moved cvs into gcc-2.92.2, configure still broken

[Aaron Weiss <weiss at clearway dot com>]

Hi there all, I'm having a bit of an issue with the new configure stuff
and I thought it perhaps merited a post to the digest.
If it turns out I'm just mistaken, I apologize in advance. :)  

I just took a copy of the code from the cvs tree as of : Tue Feb 15
20:02:13 2000 UTC (110 minutes, 20 seconds ago) by bkoz.
I followed the directions in a recent post and put it into the gcc
source directory structure. then I ran configure and it told me this:
.
.
.
.
creating ./config.status
creating mkcheck
creating Makefile
creating src/Makefile
creating math/Makefile
creating libio/Makefile
creating config.h
config.h is unchanged
./config.status: ./.././../config-ml.in: No such file or directory
running mkc++config
cat: ./bits/c++config.h: input file is output file
running mknumeric_limits
In file included from ./src/gen-num-limits.cc:42:
/home/aweiss/gcc-2.95.2/libstdc++-v3/bits/c++config.h:566: unbalanced
`#endif'
gen-num-limits failed to build, exiting.

note where it's looking for config-ml.in ... 
my pwd is : /home/aweiss/gcc-2.95.2/libstdc++
and it's a full source dist. of gcc with config-ml.in the appropriate
place.

something else weird is that on repeated runs of configure (blowing away
my config.cache, poking around with stuff , etc ..) it seems to be
appending rather
than overwriting c++config.h so it gets longer with duplicate info
repeated on each run. If I run a configure right after that one fails,
the next time through I get:
.
.
.
In file included from ./src/gen-num-limits.cc:42:
/home/aweiss/gcc-2.95.2/libstdc++-v3/bits/c++config.h:566: unbalanced
`#endif'
/home/aweiss/gcc-2.95.2/libstdc++-v3/bits/c++config.h:1132: unbalanced
`#endif'
gen-num-limits failed to build, exiting.


the unbalanced #endif that started this was because c++config.h didn't
have an #ifdef for "#endif // _CPP_CPPCONFIG_" anywhere.

I am running red hat linux, gcc 2.95.2, on i686.
I have a built and installed version of the latest officialy snapshot
that works like a champ, but I grabbed the source code because of long
long and istringstream insertion and extraction fixes.... 

Am I doing something dumb, or is there a problem... anyone else seeing
this?

thanks!
-aaron