Re: ada/6669: Irix6.5 ada bootstrap fail in ada/targparm.adb

[Rainer Orth <ro@TechFak.Uni-Bielefeld.DE>]

The following reply was made to PR ada/6669; it has been noted by GNATS.

From: Rainer Orth <ro@TechFak.Uni-Bielefeld.DE>
To: David.Billinghurst@riotinto.com
Cc: gcc-gnats@gcc.gnu.org
Subject: Re: ada/6669: Irix6.5 ada bootstrap fail in ada/targparm.adb
Date: Thu, 16 May 2002 04:29:55 +0200 (MEST)

 David.Billinghurst@riotinto.com writes:
 
 > gcc -c -g  -W -Wall -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes -Wtraditional       -gnatpg -gnata -I- -I. -I/exd4/billingd/src/gcc/gcc/ada /exd4/billingd/src/gcc/gcc/ada/targparm.adb
 > as: Error: /var/tmp/cca4KkNE.s, line 5416: value must be -128..255
 >       .byte 0x80000000
 > as: Error: /var/tmp/cca4KkNE.s, line 5417: value must be -128..255
 >       .byte 0x80000003
 > as: Error: /var/tmp/cca4KkNE.s, line 5536: value must be -128..255
 >       .byte 0x80000000
 > as: Error: /var/tmp/cca4KkNE.s, line 5537: value must be -128..255
 >       .byte 0x80000014
 > as: Error: /var/tmp/cca4KkNE.s, line 5656: value must be -128..255
 >       .byte 0x80000000
 > and so on
 > 
 > The bootstrap compiler was a mips-sgi-irix6.2 build of 
 > gnat-3.13p supplied by rainer orth  - see PR ada/6552.
 
 This only happens on the mainline, the 3.1 branch is fine.  Here's what's
 going on: targparm.adb is compiled using the .adb.o suffix rule, which
 uses $(ALL_ADAFLAGS), among others.  This variable differs between 3.1 and
 3.2: 
 
 The 3.1 gcc/ada/Makefile has
 
 ALL_ADAFLAGS = $(ADA_CFLAGS) $(X_ADAFLAGS) $(T_ADAFLAGS) $(ADAFLAGS)
 
 ADA_CFLAGS = 
 
 whereas on the mainline this is
 
 ALL_ADAFLAGS = $(CFLAGS) $(ALL_ADA_CFLAGS) $(X_ADAFLAGS) $(T_ADAFLAGS) \
 	$(ADAFLAGS)
 
 CFLAGS = -g
 ALL_ADA_CFLAGS = $(X_ADA_CFLAGS) $(T_ADA_CFLAGS) $(ADA_CFLAGS)
 
 I.e. the mainline adds -g, which breaks assembly.  This is obviously a bug
 in the bootstrap gnat.  Unfortunately, due to PR ada/6552 one cannot
 currently build a complete 3.1 based gnat, so removing -g on the mainline
 (probably stage1 only) is the only option until this is fixed.
 
 	Rainer