[Ada] Bootstrapping mainline GNAT fails (with workaround and root cause: makefile bug)

[Russ Allbery]

Andreas Schwab <schwab@suse.de> writes:

> Since I'm always compiling this way, I'd say it should work.  There is
> no real diffence anyway between the locations of OBJDIR as long as it is
> different from SRCDIR.  The complications only come from the use of
> relative vs. absolute paths.

I would wager, in fact, that the relative vs. absolute paths is the real
problem and that whether or not OBJDIR is a subdirectory of SRCDIR is a
red herring.  I bet it would work or not work the same if you used a
relative path to an OBJDIR outside of SRCDIR as well.

In any event, gcc 2.95.3 built fine using an OBJDIR inside SRCDIR
specified with a relative path.  I don't really care if this changes -- I
can always change my build procedure and only do it that way because it's
convenient and means I can rm -rf the whole thing when I'm done -- but it
did used to work.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>