[bugs #10950] GSFormat with %C, regression between 1.3.2 and 1.10.0

[Adam Fedor]

This mail is an automated notification from the bugs tracker
 of the project: GNUstep.

/**************************************************************************/
[bugs #10950] Latest Modifications:

Changes by: 
                Adam Fedor <fedor@gnu.org>
'Date: 
                Fri 12/03/2004 at 03:20 (US/Mountain)

            What     | Removed                   | Added
---------------------------------------------------------------------------
         Assigned to | None                      | fedor
              Status | Open                      | Analyzed


------------------ Additional Follow-up Comments ----------------------------
It appears that on MingW, wint_t is a short, which is technically allowed. 
However, it would be promoted to int in va_args. This patch fixes the problem:

Index: Source/GSFormat.m
===================================================================
RCS file: /cvsroot/gnustep/gnustep/core/base/Source/GSFormat.m,v
retrieving revision 1.32
diff -u -r1.32 GSFormat.m
--- Source/GSFormat.m   1 Nov 2004 08:25:04 -0000       1.32
+++ Source/GSFormat.m   3 Dec 2004 03:15:51 -0000
@@ -993,7 +993,7 @@
          break
 
        T (PA_CHAR, pa_char, int); /* Promoted.  */
-       T (PA_WCHAR, pa_wchar, wint_t);
+       T (PA_WCHAR, pa_wchar, int); /* Sometimes promoted.  */
        T (PA_INT|PA_FLAG_SHORT, pa_short_int, int); /* Promoted.  */
        T (PA_INT, pa_int, int);
        T (PA_INT|PA_FLAG_LONG, pa_long_int, long int);

Although I wonder if I should be more robust and check to see if wint_t could 
possibly be larger than an int. I'm also not sure why it worked before, but 
doesn't work now. Perhaps a change in MingW?






/**************************************************************************/
[bugs #10950] Full Item Snapshot:

URL: <http://savannah.gnu.org/bugs/?func=detailitem&item_id=10950>
Project: GNUstep
Submitted by: 0
On: Tue 11/09/2004 at 10:24

Category:  Base/Foundation
Severity:  5 - Average
Item Group:  Change Request
Resolution:  None
Privacy:  Public
Assigned to:  fedor
Status:  Analyzed


Summary:  GSFormat with %C, regression between 1.3.2 and 1.10.0

Original Submission:  Hi all,

GSFormat causes a crash when used with format "%C".

This is something that used to work with GNUstep 1.3.2 and crashes with GNUstep 
1.10.0 using MinGW. It works on OSX.

Using this code (appends an unichar 'A' to a mutable string):


On Mac OS X:

cougar:/tmp fred$ cat AppendUnichar.m
#import <Foundation/Foundation.h>

int main()
{
    NSAutoreleasePool *thePool = [[NSAutoreleasePool alloc] init];
    NSMutableString *theString = [NSMutableString string];

    [theString appendFormat:@"%C", (unichar)0x41];

    NSLog( @"[%@]", theString );

    [thePool release];

    return 0;
}

cougar:/tmp fred$ cc -o AppendUnichar -framework Cocoa AppendUnichar.m
cougar:/tmp fred$ ./AppendUnichar
2004-10-26 18:37:35.982 AppendUnichar[766] [A]

cougar:/tmp fred$

On GNUstep 1.3.2 on mingw, with gcc 3.2:

fred@FLEA /c/tmp/AppendUnichar
$ shared_obj/ix86/mingw32/gnu-gnu-gnu/AppendUnichar.exe
2004-10-26 18:44:03.988 AppendUnichar.exe[2020] [A]

fred@FLEA /c/tmp/AppendUnichar




On GNUstep 1.10.0:

AppendUnichar.e caused an Access Violation at location 6de58669 in
module gnustep-base_d.dll Reading from location ffffffff.

Registers:
eax=0022e7e0 ebx=00000000 ecx=0022e7e0 edx=00000002 esi=00000108
edi=0022e060
eip=6de58669 esp=0022e7d0 ebp=0022f718 iopl=0         nv up ei pl zr na
po nc
cs=001b  ss=0023  ds=0023  es=0023  fs=0038  gs=0000
efl=00000246

Call stack:
6DE58669  gnustep-base_d.dll:6DE58669  GSFormat  GSFormat.m:985
    ...
    
    T (PA_CHAR, pa_char, int); /* Promoted.  */
>    T (PA_WCHAR, pa_wchar, wint_t);
    T (PA_INT|PA_FLAG_SHORT, pa_short_int, int); /* Promoted.  */
    T (PA_INT, pa_int, int);
    ...

6DE69D71  gnustep-base_d.dll:6DE69D71  _i_GSMutableString__appendFormat_
 GSString.m:2803
    ...
    #endif
        }
>      GSFormat((GSStr)self, fmt, ap, nil);
      _flags.hash = 0;// Invalidate the hash for this string.
      if (fmt != buf)
    ...

004013AE  AppendUnichar.exe:004013AE  main  AppendUnichar.m:8
    ...
    NSMutableString *theString = [NSMutableString string];
    
>    [theString appendFormat:@"%C", (unichar)0x41];
    
    NSLog( @"[%@]", theString );
    ...

0040122D  AppendUnichar.exe:0040122D
00401258  AppendUnichar.exe:00401258
7C581AF6  KERNEL32.dll:7C581AF6  GetVolumeInformationA


Of course, there are plenty of workarounds, but I definitely don't understand 
what is going on there in GSFormat.m

Thanks for all the great work you all do on GNUstep,

--andre


Follow-up Comments
------------------


-------------------------------------------------------
Date: Fri 12/03/2004 at 03:19       By: Adam Fedor <fedor>
It appears that on MingW, wint_t is a short, which is technically allowed. 
However, it would be promoted to int in va_args. This patch fixes the problem:

Index: Source/GSFormat.m
===================================================================
RCS file: /cvsroot/gnustep/gnustep/core/base/Source/GSFormat.m,v
retrieving revision 1.32
diff -u -r1.32 GSFormat.m
--- Source/GSFormat.m   1 Nov 2004 08:25:04 -0000       1.32
+++ Source/GSFormat.m   3 Dec 2004 03:15:51 -0000
@@ -993,7 +993,7 @@
          break
 
        T (PA_CHAR, pa_char, int); /* Promoted.  */
-       T (PA_WCHAR, pa_wchar, wint_t);
+       T (PA_WCHAR, pa_wchar, int); /* Sometimes promoted.  */
        T (PA_INT|PA_FLAG_SHORT, pa_short_int, int); /* Promoted.  */
        T (PA_INT, pa_int, int);
        T (PA_INT|PA_FLAG_LONG, pa_long_int, long int);

Although I wonder if I should be more robust and check to see if wint_t could 
possibly be larger than an int. I'm also not sure why it worked before, but 
doesn't work now. Perhaps a change in MingW?

-------------------------------------------------------
Date: Tue 11/23/2004 at 00:03       By: Fred Kiefer <FredKiefer>
Corrected Category.

-------------------------------------------------------
Date: Tue 11/09/2004 at 10:48       By: 0 <None>
wrong group. how do i change this to 'base'?












For detailed info, follow this link:
<http://savannah.gnu.org/bugs/?func=detailitem&item_id=10950>

_______________________________________________
  Message sent via/by Savannah
  http://savannah.gnu.org/