[sdcc-devel] [ sdcc-Bugs-1722392 ] INT plus CHAR makes wrong assembly codes

[SourceForge.net <no...@so...>]

Bugs item #1722392, was opened at 2007-05-21 01:33
Message generated for change (Comment added) made by tecodev
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=1722392&group_id=599

Please note that this message will contain a full copy of the comment thread,
including the initial issue submission, for this request,
not just the latest update.
Category: pic14 target
Group: known bugs
Status: Open
Resolution: None
Priority: 5
Private: No
Submitted By: Masaki Shinomiya (shinobar)
Assigned to: Raphael Neider (tecodev)
Summary: INT plus CHAR makes wrong assembly codes

Initial Comment:
/*** simple.c for testing ***/
#include <sdcc-lib.h>

#include <pic16f84a.h>

static signed int x=0;
static unsigned char y=1;

void main() {
	x += y;
	x += y;
	while (1) {}
}
/*** source end ***/

SDCC 2.7.0 #4802 (May 21 2007) (UNIX)

Two lines of (x += y) makes different assembly codes.
It seems the first line turns wrong.

Initially, x=0x0000.
The first line of (x += y) makes x from 0x0000 to 0x0101 (shoud be 0x0001).
Next line makes x from 0x0101 to 0x0102 (correct).

'x += (singed int)y;' makes the same result as above 'x += y'.

Next sorce gets correct result (x: 0x0000 --> 0x0001 --> 0x0002).

/*** seems correct result ***/
#include <sdcc-lib.h>

#include <pic16f84a.h>

static signed int x=0, w;
static unsigned char y=1;

void main() {
	w = y;
	x += w;
	x += w;
	while (1) {}
}
/*** source end ***/


----------------------------------------------------------------------

>Comment By: Raphael Neider (tecodev)
Date: 2007-05-21 08:00

Message:
Logged In: YES 
user_id=1115835
Originator: NO

OK, here is a patch to fix the pCode optimization, along with a regression
test and improvements for the pic testsuite (src/regression).
Patch is against SDCC r4806.

File Added: pcodeopt.patch

----------------------------------------------------------------------

Comment By: Raphael Neider (tecodev)
Date: 2007-05-21 06:45

Message:
Logged In: YES 
user_id=1115835
Originator: NO

This bug is caused by the slightly buggy pcode optimization (the contents
of WREG is used, but not correctly set up by the `optimized' code). Please
compile your code with 
    --no-pcode-opt
to disable the optimisation and prevent the bug from surfacing.
Of course, this comes at the cost of increased code size...

Regards,
Raphael

I leave this open until I get to fix the optimisation...


----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=1722392&group_id=599