[fpc-pascal] Pchar from constant string

[Tomas Hajny]

On 2023-06-30 11:11, Michael Van Canneyt via fpc-pascal wrote:
> On Fri, 30 Jun 2023, Hairy Pixels via fpc-pascal wrote:
> 
>> 
>> 
>>> On Jun 30, 2023, at 9:03 AM, Hairy Pixels <genericptr at gmail.com> 
>>> wrote:
>>> 
>>> That's why that worked. Hmm with short strings the null terminator 
>>> could be truncated when copying right? Something to look out for.
>> 
>> this is what I meant about truncation. S is no longer null terminated 
>> and now the pchar will fail when passed to C libraries right?
>> 
>> var
>>  s: String;
>>  p: Pchar;
>> begin
>> 	s := 'hello';
>> 	p := @s;
> 
> S will still be null-terminated for ansistrings and shortstrings.
> 
> change to
> 
>   s:=s+s;
>   p:=@s[1]; // and not @s
>   writeln(p[Length(s)+1]=#0);
> 
> and it will print TRUE.

However, there are operations valid for Pascal shortstrings which would 
break this - as an example, older Pascal code used to change the S[0] 
(i.e. the length byte) for a shortstring which wouldn't ensure that #0 
follows such a moved end of the Pascal shortstring.

Moreover, note that #0 is a valid character in Pascal strings, i.e. in 
general it is _not_ guaranteed that the trailing #0 after the end of the 
string is the first one encountered when passing the string to a C code. 
Obviously, you can ensure this in your own code, but it's important to 
stress that neither the compiler nor the FPC RTL check or enforce full 
equivalence of Pascal strings and C null-terminated strings.

Tomas