This is the mail archive of the
gcc-prs@gcc.gnu.org
mailing list for the GCC project.
other/10182: wrong conversion [unsigned -> double] with [-masm=intel]
- From: markus dot mauhart at chello dot at
- To: gcc-gnats at gcc dot gnu dot org
- Date: 21 Mar 2003 13:20:16 -0000
- Subject: other/10182: wrong conversion [unsigned -> double] with [-masm=intel]
- Reply-to: markus dot mauhart at chello dot at
>Number: 10182
>Category: other
>Synopsis: wrong conversion [unsigned -> double] with [-masm=intel]
>Confidential: no
>Severity: critical
>Priority: medium
>Responsible: unassigned
>State: open
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Fri Mar 21 13:26:00 UTC 2003
>Closed-Date:
>Last-Modified:
>Originator: markus dot mauhart at chello dot at
>Release: gcc version 3.2 20020927 (prerelease)
>Organization:
>Environment:
cygwin under winXP, AMD Duron
>Description:
When using option "-masm=intel", then all conversions from unsigned u (with u>=0x8000 (32768)) to double fail. One effect then is that all conversions from "[U]LONGLONG" to double fail for x>0x8000.
This is with C and C++ code.
>How-To-Repeat:
Compile 123.c like ...
gcc -v -save-temps -masm=intel -o 123.exe -g 123.c
... and then run 123.exe: it then displays correct conversions (all signed -> double) and incorrect conversions (all unsigned>=32768)
//file 123.c
#include <stdio.h>
int main (int x,char const* const y[])
{
unsigned int u = 0x4000 ;
for ( ;u ;u*=2)
{
int i ;
double du ;
double di ;
i = --u , du = u , di = i ;
printf ("u=% 8x, double(u)=% 14.1f, i=int(u)=% 11i, double(i)=% 14.1f\n"
,u ,du ,i ,di) ;
i = ++u , du = u , di = i ;
printf ("u=% 8x, double(u)=% 14.1f, i=int(u)=% 11i, double(i)=% 14.1f\n"
,u ,du ,i ,di) ;
}
return 0 ;
}
>Fix:
AFAIK the only workaround is to remove the option [-masm=intel]
>Release-Note:
>Audit-Trail:
>Unformatted: