Re: [Ada] Bootstrapping mainline GNAT fails

[Mark Mitchell <mark at codesourcery dot com>]

I do not think we need to worry too much about Ada in the GCC 3.1 timeframe.

I think it should be a goal that GNAT in GCC 3.2 be built with the GNAT
in GCC 3.1.  The GNAT in GCC 3.1 will be the first official FSF release
of GNAT, and we should, like any organization, look after our users.

If GNAT 3.1 is a bit hard to build, I am not too worried.  There are two
important points:

1. How to build it should be clearly documented.  This is important so
   that system distributors can do it, and package it up.  Then, as
   time goes by, more and more people will have easy-to-install packages
   of GNAT 3.1.

2. If the GNAT used to bootstrap the compiler is not one that will work,
   and we can detect that, it would be nice if we would say "you can't
   build this GNAT with that GNAT" and stop at that point.  That will
   help people building from source know what's what.  It will also
   keep people who have some random version of GNAT around from being
   unable to build the C, C++, Java, Fortran, and Objective-C compilers
   because their build dies part way through due to a GNAT problem.

   The bottom line is that, without meaning any criticism of the language,
   there are not likely to be nearly as many Ada users as C users among
   our building-from-source users, and we should not make life difficult
   for the C users.  So, GNAT should do like Java and disable building
   itself when it is an environment where it is unlikely to succeed.

It would be nice, of course, if you could build GNAT without an installed
GNAT at all, but, then again, it would be nice if you could build all
software with nothing but a bourne shell.

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com