Re: 19990405 DUX40d: Overflow

Re: 19990405 DUX40d: Overflow

[Jeffrey A Law]

  In message < 19990413102617.A5837@ween07.elektro.uni-wuppertal.de >you write:
  > 
  > --0F1p//8PRICkK4MW
  > Content-Type: text/plain; charset=us-ascii
  > Content-Transfer-Encoding: quoted-printable
  > 
  > Hi,
  > 
  > > uname -av
  > OSF1 ween07.elektro.uni-wuppertal.de V4.0 878 alpha alpha
  > 
  > egcs-19990405 produces this while compiling format.c in libchill:
  > 
  > root@ween07:/users1/usr_local/src/egcs-19990405/alphaev5-dec-osf4.0d/libchi
  > =
  > ll>=20
  >  make /users1/usr_local/src/egcs-19990405/gcc/xgcc -B/users1/usr_local/src/
  > =
  > egcs-19990405/gcc/
  >  -B/usr/local/alphaev5-dec-osf4.0d/bin/ -c -I. -I.   -O2  format.c
  > as0: Error: format.c, line 6: Overflow: 0x010000000000000000cccccccd
  >       .quad 0x10000000000000000cccccccd
  >       make: *** [format.o] Error 1
This has already been fixed and should work in the next snapshot.

Thanks,
jeff

19990405 DUX40d: Overflow

[Kurt Garloff]

Hi,

> uname -av
OSF1 ween07.elektro.uni-wuppertal.de V4.0 878 alpha alpha

egcs-19990405 produces this while compiling format.c in libchill:

root@ween07:/users1/usr_local/src/egcs-19990405/alphaev5-dec-osf4.0d/libchill> 
 make /users1/usr_local/src/egcs-19990405/gcc/xgcc -B/users1/usr_local/src/egcs-19990405/gcc/
 -B/usr/local/alphaev5-dec-osf4.0d/bin/ -c -I. -I.   -O2  format.c
as0: Error: format.c, line 6: Overflow: 0x010000000000000000cccccccd
      .quad 0x10000000000000000cccccccd
      make: *** [format.o] Error 1

Here is the offending assembler code:
.rdata
$LC12:
	.ascii "CHOBF\0"
$LC13:
	.ascii "X<>T\0"
$LC14:
	.ascii "/+-?!=\0"
	.align 3
$LC15:
	.quad 3435973837
	.align 3
$LC16:
	.quad 0x10000000000000000cccccccd
.text
	.align 5
	.ent scanformcont
scanformcont:
	.frame $30,96,$26,0
	.mask 0x400fe00,-96
	ldgp $29,0($27)
$scanformcont..ng:
	subq $30,96,$30
	stq $26,0($30)
	stq $9,8($30)
					 
Interestingly, when -g is added to the flags, the assembler won't complain and
no constant 0x10000... can be found in the assembly code.

With -O (instead of -O2), the bug remains the same, without optimization,
the bogus value can even be found twice in the assembly code.
It might be associated with the isXXX() macros, but I don know, cause my
alpha assembly knowledge very small.

Hope somebody can have a look at it.

-- 
-----BEGIN PGP SIGNATURE-----
Version: 2.6.3ia

iQCVAwUBNxL/qKsDVfmWq1wZAQEBUgP/b86Kx1xcz6xzEbTTS8tkgmUrSsmNo742
60vK6hmusMdOwN7SDC/Zl/8p0s/H6iLdvhdCmGSO5JlJ23/7ZiOnfJ5kdyVnWRML
NIaPMekdy7jw417OI6vo3n8ERAyxC7MDN7lvXs3ZzHgG0l8CnaZx9Jcl9Taqd8F5
XjD+s2UhBHM=
=sj0Z
-----END PGP SIGNATURE-----
>From law@upchuck.cygnus.com Fri Apr 30 23:15:00 1999
From: Jeffrey A Law <law@upchuck.cygnus.com>
To: brendan@zen.org (Brendan Kehoe)
Cc: egcs-bugs@egcs.cygnus.com
Subject: Re: crash compiling mozilla using egcs CVS tree of 9/9/1998 
Date: Fri, 30 Apr 1999 23:15:00 -0000
Message-id: <12918.924072883@upchuck>
References: <13815.2719.807435.101262@galway.zen.org>
X-SW-Source: 1999-04n/msg00393.html
Content-length: 549

  In message <13815.2719.807435.101262@galway.zen.org>you write:
  > The code below, when compiled with -fPIC, will fail with
  > 
  > crash1a.c: In function `XmlInitUnknownEncoding':
  > crash1a.c:12: internal error--unrecognizable insn:
  > (insn 28 26 30 (set (reg:SI 27)
  >         (const (unspec[ 
  >                     (symbol_ref/v:SI ("latin1_encoding"))
  >                 ]  7))) -1 (nil)
  >     (nil))
  > ../../egcs/gcc/toplev.c:1365: Internal compiler error in function fatal_insn
This (very old) bug was fixed a while ago.

jeff