There's no real reason that lang/gcc should be marked as only available
for 2-3 platforms.

The in-tree copy of gcc 2.95.3 has been fully ported to all the major archs
and except for pc532 really should be workable on everything.

In any case I've made a new patchset based on the latest -current copy of
gcc and put a copy of this on

ftp.netbsd.org:/pub/NetBSD/misc/jmc/gcc.tar

It unpacks as lang/gcc/

I'd like folks to test this if possible and let me know if they encounter
any errors (I'd be surprised as it's built clean for me so far on x86, powerpc
and sparc64 and this build mimics what tools/toolchain does for its
bootstrap buid).

After that I had a few questions since the cvs logs are deficient in stating
why certain things are as they are:


1. A few (10 or so) packages for SunOS require lang/egcs. Can anyone think of
   a reason that shouldn't be turned into lang/gcc?

2. Why was there a special PLIST.NetBSD-sparc? My sparc box is currently 
   offline but I know 2.95.3 works on it. This is definitly an arch if someone
   has a box available testing this would be nice.

3. Look in Makefile.gcc and let me know what people think. Basically the idea
   is anything which relies on lang/gcc is doing this because of features
   required. So the test for egcs vs. 2.95.3 is in here and only if the 
   installed gcc isn't 2.95.3 will this package get pulled in as a build 
   dependency. There's a var to force it anyways if need be.

(What prompted all of this was powerpc not having lang/gcc support and there's
30 some odd packages that don't build in my bulk builds then)

James