[sdcc-devel] [ sdcc-Bugs-1745717 ] bit returned in dpl #2

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

Bugs item #1745717, was opened at 2007-06-30 15:10
Message generated for change (Comment added) made by maartenbrock
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=1745717&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: Accepted
Priority: 5
Private: No
Submitted By: Nobody/Anonymous (nobody)
>Assigned to: Maarten Brock (maartenbrock)
Summary: bit returned in dpl #2

Initial Comment:
the following tiny piece of C generates incorrect code
with sdcc -v version

SDCC : mcs51 2.7.2 # (Jun 28 2007) (UNIX)
svn #4870

bit foo(char arg) {
  return arg & 1;
}

the generated code looks like

   0000 E5 82               108 	mov	a,dpl
   0002 54 01               109 	anl	a,#0x01
   0004 F5 82               110 	mov	dpl,a
   0006 22                  111 	ret

thus, the function returns a char in dpl, instead of a bit.

greetings,
  Robert Larice

larice 0x40 vidisys 0x2e de

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

>Comment By: Maarten Brock (maartenbrock)
Date: 2007-07-11 16:43

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

Thanks, this indeed fixes both bugs.
Applied in SDCC 2.7.2 #4879.

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

Comment By: Nobody/Anonymous (nobody)
Date: 2007-07-04 22:56

Message:
Logged In: NO 

the problem seems to have its root in geniCodeCast() where a call to
compareType()
  returns 1 (that is match) though the destination has type bit and the
operand has type char.
looking at compareType() i'd propose (in unbelivable naivity) something
like


--- SDCCsymt.c.orig	2007-06-28 11:16:45.000000000 +0200
+++ SDCCsymt.c	2007-07-04 21:26:28.000000000 +0200
@@ -2130,19 +2130,20 @@
   if (SPEC_NOUN (dest) != SPEC_NOUN (src))
     {
       if (SPEC_USIGN (dest) == SPEC_USIGN (src) &&
           IS_INTEGRAL (dest) && IS_INTEGRAL (src) &&
           /* I would prefer
           bitsForType (dest) == bitsForType (src))
              instead of the next two lines, but the regression tests fail
with
              them; I guess it's a problem with replaceCheaperOp  */
           getSize (dest) == getSize (src) &&
-          !(!IS_BIT (dest) && IS_BIT (src)))
+          (IS_BIT (dest) == IS_BIT (src)))
+        //   !(!IS_BIT (dest) && IS_BIT (src)))
         return 1;
       else if (IS_ARITHMETIC (dest) && IS_ARITHMETIC (src))
         return -1;
       else
         return 0;
     }
   else if (IS_STRUCT (dest))
     {
       if (SPEC_STRUCT (dest) != SPEC_STRUCT (src))


at least for the simple test function 
  bit foo(char arg) { return arg & 1; }

this seems to do the job.
perhaps somebody is so kind and tries to run the regression tests ?
  i wasn't able to do myself.

Robert Larice
larice 0x40 vidisys 0x2e de


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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=1745717&group_id=599