egcs-980321 alphaev5-dec-osf4.0b bootstrap failure ...

[Kaveh R. Ghazi]

 > From: Jim Wilson <wilson@cygnus.com>
 > 
 > I believe I found the problem.
 > 
 > When Richard rewrote the GEN_INT macro, he dropped the HOST_WIDE_INT cast.
 > Unfortunately, this causes us to get different results depending on whether
 > there is a prototype for gen_rtx_CONST_INT in scope.  The cc built stage1
 > has no prototype, the gcc built stage2 has a prototype, and so they generate
 > different code.  This is not good.
 > 
 > I am trying a rebuild with the following patch, but the alpha machine I am
 > doing it on is so slow that I won't know the result until tomorrow.
 > 
 > Wed Apr  1 14:07:52 1998  Jim Wilson  <wilson@cygnus.com>
 > 
 > 	* rtl.h (GEN_INT): Readd HOST_WIDE_INT cast lost Jan 8.
 > 

	When I apply the above patch against the older egcs-980321, it
solves the "make compare" failures when using cc for stage1.  Thanks for
the excellent sleuthing!

	However the more recent egcs-980328 crashes during bootstrap on
OSF4.  This happens regardless of whether I use cc or gcc for stage1. 
Whether I do or don't apply the rtl.h or your recent alpha.md patch
doesn't affect this either.  All crash the same. 

 > ./xgcc -B./  -DIN_GCC    -W -Wall -O2 -g -I./include  -I. -I. -I./config \
 > -c ./objc/hash.c -o objc/hash.o
 > xgcc: Internal compiler error: program cc1 got fatal signal 6
 > make[2]: *** [objc/hash.o] Error 1


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