Re: Reporting the Microkernel in the GNU triple?

[Roland McGrath]

We're agreed that we don't want a distinct tuple for most purposes.  I
suspect we don't need one even for the few special cases that distinguish.
The Hurd itself will always be the primary special case that is
substantially different depending on the microkernel.  It is also the
terminus of the dependency graph, i.e. everything depends on libc and libc
depends on it.  (Actually the end of the graph is circular, but you have to
start with hurd before you can do libc and then you can do hurd and
everything else without circularity.)  I think we can make everything else
key off that, which is to say the Hurd headers and libraries installed can
indicate everything more specific than "we be GNU" that anything else needs
to know.

As to what we do in libc and in gdb and each other case with separate
builds that should be packaged separately, that discussion should start up
on debian-hurd.