[PATCH] libio: Fix race in _IO_new_file_init_internal initialization order [BZ #33785]

[Florian Weimer]

* Shamil Abdulaev:

> +  while (!atomic_load_explicit (&stop, memory_order_acquire))
> +    {
> +      FILE *fp = fopen ("/dev/null", "r");
> +      if (fp != NULL)
> +        fclose (fp);
> +    }

Sorry, missed this: Is there any reason not to treat fopen failure as a
test failure?  You could perhaps use xfopen/xfclose, but some of us
prefer not to use error-checking wrappers for functions under test.

Thanks,
Florian