RE: 3.0 branch: wrong CFLAGS for gen*

["Billinghurst, David (CRTS)" <David dot Billinghurst at riotinto dot com>]

That gets over the original hurdle, but fails building g77

if [ -f lang-f77 ]; then \
  stage1/xgcc -Bstage1/ -B/exd4/billingd/tmp/gcc-3.0/mips-sgi-irix6.5/bin/
-c  -DIN_GCC    -g -O2 -W -Wall -Wwrite-strings -Wstrict-prototypes
-Wmissing-prototypes -Wtraditional -pedantic -Wno-long-long  -DHAVE_CONFIG_H
-I. -I. -I/exd4/billingd/src/gcc-3.0/gcc -I/exd4/billingd/src/gcc-3.0/gcc/.
-I/exd4/billingd/src/gcc-3.0/gcc/config
-I/exd4/billingd/src/gcc-3.0/gcc/../include \
	/exd4/billingd/src/gcc-3.0/gcc/f/g77spec.c -o g77spec.o; \
else true; fi
echo '#include "config.h"' > conftest.c
echo '#include "except.h"' >> conftest.c
echo 'xyzzy USING_SJLJ_EXCEPTIONS' >> conftest.c
cc -E conftest.c  -DIN_GCC    -g -O2 -W -Wall -Wwrite-strings
-Wstrict-prototypes -Wmissing-prototypes -Wtraditional -pedantic
-Wno-long-long  -DHAVE_CONFIG_H -I. -I. -I/exd4/billingd/src/gcc-3.0/gcc
-I/exd4/billingd/src/gcc-3.0/gcc/. -I/exd4/billingd/src/gcc-3.0/gcc/config
-I/exd4/billingd/src/gcc-3.0/gcc/../include > conftest.out
cc ERROR parsing -W:  bad syntax for option
cc WARNING:  phase key (-)  is no longer supported
cc ERROR parsing -W:  bad phase for -W option
cc WARNING:  phase key (W)  is no longer supported
cc ERROR parsing -W:  bad phase for -W option

The rule being used is 

cs-tconfig.h: Makefile $(CONFIG_H) except.h
	HEADERS="$(xm_file)" DEFINES="" TARGET_CPU_DEFAULT="" \
	$(SHELL) $(srcdir)/mkconfig.sh tconfig.h
# ??? Need extra help to get this definition copied to where libgcc can
# see it.  It shouldn't matter which compiler we use to get there.
	echo '#include "config.h"' > conftest.c
	echo '#include "except.h"' >> conftest.c
	echo 'xyzzy USING_SJLJ_EXCEPTIONS' >> conftest.c
	$(HOST_CC) -E conftest.c $(ALL_CFLAGS) $(INCLUDES) > conftest.out
	sed -e 's/xyzzy/#define USING_SJLJ_EXCEPTIONS/' -e '/#define/p' \
		-e d conftest.out >> tconfig.h
	rm conftest.*

This looks bogus to me as we should be using the newly built gcc, not
$(HOST_CC), to determine the exception model. 


-----Original Message-----
From: Loren James Rittle [mailto:rittle@latour.rsch.comm.mot.com]
Sent: Thursday, 4 October 2001 11:48 
To: robertlipe@usa.net
Cc: Billinghurst, David (CRTS)
Subject: Re: 3.0 branch: wrong CFLAGS for gen*


In article <20011003203917.B7813@rjlhome.caldera.com>,
Robert Lipe <robertlipe@usa.net> writes:

>> >This was broken sometime in the last week.  Without testing it, this
>> >change looks the most suspicious:

>> Yes, I have already accepted blame.  I will be checking in the
>> pre-approved patch after I finish testing it.

> OK, thanx.  Since this is the third bootstrap error in a row (and this
> on the stable branch!) I was probably unduly annoyed. :-)

I hope you are not as mad at me as I am with myself.  ;-)

Would you be so kind as to try this patch:

Index: Makefile.in
===================================================================
RCS file: /cvs/gcc/egcs/gcc/Makefile.in,v
retrieving revision 1.742
diff -c -r1.742 Makefile.in
*** Makefile.in	2001/10/02 07:12:19	1.742
--- Makefile.in	2001/10/04 01:40:03
***************
*** 2959,2964 ****
--- 2959,2967 ----
  # the context of the stage_x rule.
  STAGE2_FLAGS_TO_PASS = \
  	HOST_CC="$(HOST_CC)" \
+ 	HOST_CFLAGS="$(HOST_CFLAGS)" \
+ 	HOST_LDFLAGS="$(HOST_LDFLAGS)" \
+ 	HOST_CPPFLAGS="$(HOST_CPPFLAGS)" \
  	CFLAGS="$(BOOT_CFLAGS)" \
  	LDFLAGS="$(BOOT_LDFLAGS)" \
  	WARN_CFLAGS="\$$(GCC_WARN_CFLAGS)" \

Alexandre Oliva <aoliva@redhat.com> and I both think it might be
right.  It passes bootstrap for me just as well at what is on mainline
and 3.X branch but since I only have gcc as a bootstrap compiler, I
have no idea how to see the failure I caused you guys.

Regards,
Loren