Bug#683847: unblock: sgml-base/1.26+nmu4

[Helmut Grohne <helmut@subdivi.de>]

On Tue, Oct 09, 2012 at 11:43:19AM +0200, Julien Cristau wrote:
> I'm not interested in "you are on sid" so much as "you're upgrading from
> squeeze to wheezy".  And considering the amount of bugs this whole thing
> has uncovered (whether in the transition stuff itself, in dpkg, or
> somewhere else) I'm fairly convinced this whole thing is in the "not
> worth it" category.  And even in the "you've already upgraded"
> situation, dpkg's failing at trigger handling means I'm fairly nervous
> about the next dist-upgrade.

I do see your reasoning here. On the bright side, I can say that the
stream of bugs seems to have stopped. During the last two months the
only new thing that popped up was a revival of #680291 which is due to
xml2rfc being buggy in squeeze. It is not the case that our previous
state worked that well. Instead what we see here is simultaneous rising
of quality levels by doing more extensive piuparts tests and declaring
failures as rc instead of important.

> Not blaming you, as you couldn't have predicted most of these bugs, just
> saying at some point you have to stop the trainwreck.

Thanks. What I fail to see here is how to stop the trainwreck. You
cannot simply take the squeeze packages, bump their versions and upload.
That would severely break sid and wheezy. You would have to reverse the
transition. So what I am suggesting here is that the brake is the worse
option in terms of breakage to wheezy.

Note that even though I invested a fair amount of time in developing the
trigger based sgml-base catalog update, I am trying not to be biased by
having that work done. If you can show me a different solution, I will
try to have an honest look. "go back" is just too vague to count as a
solution at this point.

Helmut