Daniel Stenberg writes:
>
> Ah. Redhat 7... If I'm just mistaking, that is more correct according to the
> Linux Hierarchy Standard, isn't it?

 Yes, according to FHS-2.1. Anyway, using proper names is better.

> I need to figure out a way to build release archives that work for me and
> still keeps the 'make dist' things work. I will probably just convert the
> script to make some fiddling first and then invoke 'make dist'. That should
> make all of us happy! ;-)

 That would be great, indeed. Updating EXTRA_DIST entries or FILES list is
a pain anyway, using a single source is less annoying.

 In the same spirit, make distcheck do make sure that sources build
properly from the generated tarbal and this is definitely usefull:
you're sure the tarbal shipped compiles. However, since the 'make
check' goal is disconnected from the 'make test' goal, the tests are
not run. Making this work would require some extra work (mainly using
$(srcdir) to find files in the tests) since the 'make distcheck' and
rpm building do the test when compiling outside the source tree. That
would bring an additional benefit : make distcheck would make sure
that 1) the distribution compiles + install ok 2) the tests all run
ok. And 'make distcheck' works on Solaris ;-) For my own packages I
recommend that contributions are only commited to the CVS tree when
'make distcheck' runs properly.

 Would you like this to work or do you see complications or is it for
some reason against your policy ?

 Cheers,

-- 
Loic   Dachary         http://www.dachary.org/  loic@dachary.org
24 av Secretan         http://www.senga.org/      loic@senga.org
75019    Paris         Tel: 33 1 42 45 09 16        loic_at_gnu.org