Re: [PATCH 2/4] ptsname_r-tests: new test module

[Eric Blake]

On 11/10/2011 05:57 AM, Bruno Haible wrote:
> > * tests/test-ptsname_r.c: New file.
>
> You are testing that the return value in case of failure is == errno.
> This is not documented, neither in the glibc documentation, nor in the
> Linux man page. Why should people write
>
>      if (result == EINVAL)
>
> when they can just as well write
>
>      if (errno == EINVAL)
>
> ? The latter code is more future-proof.

Not necessarily.  On many of the existing standardized *_r functions, 
and more particularly on ptsname_r, POSIX requires that the function 
returns an error value, and leaves errno unspecified or requires errno 
to be unchanged..

> Here's a suggested patch:
>
>
> --- tests/test-ptsname_r.c.orig Thu Nov 10 13:52:13 2011
> +++ tests/test-ptsname_r.c      Thu Nov 10 12:38:25 2011
> @@ -79,7 +79,6 @@
>         else
>           {
>             ASSERT (result != 0);
> -          ASSERT (result == errno);
>             ASSERT (errno == ERANGE);
>             ASSERT (buffer[0] == 'X');

That's backwards compared to POSIX conventions.  If anything, I'd expect 
a POSIX standardization of ptsname_r to require the return value to be 
ERANGE, and leave errno unspecified.

--
Eric Blake   address@hidden    +1-801-349-2682
Libvirt virtualization library http://libvirt.org