Re: Depencencies of gcc on glibc/eglibc and libc6-i386
Ludovic Brenta a écrit :
> Matthias Klose <doko@debian.org> writes:
>> On 29.08.2009 18:38, Ludovic Brenta wrote:
>>> I would like to better understand the depencencies between the various
>>> gcc packages and libc6{,-i386}, in particular as they relate to the
>>> transition to /lib32.
>>>
>>> In debian/rules.conf we have:
>>>
>>> libc_ver := 2.5
>>> ifneq (,$(findstring gnat,$(PKGSOURCE)))
>>> libc_ver := 2.9-21
>>> endif
>> [...]
>>> How about unifying all the version numbers to 2.9-22?
>> I don't see why the b-d is necessary for the gcc build; it's an
>> upgrade scenario only, and the buildd's install multilib related
>> packages from scratch for each build. I'd like to keep it at 2.5, so
>> that people can build the package on older releases as well.
>>
>> I think that Aurelian did argue having the conflicts in the binary
>> packages is enough.
>
> But in a machine with glibc 2.5 and libc6-i386 2.5 that builds gcc
> produces binary packages that are uninstallable on the same machine
> because they conflict with glibc (<< 2.9-22). Is that intentional? To
> me it seems to break the Law of Least Astonishment; if I build on a
> machine, the least I can reasonably expect is to be able to install on
> that same machine (unless of course I'm cross-compiling, which is not
> our case).
This is due to the lib32 transition on amd64. Current version of gcc put
the 32-bit version of the library in /lib32, while previous versions put
them into /emul/ia32-linux. Therefore while the current gcc will build
against glibc 2.5, the resulting lib32* packages won't be installable on
amd64. Other packages will be installable. Note that it does not concern
the other architectures.
--
Aurelien Jarno GPG: 1024D/F1BCDB73
aurelien@aurel32.net http://www.aurel32.net
Reply to: