very occasional failures with threads

[rjbs]

See, for example, http://matrix.cpantesters.org/?dist=Data-UUID-1.220;reports=1

originally reported by @andk as https://rt.cpan.org/Ticket/Display.html?id=85978

[macar-cm]

I noticed that this module is not thread-safe. new() and create() in UUID.xs attempt to save and restore the process's umask:

      mask = umask(_DEFAULT_UMASK);
      if ((fd = fopen(UUID_NODEID_NV_STORE, "wb"))) {
         fwrite(&(RETVAL->nodeid), sizeof(uuid_node_t), 1, fd);
         fclose(fd);
      };
      umask(mask);

If two threads attempt to create UUIDs at the same time, the original umask will be lost.

Changing the process umask seems to have been introduced in Data-UUID-0.05 ; I'm not sure why the umask was being reset instead of just specifying a mode in fopen() because it looks like fopen()'s mode argument doesn't include permissions, just how you're going to use the file.