Why not gnat Ada in gcc?

[Robert Dewar]

<<And that's precisely the procedure you would use if you did a merge from
the external (potentially unstable) tree into your internal tree.  Those
tests will likely find problems, you would fix them and submit the fixes
to the net sources.
>>

But it is VERY rare for regressions to occur on nightly runs, I would
say that 90% of the time the nightly builds are completely clean and
run the entire regression suites completely. I cannot believe that
this is true of the current gcc open tree (unless the test suites
are very weak indeed). 

<<Right.  THe problem is strictly your problem as a business, it is not a
problem for the GNU project.  Thus, you should confine that problem to your
own internal sources and not force it on the official GNU sources.
>>

No one is forcing anything on anyone! Indeed if anything it seems like
Jeff is trying to force some unworkable model here, although it is a little
hard to follow, since in practice, I think the situation with GNAT will
be quite similar to that of gcc, avoiding hopefully the phenomena of
out-of-the-blue major things like the ia32 port.

<<I agree, and it is a problem.
>>

(it = performance problems)

Yes, and it is a problem that absolutely has to be dealt with in 
our environment. As I said, I am quite surprised that this is not
a problem for Redhat and Cygnus (I know it is a significant problem
for other companies trying to make more use of GCC).