[sdcc-devel] [ sdcc-Bugs-868103 ] gen.c line 7593 : need pointerCode

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

Bugs item #868103, was opened at 2003-12-30 21:09
Message generated for change (Comment added) made by maartenbrock
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=868103&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: Icode generator
Group: fixed
>Status: Closed
Resolution: Fixed
Priority: 5
Private: No
Submitted By: Ico Doornekamp (zevv)
Assigned to: Maarten Brock (maartenbrock)
Summary: gen.c line 7593 : need pointerCode 

Initial Comment:

Compiling the following source generates the error message :

$ sdcc main.c 
main.c:14: error: FATAL Compiler Internal Error in file 'gen.c' line 
number '7593' : need pointerCode 
Contact Author with source code

---------------------------
#include <stdlib.h>
    
struct foo {        
        char *bar[2];      
};               
                 
                 
int main(void)   
{                
        struct foo f;        
                 
        f.bar[1] = NULL;
                 
        return 0;
}
---------------------------

$ sdcc -v
SDCC : 
mcs51/gbz80/z80/avr/ds390/pic14/pic16/TININative/xa51/ds400/h
c08 2.3.5 (Nov 19 2003) (UNIX) 

OS: Linux / debian
Mail: sd...@ze...



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

>Comment By: Maarten Brock (maartenbrock)
Date: 2007-07-13 14:29

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

Enabled the test and fixed the bug in SDCC 2.7.2 #4882.

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

Comment By: Maarten Brock (maartenbrock)
Date: 2007-07-11 16:20

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

It appears this bug is no longer tested and when reenabled it fails the
assertion.

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

Comment By: Maarten Brock (maartenbrock)
Date: 2004-09-08 07:52

Message:
Logged In: YES 
user_id=888171

Fixed in SDCC 2.4.4 #831

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

Comment By: Maarten Brock (maartenbrock)
Date: 2004-08-26 22:17

Message:
Logged In: YES 
user_id=888171

Update:

I've been tracking this one down and it boils down to some 
strange architecture in SDCC. It starts calculating the 
address of the struct member without actually 
inserting "address of" or dereference iCodes. It only sets the 
isaddr bit. Later on when it's too late it looks at this bit and if 
it's no pointer yet, it gets turned into a pointer. Next time 
around it will be a pointer and it won't be turned into pointer 
again. But if the struct member was a pointer to begin with 
there's no way to fix things. This is called a "small kludge" (?) 
in the source code (SDCCcse.c line 1897).


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

Comment By: Maarten Brock (maartenbrock)
Date: 2004-04-20 21:02

Message:
Logged In: YES 
user_id=888171

This bug is easily reproduced using SDCC 2.4.0 or 2.4.1. It 
looks like the address of the combination of a pointer array in 
a struct gets the storage class of the pointer. In this case 
the compiler tries to store the generic NULL-pointer at a 
generic-pointed-to address instead of an address in data-
memory.

If you add a storage class to the bar pointer you get these 
results:
char data *bar[2]; // works
char idata *bar[2]; // works
char pdata *bar[2]; //NULL written in wrong memory
char xdata *bar[2]; //NULL written in wrong memory
char code *bar[2]; // FATAL...
char *bar[2]; // FATAL...

I just can't seem to pinpoint where this pointer gets set.

Greets,
Maarten

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

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100599&aid=868103&group_id=599