-B${prefix} during bootstrap

[Phil Edwards <pedwards at disaster dot jaj dot com>]

Something that's bothered me for years that I'm just now asking...

Why is -B$prefix passed to the stage1 compiler when building the stage2
compiler?  If $prefix isn't there yet (normal when bootstrapping, I would
think), then what's the point?

When bootstrapping with a previous gcc installed in the same $prefix, I
once ended up with a permanently buggy toolchain.  Some back-end binary
was being picked up because of this, and propogating its bugs into the
newly-bootstrapped compiler.  Installing the new compiler over the old
one didn't fix the bug (the goal of the exercise).  Lather, rinse, repeat.
(Thank goodness for prebuilt binaries.)

It was long enough ago that I don't remember details, and I didn't understand
the nature of the bug.  Once the problem with -B was pointed out to me
(some kind soul on a mailing list), I started using different $prefix'es
for every build, but always wondered what the rationale for this was.


Phil

-- 
pedwards at disaster dot jaj dot com  |  pme at sources dot redhat dot com
devphil at several other less interesting addresses in various dot domains
The gods do not protect fools.  Fools are protected by more capable fools.