egcs-19980803 (pre-1.1) powerpc-ibm-aix4.1.4.0 regressions

[Kaveh R. Ghazi]

 > From: David Edelsohn <dje@watson.ibm.com>
 > 
 > Kaveh> BOOT_CFLAGS:             -g -O3 -funroll-all-loops -mminimal-toc
 > 
 > Kaveh> The following bootstrap error is a regression from prior
 > Kaveh> releases/snapshots.  (However, I am able to bootstrap when I lower
 > Kaveh> optimization to "-g -O2 -mminimal-toc".)
 > 
 >         You realize that this simply does not make sense.  I have been
 > bootstrapping without problem and no one else has reported a similar
 > problem.  Is there something corrupted on your system?  Nothing should
 > have changed between this snapshot and the previous one.
 > David


	I guess in pointing the finger back at me, you really did miss
the line saying that I was using extra BOOT_CFLAGS, even though my
message did state this.  I'm generally not an idiot, hopefully next
time your first reflex will not be to assume that my reports are bogus. :-)



 > From: David Edelsohn <dje@watson.ibm.com>
 > 
 > >>>>> "Kaveh R Ghazi" writes:
 > 
 > Kaveh> Just to be clear, you use "-g -O3 -funroll-all-loops -mminimal-toc"
 > Kaveh> and pass "make compare"?
 > 
 > 	First, you shouldn't be using -mminimal-toc as bootstrap is
 > suppose to use AIX -bbigtoc option.




	David, you yourself confirmed that there are bugs in AIX's
linker which make -mminimal-toc necessary even though -bbigtoc is
inserted automatically during bootstrap.  Please see the following
discussion:

http://www.cygnus.com/ml/egcs/1997-Sep/0632.html
http://www.cygnus.com/ml/egcs/1997-Sep/0636.html
	and especially where you join in:
http://www.cygnus.com/ml/egcs/1997-Sep/0648.html

	Its quite possible that since then you or someone else has
fixed gcc to work around this AIX linker bug, but I wasn't aware.  Can
you please comment on this?



 > 
 > 	Second, if you were building snapshots this way before, nothing
 > should have changed in this pre-release snapshot which would cause this
 > regression.
 > David


	Right, I already apologized to you for being unclear as to
when this bug first started happening.  It has been with us since at
least early July if not before.  I have a foggy recollection that it
may have started a couple of weeks earlier, perhaps around the time of
the last gcc2 merge, though I can't be sure.

		--Kaveh
--
Kaveh R. Ghazi			Project Manager / Custom Development
ghazi@caip.rutgers.edu		Icon CMT Corp.