Comment 23 for bug 25865

Message-ID: <email address hidden>
Date: Thu, 24 Nov 2005 22:49:57 +0100
From: Matthias Klose <email address hidden>
To: "<email address hidden> Adeodato =?utf-8?B?U2ltw7M=?=" <email address hidden>
Cc: <email address hidden>
Subject: Re: libstdc++ allocator change

Adeodato =?utf-8?B?U2ltw7M=?= writes:
> * Matthias Klose [Thu, 24 Nov 2005 22:29:15 +0100]:
>
> > right, that approach looks better, although currently you have to
> > rename both the package.
>
> I discussed it with vorlon, and we were going with the "make libpcre3
> conflict with everything that used libpcrecpp.so, which we expect will
> be very little" option. Do you see some sort of problem with this
> approach? (If I'm not mistaken, has been used for at least one package
> in the previous ABI transition).

well, I disagree, because that leaves out any 3rd party package, which
you don't know. and vorlon did agree on that? *cough*

let's take the ways we did go incompletely for libgmp3:

- split libpcre3 into libpcre3c2a and libpcre3++c2a
- libpcre3c2a depends on libpcre3++c2a
- libpcre3c2a's shlibs file reads:

 libpcre 3 libpcre3c2a | libpcre3 (>= 4.5)
 libpcreposix 3 libpcre3c2a | libpcre3 (>= 4.5)

This way, packages depending on the renamed package can move to
testing before pcre3 moves to testing.

the shlibs file can be reduced once the new pcre3 is in testing.

  Matthias