Message-Id: <199903061623.LAA10573@delorie.com>
Mailing-List: contact cygwin-help AT sourceware DOT cygnus DOT com; run by ezmlm
Sender: cygwin-owner AT sourceware DOT cygnus DOT com
Delivered-To: mailing list cygwin AT sourceware DOT cygnus DOT com
From: "Paul Coward" <paul AT asgard DOT net DOT au>
Organization: DigiSoft Pty Ltd
To: cygwin AT sourceware DOT cygnus DOT com
Date: Sun, 7 Mar 1999 02:24:36 +1000
MIME-Version: 1.0
Content-type: text/plain; charset=US-ASCII
Content-transfer-encoding: 7BIT
Subject: Stack problem in Beta 20.1 ???
Reply-to: paul AT asgard DOT net DOT au
X-mailer: Pegasus Mail for Win32 (v3.01b)

Hi,

The following code fragment fails to compile correctly under 20.1 unless 
the setjmp/longjmp pair are included to fix an apparant stack problem.

This code is run inside a loop and prints the correct values the first time 
through but totally stuffs up on subsequent passes (unless the 
setjmp/longjmp are present).

As near as I can tell the code seems to get the stack alignment wrong 
after the first pass.  The fact the the setjmp/longjmp pair seem to have 
fixed the problem seems to support this assumption.

This is the first time that I have compiled a program with cygwin that 
passes so many args on the stack (9 in the case of 
PRN_SetPrintText()).  Could this be the cause of the problem?

The values passed don't matter.  Even the fprintf call stuffs up with too 
many parameters passed.

      if (setjmp(jmpBuf) == 0) {
#ifdef DEBUG_PRN
       fprintf(stderr, " Push constants\n");
       fprintf(stderr, "   (left) %ld\n   (top) %ld\n   (charType) %d\n   
(charWidth) 
%d\n   (charHeight) %d\n   (charRot) %d\n   (stringRot) %d\n   
(charSpace) %d)\n",
                ((long) (((double) text->tfLeft) / TWIPS_PER_THERM)),
                ((long) (((double) text->tfTop) / TWIPS_PER_THERM)),
                1, 1, 1, 0, 0, 3);

       fprintf(stderr, " Push normal\n");
       fprintf(stderr, "   (left) %ld\n   (top) %ld\n   (charType) %ld\n   
(charWidth) 
%ld\n   (charHeight) %ld\n   (charRot) %ld\n   (stringRot) %ld\n   
(charSpace) 
%ld)\n",
                ((long) (((double) text->tfLeft) / TWIPS_PER_THERM)),
                ((long) (((double) text->tfTop) / TWIPS_PER_THERM)),
                charType, charWidth, charHeight, charRot, stringRot, 
charSpace);
       fprintf(stderr, "PRN_SetPrintText(\"%s\"\n\t(left) %d\n\t(top) 
%d\n\t(charType) 
%d\n\t(charWidth) %d\n\t(charHeight) %d\n\t(charRot) %d\n\t(stringRot) 
%d\n\t(charSpace) %d)\n",
                text->tfText,
                ((long) (((double) text->tfLeft) / TWIPS_PER_THERM)),
                ((long) (((double) text->tfTop) / TWIPS_PER_THERM)),
                charType, charWidth, charHeight, charRot, stringRot, 
charSpace);
#endif
       result = PRN_SetPrintText(text->tfText,
               ((long) (((double) text->tfLeft) / TWIPS_PER_THERM)),
               ((long) (((double) text->tfTop) / TWIPS_PER_THERM)),
               charType,
               charWidth,
               charHeight,
               charRot,
               stringRot,
               charSpace);
       if (result != PRN_SUCCESS) {
        LogError("Failed SetPrintText() call for \"%s\" (return code 
0x%08x)", 
text->tfText, result);
        Shutdown(1);
       }
       longjmp(jmpBuf, 1);
      }


Of course, it could be my fault.  I've spent 2 days trying to find a 
problem in my code.  It wasn't until I started to suspect the compiler 
and included setjmp that the problem disappeared.

If this is found to be a bug in cygwin could someone please mail me 
direct as I don't subscribe to this list.

Thanks
Paul Coward.



--
Want to unsubscribe from this list?
Send a message to cygwin-unsubscribe AT sourceware DOT cygnus DOT com