lvstring: fix size handling

[benoit-pierre]

The chunk size is the allocated size minus one (the terminating \0). Fix methods that don't have the same definition.

---

This change is 

[poire-z]

Looks like it is the case indeed.
This is all confirmed by the fact there is n+1 in all the calls to malloc & realloc : sizeof(char8/32) * (n+1), right?
So, we were just always allocating one slot too much, that ended up just being not used ?
Any other method that was using size right ?

[benoit-pierre]

Looks like it is the case indeed. This is all confirmed by the fact there is n+1 in all the calls to malloc & realloc : sizeof(char8/32) * (n+1), right? So, we were just always allocating one slot too much, that ended up just being not used ?

No, the +1 allocation is right (for the extra terminating \0), again size is the usable size.

The problem is setting size to n+1, which means a following call to append('c') for example would overflow the buffer by one character.

Any other method that was using size right ?

Yes, reserve.

[poire-z]

No, the +1 allocation is right (for the extra terminating \0), again size is the usable size.

I meant it the other way around:
It's because size is indeed the usable size that the code was already using everywhere malloc & realloc : sizeof(char8/32) * (n+1).
Or I still don't get it ? :/

[benoit-pierre]

No, the +1 allocation is right (for the extra terminating \0), again size is the usable size.

I meant it the other way around: It's because size is indeed the usable size that the code was already using everywhere malloc & realloc : sizeof(char8/32) * (n+1).

Yes, and setting size = n, and checking for not enough capacity with pchunk->size < len.

[poire-z]

Same dilemna as in #543 (comment) in here. Waiting for a common decision :)