[sdcc-devel] [ sdcc-Bugs-1994062 ] pointer compare wrongly done with signed math

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

Bugs item #1994062, was opened at 2008-06-14 21:12
Message generated for change (Settings changed) made by maartenbrock
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=1994062&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: C-Front End
>Group: fixed
>Status: Closed
>Resolution: Fixed
Priority: 5
Private: No
Submitted By: Andy (hal42usa)
>Assigned to: Maarten Brock (maartenbrock)
Summary: pointer compare wrongly done with signed math

Initial Comment:
mcs51 port:
pointer compares are done with signed logic (attested to by 'xrl a,#80'), which is usually not a problem with xdata pointers since xdata space is usually only 1k, but is a disaster with idata pointers.
I have the source compiling with VS2005, but am having trouble figuring out where pointers are getting tagged as signed.

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

>Comment By: Maarten Brock (maartenbrock)
Date: 2010-08-04 16:05

Message:
Fixed both signed and xdata bug in SDCC 2.9.7 #5909.
Also added requested optimization.
1458678 was not related.

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

Comment By: Patryk (patryks)
Date: 2008-06-16 15:21

Message:
Logged In: YES 
user_id=1788180
Originator: NO

Maybe related to pointer/value constness (long standing SDCC problem):
[ 1458678 ] unsigned long *const w, xtea stream cypher


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

Comment By: Maarten Brock (maartenbrock)
Date: 2008-06-16 11:31

Message:
Logged In: YES 
user_id=888171
Originator: NO

It's worse. If you have the pointers point toward unsigned char's the
compiler uses unsigned math.

Or if you move everything to xdata this assignment fails.

char xdata *one;
char xdata *two;
two = &one[140];

error 47: indirections to different types assignment
from type 'unsigned-char near* '
to type 'unsigned-char xdata* '

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

Comment By: Andy (hal42usa)
Date: 2008-06-14 22:52

Message:
Logged In: YES 
user_id=2117929
Originator: YES

getting closer, the compiler is comparing the pointers as bytes that are
neither specifically signed or unsigned. I will try 'default char type
unsigned' as a compiler switch ...
yep, that fixes it, as a workaround.
Now if I could find where the compiler decides to treat idata pointers as
default chars I could fix this for real.


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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=1994062&group_id=599