[NTLUG:Discuss] Re: OT C question

[Johnny Cybermyth]

I just posted the two example files that I'm working with, so have a 
peek at those.  The compiler does complain of a "suspicious pointer 
conversion" but works as posted (again, see my other email).

I'm totally willing to believe that this is a compiler specific quirk. 
I think that is my main motivation for bringing it up.  I've posted on 
microchip's forums and have submitted a support ticked directly to 
Microchip, but no one wants to comment.  I was hoping there were some 
folks listening here who could get gcc or some such to prove this wrong 
or right.

The best case would be to see if native gcc on Linux and MinGW agree on 
the situation.  I've tried it myself, but admittedly, I don't trust my 
experiment fully.

Thanks for the thoughts.  I hope my files clear up the situation and 
that we'll all have a good hearty cup of C pointer semantics today!

Also, if we find that the compiler's funny, it will be something to 
report to Microchip since they proclaim to be strictly gcc compliant.

Todd Robinson wrote:
> 
>> I just recently ran into this funny situation and I don't understand 
>> the fix very well.  Here is the example I was working with:
>>
>> extern unsigned char *myarray;
>>
>> void myfunc(unsigned char *somearray);
>>
>> void main(void)
>> {
>>     myfunc(&myarray);
>>     myarray[0] = '*';
>>     *(myarray+1) = '*';
>> }
>>
>> 1> The function call only worked when the & was added in front of 
>> myarray.  This doesn't make any kind of sense to me 
> 
> Have you confirmed that myarray is initialized correctly? Is it pointing 
> to a char array or some random address?
> 
> myfunc(&myarray) results in giving myfunc() the address of a pointer to 
> a char (char **). In fact, some compilers would complain about that. 
> What issues did you have with passing without '&'? Perhaps myfunc() 
> derefs somearray twice ( ie: *somearray[3] ). Maybe I'm reading into the 
> example too much but the code almost suggests that myfunc() is providing 
> the char array to be used by the subsequent statements. In that case, 
> calling it with the address of the char * variable makes sense (despite 
> myfunc's parm not being defined as char **somearray). If myfunc() is 
> supplying the char array, obviously safeguards need to be in place to 
> make sure it was successful before using myarray later.
> 
> The other two statements in main() look perfectly legal (as long as 
> myarray is pointing to a char array of at least 2 bytes). What symptom 
> did you see when using the bracket notation? My guess is it's some quirk 
> with the rare compiler you are using.
> 
> Using the ... char myarray[] for defining an array of char is a better 
> coding practice (IMO) but the compiler should allow both approaches. 
> Perhaps the compile rules could be made less restrictive (assuming you 
> "must" code the way you showed).
> 
> 
> _______________________________________________
> https://ntlug.org/mailman/listinfo/discuss
>