Todd Vierling <tv@pobox.com> writes:

> On 11 Sep 1998, Klaus Klein wrote:
> 
> : >   For instance, strtok() would be hard to make make thread-safe, I
> : >   think.
> 
> Partly, that's why strtok_r() was invented in various standards.
> 
> : It wouldn't be hard; allocating a piece of thread-specific storage to
> : store the internal `last' pointer is easy enough.
> 
> If we want to do so; however, I believe that strtok() is specifically said
> to be unsafe in the SUS98 under threads anyway....

That's correct, it's not required to be reentrant, hence strtok_r(),
and due to its unfortunate design forcing you to tokenize one string
at a time, even when using thread-specific storage, strtok() hardly
qualifies for such a change.

There are, however, some interfaces not required to be reentrant that
may be worthwhile to be converted as described above, i.e. strerror()
comes to mind.