standards/171572: ptsname(-1) returns errno 0

[Ed Maste]

The following reply was made to PR standards/171572; it has been noted by GNATS.

From: Ed Maste <emaste at freebsd.org>
To: Sergey Kandaurov <pluknet at gmail.com>
Cc: bug-followup at freebsd.org
Subject: Re: standards/171572: ptsname(-1) returns errno 0
Date: Wed, 12 Sep 2012 09:53:45 -0400

 On 12 September 2012 09:40, Sergey Kandaurov <pluknet at gmail.com> wrote:
 > This is regression (or was it an intended change?) from 8.x's MPSAFE TTY:
 >
 > in 8.x ptsname() was changed so that:
 > /* Make sure ptsname() does not overwrite errno. */
 
 It looks like it was intentional, probably based on:
 
 http://pubs.opengroup.org/onlinepubs/007904975/functions/ptsname.html:
 
 ERRORS
 
     No errors are defined
 
 I guess this was interpreted as "shall not set errno."
 
 However - http://austingroupbugs.net/view.php?id=503:
 
 At line 54750, after "shall return a null pointer", add ", and may set errno".
 
 At line 54754, replace "No errors are defined." with:
 
 The ptsname( ) function may fail if:
 [EBADF] The fildes argument is not a valid file descriptor.
 [ENOTTY] The file associated with the fildes argument is not a master
 pseudo-terminal device..
 
 There's some further discussion on ENOTTY vs EINVAL there.  It looks
 like we should at least set errno though:
 
 Index: stdlib/ptsname.c
 ===================================================================
 --- stdlib/ptsname.c    (revision 240164)
 +++ stdlib/ptsname.c    (working copy)
 @@ -77,7 +77,6 @@
  {
         static char pt_slave[sizeof _PATH_DEV + SPECNAMELEN] = _PATH_DEV;
         char *ret = NULL;
 -       int sverrno = errno;
 
         /* Make sure fildes points to a master device. */
         if (__isptmaster(fildes) != 0)
 @@ -87,7 +86,6 @@
             sizeof pt_slave - (sizeof _PATH_DEV - 1)) != NULL)
                 ret = pt_slave;
 
 -done:  /* Make sure ptsname() does not overwrite errno. */
 -       errno = sverrno;
 +done:
         return (ret);
  }