilt package the previous left-over
> package catalog would be treated as a modified configuration file. This
> was fixed in debhelper/9.20120830 (migrated). It can indeed affect
> upgrades from squeeze to wheezy, if a package removed by an
> administrator is later reintroduced as a dependency during the upgrade.
>
> [...]
Just if I understand it correctly - the requirement for triggering this
bug is to:
* install the Squeeze version of one of the affected packages below
* remove said package (but do not purge it)
* install the Wheezy version of the affected package
Is that correctly asserted of me?
>
> Here is the result from my sid work machine (as a lower bound of what
> would need to be fixed).
>
> =====
> [...]
> jade_1.2.1-47.1+b1_amd64.deb: affected
- has lib, but it is not M-A:same AFAICT
> [...]
> linuxdoc-tools_0.9.68_amd64.deb: affected
- no libs
> [...]
> openjade1.3_1.3.2-11.1+b1_amd64.deb: affected
- no libs
> openjade_1.4devel1-20.1+b1_amd64.deb: affected
- has lib, but it is not M-a:same AFAICT
> [...]
>
> [...]
>
> What do you think about the RCness of this issue? Should it be fixed for
> wheezy? Adam already indicated that he leans towards "no".
>
> Helmut
>
>
If my assertion above is correct, then I inclined to agree. Though if
there are other ways to trigger the issue in these packages we might
want to at least fix a couple of arch:all cases as well (e.g. sgml-base
has a popcon of 70k and it is not the only one).
Plus we might as well get those 4 packages binNMU'ed regardless since
they are practically "free" fixes.
~Niels
Reply to: