c++/1687: Extreme compile time regression from 2.95.2

[Steven Bosscher]

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=1687

I cannot reproduce this with:
- gcc-3.3 (GCC) 3.3 20030407 (prerelease) or
- gcc-3.4 (GCC) 3.4 20030411 (experimental).

Both give subsecond compile times.  Which is particularly nice
because it means that we can say at least one positive thing about
compile time performance of 3.3 (this is a regression that is not
marked as one for some reason...).

Wolfgang, you were the last to re-confirm the PR, do you still
see it?

Greetz
Steven