CVS head and 3.0 branch, bootstrap with 2.96-rh?

[Gordon Sadler <gbsadler1 at lcisp dot com>]

First, I understand 2.96-rh is 'unofficial' as far as gcc is concerned.
I'm not asking this be considered a bug, I'm interested in the build
process. Does it perhaps look for 2.95.x and/or 3.x, and cause undefined
behaviour for all other values of a gcc version used in the bootstrap?

I have a few months experience bootstrapping gcc from cvs virtually
every day at home. Then going through some of the logs to try to
identify daily differences/changes/regressions. I think I have a decent
understanding of using make bootstrap on a ix86 host native build.

At work, for the time being, I have linux computer running Redhat 7.1
with the 2.96-rh (latest version -85?) compiler as the default gcc/g++/etc..
As it is a pentium IV with 256 MB RAM and considerable spare cycles, I've
tried to get it to do daily bootstraps as well. Now I've noticed that using
this compiler to bootstrap never seems to produce stages.

(ie.
$grep stage1 build.log

$grep stage2 build.log

)

However all of my bootstraps at home using 2.95.x or [pre]3.0 to
bootstrap cvs head or 3.0 branch AFAICS always produces stages where the
result is compared after stage1 makes stage2.

I have also installed Redhat's kgcc on the system, so I have another
compiler available for the bootstrap. I guess I'm curious as to why make
bootstrap seems to behave so differently with 2.96-rh?

-- 
Gordon Sadler