Re: [PATCH] Avoid strtok for thread safety

[Nathan Myers <ncm-nospam at cantrip dot org>]

Hi Paolo,

I understand that you're not introducing new bugs, but we need to fix 
any we find as we go along.  I'm thinking something like this:

   {
     size_t __i = 0;
     const char* __save = __s;
     while (__i < _S_categories_size + _S_extra_categories_size)
       {
         char* __next = __save + strcspn(__save, "=;");
         char* __new = new char[__next - __save + 1];
         memcpy(__new, __save, __next - __save);
         __new[__next - __save] = '\0';
         _M_names[__i] = __new;
         ++__i;
	 __save = __next;
         if (*__save == '\0')
           break;
         ++__save;
       }
     if (__i != _S_categories_size + _S_extra_categories_size ||
         *__save != '\0')
       throw runtime_error(); 
  }

Again, untested, but maybe better than my last posting.
(The exception chosen, "runtime error", is according to Josuttis.
I don't have the standard handy at the moment.)

Nathan Myers
ncm-nospam@cantrip.org

On Wed, Nov 27, 2002 at 01:02:24AM +0100, Paolo Carlini wrote:
> Nathan Myers wrote:
> 
> >strpbrk() returns NULL if it doesn't find any of the characters
> >searched for.  What happens then?
> >
> Everything can happen, probably a segmentation fault. There are no checks on
> the validity of the string, *exactly* as there weren't before.
> 
> If the string is constructed by locale::locale(const char*) this is not 
> really an issue,
> but, as you said before, if it's passed by the user it can be, if we 
> want: someone told
> me that this area is basically QoI, since the standard prescribes 
> nothing about
> all of it.
> 
> So...
> 
> We should decide together which checks to add, how much expensive and 
> what to do
> if they are not ok. Perhaps throw an exception? Which one?
> 
> Ciao, Paolo.
> 
>