Re: Solaris bootstrap failure

["Kaveh R. Ghazi" <ghazi at caip dot rutgers dot edu>]

 > From: Mark Mitchell <mark@codesourcery.com>
 > 
 >   I also ran 2001-02-08 & 09 this morning to narrow it down and got:
 > 
 >   2001-02-08: pass
 >   2001-02-09: fail (cp/tree.o differs)
 > 
 > Thanks!  (I was out of town yesterday; sorry for the slow reply.)
 > 
 > I'm not sure what to make of this.  As you say, it's disappointing
 > that it succeeds again later, as that makes it hard to blame a
 > particular change.  Still, I will stare at those diffs as closely as I
 > can.  Also, do you find that the difference is reproducable using the
 > 2001-02-09 build?  That's what we really need is a case where the
 > difference can be reproduced when running outside of the `make
 > bootstrap' process.

Sadly, I cannot reproduce the diffs when running by hand outside the
bootstrap process with 2-09 or 2-10.  The bootstrap still fails make
compare.  But both hand-run stages created a cp/tree.o which
consistently matched the stage2 bootstrap cp/tree.o.  Strangely, when
I run the same invocations from another solaris box which NFS mounts
my build directory, the hand-run invocations produce a third different
cp/tree.o but are consistent with each other. :-(

(BTW, I'm using cc for stage1 and native as/ld, in case that matters.)

I also tried bootstrapping both checkouts with -g -O2 --save-temps to
see if it would leave behind some clues, but that one passed "make
compare".

Any other ideas?

		--Kaveh
--
Kaveh R. Ghazi			Engagement Manager / Project Services
ghazi@caip.rutgers.edu		Qwest Internet Solutions