Having spent the better part of two days tracking down this problem I thought I
add to the collective painful lesson database.
ALWAYS turn on the Codewarrior "Require Function Prototypes". Without this
Codewarrior is going to generate screwed up code without
any warnings at all! I spent hours and hours trying to track down this weird
bug which only happened on OS versions 3.0 - 3.3 and
not on any others. When I finally found the problem, I realized that it was
all due to the "Require Function Prototypes" having
been turned off.
This is how. In one source file I had a routine, CreateRecord(DBRecord *)
which was successfully getting called from other
routines. In another source file I also called this CreateRecord(DBRecord *).
BUT, when I was moving routines around into more
distinct source files I didn't add a global prototype for CreateRecord().
Thus, in the module that didn't have the prototype it was silently treating the
CreateRecord() routine as CreateRecord(int) instead
of the CreateRecord(long). Then when the CreateRecord() was being called it
wasn't getting the correct parameter off the stack.
The REALLY annoying thing is that is was always getting a 1, instead of the 0
that was trying to be passed. My hair is finding
other locales fast enough without my help pulling it out.
So, that's it. I hope this helps somebody out someday.
Rick Wagner
Mindweave Software, Inc.
PS: Rereading this someone might think that I'm blaming codewarrior. Not at
all! It was definitely a user error (most likely
mine).
Attached below is the output. I'm including this so that it's searchable.
Here is the mixed assembler/C source from the debugger:
err = CreateRecord(NULL);
00048C8A: 42A7 clr.l -(a7)
00048C8C: 4EBAF37C jsr *-3202 ; 0x0004800a
Here is the another snippet from another area of the program:
case NewButton:
if(CreateRecord(NULL) == 0) { //create new empty loan
00047190: 4267 clr.w -(a7)
00047192: 4EBA0EB6 jsr *+3768 ; 0x0004804a
Note that on one line this is a clr.l, and on another line it is clr.w.
(these were from two different compiles so the jsr address is different).
--
For information on using the ACCESS Developer Forums, or to unsubscribe, please
see http://www.access-company.com/developers/forums/
