Re: [Ada] Bootstrapping mainline GNAT fails

[dewar at gnat dot com (Robert Dewar)]

One more comment here.

I know that many people must be getting tired of this thread (one more
argument to go get yourself a thread reading mailer that can easily kill
threads :-)

The reason I am pushing on this issue is that I am trying to do everything
I can to make sure that the ACT and FSF trees stay as close as possible.
I think that's a very important goal for both the Ada community and the
GNU project.

Of course there are some places where we really need to be sure to fit into
the structure of GCC smoothly. We can't have GNAT being too much of a pain
in the neck for mainstream GCC development, or it won't be a welcome citizen,
and we definitely want it to be as welcome as possible.

For example, the recent discussion of version numbers in headers is a point
at which the requirements in the two trees are very different. We have to
have the version numbers in the headers at ACT, since it is an important
part of our over all configuration management system. But as we have
discussed, we clearly don't want them in the GCC tree. So we sat down and
figured out how to automate things so that this won't be a barrier that
makes synchronization of the tree (the solution, for those interested, is
that we remove the version headers from the tree, and have our checkout
procedures are ACT reinsert them at the point of checkout when needed).

But I really see the insistence on being able to bootstrap with arbitrarily
earlier versions of GNAT as one that could introduce problems that have no
feasible solution, and that's why I am devoting energy to makeing sure that
everyone understands the issues.

I find Laurent Guerby's comments significant. Laurent is one of the people
who have contributed significantly to GNAT in the past, and I hope will in
the future. That's the kind of voice that we really need to listen to.

Robert Dewar