Re: About a mass bug report not based on Sid or Jessie.
Ian Jackson <ijackson@chiark.greenend.org.uk> writes:
> Simon McVittie writes:
>> AIUI, it's for build systems that make decisions based on the
>> "canonical" GNU architecture name of the build architecture (or,
>> indirectly, the host architecture, since the default host architecture
>> is "whatever the build architecture is"). For instance, "if this is
>> Linux, do foo, else do bar"
> Ah. I never do that. I actually test for the presence of the
> features I care about.
That's the gold standard, but it's not always possible. For example, see
the code in libpam-afs-session's kafs directory and the corresponding
Autoconf macros in m4/kafs.m4. I have to select between available kernel
interfaces for creating AFS PAGs, and it's very difficult to probe,
particularly given that the package may be built on a system that doesn't
have AFS loaded.
That said, that's one of the few packages for which I use
AC_CANONICAL_HOST directly. Usually it's pulled in by Libtool, which
appears to use it extensively. (Libtool is a classic example of the worst
possible tool for the job, except for all of the other available tools.)
--
Russ Allbery (rra@debian.org) <http://www.eyrie.org/~eagle/>
Reply to: