[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: C++ ABI change for etch -- freeze unstable for all C++ libs with changed or new sonames

On Sun, Jun 05, 2005 at 03:04:45PM +0200, Petter Reinholdtsen wrote:
> [Matthias Klose]
> > Details of the planned C++ ABI change can be found at
> > 
> >   http://lists.debian.org/debian-release/2005/04/msg00153.html

> There I find this remark:

>   Appended is an updated version, the most notable change is to drop
>   the 'c102' suffix from packages, if it exists. This way, we get rid
>   off the "ugly" extension, and we don't support direct upgrades from
>   woody to etch anyway.

> How will this work for already installed non-debian binaries.  I am
> talking about binaries installed a long time ago by the system
> administrator, using the C++ ABI in woody.  If this third party
> package depends on libfoo (with old C++ ABI, before c102 was added),
> how do we avoid that the program break when the machine in question is
> upgraded from sarge to etch?

> To elaborate, I talk about a machine installed with woody, where
> someone built their own package with some binary using the woody C++
> ABI, next, they upgrade to sarge and get libfooc102 in addition to the
> libfoo library they are using, and then when etch is released, they
> upgrade to etch, and libfoo from woody is replaced with libfoo from
> etch, with a completely different C++ ABI.  Is this the way it will
> work?  Do we want it?

Since the C++ transition *to* c102 naming involved conflicting with the
previous version of the library (since the package name changed but the
soname -- and therefore the filename -- did not), such third-party binaries
are already unusable in sarge unless the user did not install the c102

So, this is only a problem for users that didn't have anything on their
system which depends on the sarge version of the library.

> (I suspect we could avoid it by making sure all libraries using the
> c102 prefix use the c2 prefix in etch.)

True, to the extent that it applies.

Steve Langasek
postmodern programmer

Attachment: signature.asc
Description: Digital signature

Reply to: