[Coco] Fwd: MW-C Cross compiler bug - Solved

[Luis Antoniosi (CoCoDemus)]

ow great seems to compile. didn't test yet.

but seems that using long constant expressions is not a good idea as
the optimizer does not resolve them right away, instead I see in the
asm code it is calling _lmul when doing 5L*80L for example.



On Sun, Mar 9, 2014 at 8:14 PM, Jamie Cho <silurian at icloud.com> wrote:
> Would it be OK to move the code to GitHub? It is easier to share patches on GitHub and will probably generate less clutter for people's inboxes.
>
> The following code in comp_06.c around line 165 fixes the crashing problem:
>
>                     /* L3696 */
>                 if ( (((int *)var4)[0] == 0)              &&
>                      (valptr = shftcount (var4->varsize))   )
> #else
>                 long _lval = ((long)(cref->cr_Right)->cmdval);
>
>                 if ((_lval <= 0xffff) && (valptr = shftcount (_lval)))
> #endif
>                 {
>                     var4 = cref->cr_Right;
>                     /* for non-coco systems, cmdval is already a direct long */
> #ifdef COCO
>                     var4->cmdval = valptr;
>
> On Mar 9, 2014, at 7:11 PM, Luis Antoniosi (CoCoDemus) <retrocanada76 at gmail.com> wrote:
>
>> ok it will accept:
>>
>> func(400L)
>>
>> but not
>>
>> func(5L*80L)
>>
>> or
>>
>> func((long)5*80)
>>
>>
>> On Sun, Mar 9, 2014 at 6:38 PM, Luis Antoniosi (CoCoDemus)
>> <retrocanada76 at gmail.com> wrote:
>>> and this will cause core dumped:
>>>
>>> func((long)10*80);
>>>
>>> On Sun, Mar 9, 2014 at 6:34 PM, Luis Antoniosi (CoCoDemus)
>>> <retrocanada76 at gmail.com> wrote:
>>>> ok found other bug. when you have long as parameters and you call a
>>>> function using a immediate:
>>>>
>>>> void func(x)
>>>> long x;
>>>> {
>>>> }
>>>>
>>>>
>>>> main()
>>>> {
>>>> func(10*80);
>>>> }
>>>>
>>>> I think it's still pushing an int instead of a long. If you use a long var as:
>>>>
>>>> long v;
>>>> v= 10*80;
>>>> func(v);
>>>>
>>>> it seems to work.
>>>>
>>>>
>>>> On Sun, Mar 9, 2014 at 5:27 PM, Jamie Cho <silurian at icloud.com> wrote:
>>>>> Thanks for the link! I knew it worked for me before even though it was apparently always illegal.
>>>>>
>>>>> On Mar 9, 2014, at 5:13 PM, Chris Osborn <fozztexx at fozztexx.com> wrote:
>>>>>
>>>>>>
>>>>>> On Mar 9, 2014, at 2:01 PM, Gene Heskett wrote:
>>>>>>
>>>>>>> Are you saying that both of my copies (original and the "ANSI C" versions
>>>>>>> of the K&R books are wrong?
>>>>>>
>>>>>> Page 250 of my ANSI K&R 2nd Edition book says:
>>>>>>
>>>>>> void *memcpy(s,ct,n)          copy n characters from ct to s, and return s.
>>>>>> void *memmove(s,ct,n) same as memcpy except that it works even if the objects overlap.
>>>>>>
>>>>>> So yes, you're supposed to use memmove.
>>>>>>
>>>>>> The memcpy vs. memmove saga:
>>>>>>
>>>>>> http://www.win.tue.nl/~aeb/linux/misc/gcc-semibug.html
>>>>>>
>>>>>> --
>>>>>> Coco mailing list
>>>>>> Coco at maltedmedia.com
>>>>>> http://five.pairlist.net/mailman/listinfo/coco
>>>>>
>>>>>
>>>>> --
>>>>> Coco mailing list
>>>>> Coco at maltedmedia.com
>>>>> http://five.pairlist.net/mailman/listinfo/coco
>>>>
>>>>
>>>>
>>>> --
>>>> Long live the CoCo
>>>
>>>
>>>
>>> --
>>> Long live the CoCo
>>
>>
>>
>> --
>> Long live the CoCo
>>
>> --
>> Coco mailing list
>> Coco at maltedmedia.com
>> http://five.pairlist.net/mailman/listinfo/coco
>
>
> --
> Coco mailing list
> Coco at maltedmedia.com
> http://five.pairlist.net/mailman/listinfo/coco



-- 
Long live the CoCo