Re: Frank Küster 2006-04-03 <[🔎] 86acb226fs.fsf@alhambra.kuesterei.ch> > the backport of libpoppler creates a binary package libpoppler0c2. IIRC > the c2 suffix means that the package has been linked against the new > libstdc++ which is ABI-incompatible to sarge's. Shouldn't the packages > be renamed to libpoppler0, libpoppler0-glib etc.? Yes. Re: Daniel Baumann 2006-04-03 <[🔎] 443123AF.70708@panthera-systems.net> > If a library is not part of stable, the library packages names should > not be altered, so other backports don't need to have backport-specific > Build-Depends with respect to that very library. > > The upgrade path from stable+backports to testing is not affected by this. > > If a library was already part of stable, the library packages names > should be altered to match the stable ones *if* the soname didn't > change. Look at libextractor for an example :) That's only half of the issue here. The PITA with the C++ ABI change is that the ABI of every library broke *without* the soname changing. Therefore, we have to change the library package name to include something like "sonumber plus foo" where 'foo' is the C++ ABI version (and the corresponding shlibs file). It doesn't make a difference if the package is part of Sarge, since bpo acts as a Sarge replacement. Remember that the sarge-backports suite uses the Sarge g++, so any library in bpo that uses the Sarge C++ ABI must have a different name than in etch to enable partial upgrades (and since partial upgrades include the case "there is software locally compiled that links against this C++ lib", this is a very strong requirement). Christoph -- cb@df7cb.de | http://www.df7cb.de/
Attachment:
signature.asc
Description: Digital signature