Paul Eggert writes:
The basic idea
here is that coreutils will attempt to detect the problem, and pass
the -qlanglvl=extc89 option to IBM's compiler. (It will pass
-qlanglvl=ansi to older versions of IBM's compilers.) I assume this
will work around the problem; if not, please let me know.
Yes, -qlanglvl=extc89 fixes it so we are good to go. Thanks.
-Original Message-
From: Paul Eggert [mailto:[EMAIL PROTECTED]
Sent: Monday, November 21, 2005 2:49 PM
To: Lemley James - jlemle
Cc: bug-coreutils@gnu.org
Subject: Re: FW: PMR 52061,370,000
Lemley James - jlemle [EMAIL
IBM's test case is fine, but it's not related to the bug. The
bug, as I understand it, is exhibited when you use bool. IBM's
test case does not involve bool.
Well, no. I've discovered that bool isn't really involved; it's an
unsigned problem when dealing with 64-bit math. The fact that we
Lemley James - jlemle [EMAIL PROTECTED] writes:
--- begin IBM note ---
There is nothing to fix in cc. This compiler invocation is intended to
use an older language level, which uses -qupconv.
I understand why IBM cc emulates older compilers when promoting
unsigned short. That is because
From: Paul Eggert [mailto:[EMAIL PROTECTED]
Sent: Friday, October 07, 2005 2:00 PM
Also, could you please report the xlc bug to IBM?
After going back and forth with IBM some, it appears they don't think
there is problem with their compiler, even though with the default
options and
Lemley James - jlemle [EMAIL PROTECTED] writes:
--- begin IBM note ---
Hi James,
This is not a defect. I wrote a new testcase that exhibits the same
behaviour as the one that was submitted.
IBM's test case is fine, but it's not related to the bug. The
bug, as I understand it, is
How about if we install the following coreutils patch?
It should catch the more-general problem, if I understand it aright.
My understanding is that the following little C program:
#include stdbool.h
#include assert.h
int main (void) {
assert (-1 - (_Bool) 0 0);
assert (-1 -
Paul Eggert [EMAIL PROTECTED] wrote:
How about if we install the following coreutils patch?
It should catch the more-general problem, if I understand it aright.
My understanding is that the following little C program:
#include stdbool.h
#include assert.h
int main (void) {
Paul, Jim and other interested parties:
The problem isn't with bool or _Bool; the problem reproduces with char
and unsigned char as well. Moreover, the assert won't catch it.
Looks like a rather severe compiler bug to me.
Example program output:
-2 + b is -1
-2 + b is -1
-2 + b is -1
d[-2
So, the problem occurs only in a subscript context? I really would
like to catch it if possible.
Near as I can tell, that's right. At least printf prints the right
thing.
Does the following program catch it? I'll looking for a way to detect
that compiler bug at compile-time.
Just what
Lemley James - jlemle [EMAIL PROTECTED] writes:
this detects it:
#include stdbool.h
#include stdio.h
char *c = NULL;
int main()
{ return (c + 5)[-2 + (bool) 1] == c[4] ? 0 : 1; }
Thanks. We might use that if that's the best we can do, but I want
something that works at
Oh, I get it. I didn't realize you wanted fail at compile time. Hm...
No, that fails to compile in both 64-bit and 32-bit mode, as well as my
old development Linux box (gcc 2.96 on Redhat 7.2)
AIX says:
eggert.c, line 3.12: 1506-221 (S) Initializer must be a valid constant
expression.
GCC
It dumps core as before.
-Original Message-
...
Would you please see if it works with this syntax?
bool swap = !!(0 compare (lines[-1], lines[-2]));
...
**
The information contained in this communication is
Jim Meyering [EMAIL PROTECTED] wrote:
Lemley James - jlemle [EMAIL PROTECTED] wrote:
when compiling coreutils 5.90 on AIX 5.3 in 64 bit mode
(OBJECT_MODE=64),
sort dumps core at line 1724 or 1725 (coreutils 5.90).
It works fine in 32-bit mode, and debugging statements I add make it
seem
Lemley James - jlemle [EMAIL PROTECTED] writes:
When this change is made, it runs just fine (well, it works for at least
one input) after being compiled in 64-bit mode:
1723c1723
bool swap = (0 compare (lines[-1], lines[-2]));
---
int swap = (0 compare (lines[-1],
15 matches
Mail list logo
