On Thu, Dec 16, 2004 at 08:13:31PM +0100, Adeodato Simó wrote: > no, there is no error on your part. the package 'gcc-defaults' gets to > decide where do the cpp, gcc, g++, etc. links point to, and at the > moment those point to 3.3 versions in both sid an sarge. > however, note that in all architectures except hppa and m68k, the > libgcc1 package is provided by the gcc-3.4 source package and not by > gcc-3.3, which makes packages depend on libgcc1 =~ 3.4 and so (as is > happening now for a bunch of packages) gcc-3.4 can block those > entering from sarge. > for the reasons for having libgcc1 provided by gcc-3.4 (which is not > the default compiler) and not by gcc-3.3. (which is), you should ask > someone like Matthias Klose, I guess. though I'd say that the change > has been made in all architectures for which the 3.3 -> 3.4 GCC > transition does not mean an ABI change (since libgcc1 =~ 3.4 is > probably "better" than =~ 3.3 and there is no reason against using it). gcc 3.3->3.4 should only be an ABI change for *C++* code; libgcc is implemented in C, and is therefore not affected by the ABI transition, making it reasonable to use the newer version of libgcc in sarge. > > Secondly, I'd like to learn what this libunwind is about and why > > tetex-bin is linked against it on some (many!) arches, but not on > > i386. The package description wasn't really helpful there. > nfi about this, sorry. This is not "many", or even "some"; there is precisely one architecture which includes libunwind7, ia64, where AIUI this library plays a role in being able to properly debug program call stacks. Why this library was suddenly deemed critical for the architecture after we were already 3 months into a toolchain freeze is another question. FWIW, these questions seem more appropriate to debian-devel than debian-mentors; these are not what I would call novice packaging questions. :) -- Steve Langasek postmodern programmer
Attachment:
signature.asc
Description: Digital signature