h_errno?

h_errno?

[Eli Zaretskii]

In process.c we have:

  #ifndef HAVE_H_ERRNO
  extern int h_errno;
  #endif

There's also a configure-time test for this variable.  But no code
ever references h_errno in process.c, and the only place that it is
used is in sysdep.c:

  #else /* !HAVE_GETADDRINFO */
	  struct hostent *hp;
	  for (count = 0;; count++)
	    {

  #ifdef TRY_AGAIN
	      h_errno = 0;
  #endif
	      hp = gethostbyname (hostname);
  #ifdef TRY_AGAIN
	      if (! (hp == 0 && h_errno == TRY_AGAIN))
  #endif

		break;

	      if (count >= 5)
		break;
	      Fsleep_for (make_number (1), Qnil);
	    }

What is this variable about, and why do we have its declaration in
process.c?  Can we use errno instead of h_errno in sysdep.c?

Re: h_errno?

[Dmitry Antipov]

On 02/15/2013 01:10 PM, Eli Zaretskii wrote:

> What is this variable about, and why do we have its declaration in
> process.c?

http://www.linuxmanpages.com/man3/gethostbyname.3.php

> Can we use errno instead of h_errno in sysdep.c?

IIUC errno is orthogonal here. The whole code snippet is obsolete
and should be rewritten to use getaddrinfo/getnameinfo.

Dmitry

Re: h_errno?

[Eli Zaretskii]

> Date: Fri, 15 Feb 2013 13:46:17 +0400
> From: Dmitry Antipov <dmantipov@yandex.ru>
> CC: emacs-devel@gnu.org
> 
> On 02/15/2013 01:10 PM, Eli Zaretskii wrote:
> 
> > What is this variable about, and why do we have its declaration in
> > process.c?
> 
> http://www.linuxmanpages.com/man3/gethostbyname.3.php

Thanks.  I guess most of w32 functions that set h_errno can stop doing
that now.

Re: h_errno?

[Paul Eggert]

On 02/15/2013 02:19 AM, Eli Zaretskii wrote:
> I guess most of w32 functions that set h_errno can stop doing
> that now.

If memory serves, h_errno is part of an obsolete POSIX
interface that Emacs no longer uses now, if the platform
supports the current POSIX interface.  As I understand it,
the w32 code emulates the obsolete interface, so if the
policy is to emulate the interface rather than do the
minimal work needed to support Emacs, then w32 code needs to
set h_errno when it emulates that interface.

The declaration of h_errno isn't needed in src/process.c;
it's a leftover from some old code that was removed.  So I
removed the unused decl in trunk bzr 111791.

Re: h_errno?

[Eli Zaretskii]

> Date: Fri, 15 Feb 2013 11:09:14 -0800
> From: Paul Eggert <eggert@cs.ucla.edu>
> CC: Dmitry Antipov <dmantipov@yandex.ru>, emacs-devel@gnu.org
> 
> On 02/15/2013 02:19 AM, Eli Zaretskii wrote:
> > I guess most of w32 functions that set h_errno can stop doing
> > that now.
> 
> If memory serves, h_errno is part of an obsolete POSIX
> interface that Emacs no longer uses now, if the platform
> supports the current POSIX interface.  As I understand it,
> the w32 code emulates the obsolete interface, so if the
> policy is to emulate the interface rather than do the
> minimal work needed to support Emacs, then w32 code needs to
> set h_errno when it emulates that interface.

The problem is that (a) w32 code sets h_errno to values that come from
errno.h, which is clearly wrong, and (b) it sets h_errno in many
network functions that are not documented to set h_errno, like accept,
recvfrom, sendto, listen, etc.  I will fix both of these soon.

> The declaration of h_errno isn't needed in src/process.c;
> it's a leftover from some old code that was removed.  So I
> removed the unused decl in trunk bzr 111791.

Thanks.

Re: h_errno?

[Eli Zaretskii]

> Date: Fri, 15 Feb 2013 21:44:54 +0200
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: dmantipov@yandex.ru, emacs-devel@gnu.org
> 
> > Date: Fri, 15 Feb 2013 11:09:14 -0800
> > From: Paul Eggert <eggert@cs.ucla.edu>
> > CC: Dmitry Antipov <dmantipov@yandex.ru>, emacs-devel@gnu.org
> > 
> > On 02/15/2013 02:19 AM, Eli Zaretskii wrote:
> > > I guess most of w32 functions that set h_errno can stop doing
> > > that now.
> > 
> > If memory serves, h_errno is part of an obsolete POSIX
> > interface that Emacs no longer uses now, if the platform
> > supports the current POSIX interface.  As I understand it,
> > the w32 code emulates the obsolete interface, so if the
> > policy is to emulate the interface rather than do the
> > minimal work needed to support Emacs, then w32 code needs to
> > set h_errno when it emulates that interface.
> 
> The problem is that (a) w32 code sets h_errno to values that come from
> errno.h, which is clearly wrong, and (b) it sets h_errno in many
> network functions that are not documented to set h_errno, like accept,
> recvfrom, sendto, listen, etc.  I will fix both of these soon.

Done in trunk revision 111804.

Re: h_errno?

[Andreas Schwab]

Dmitry Antipov <dmantipov@yandex.ru> writes:

> The whole code snippet is obsolete and should be rewritten to use
> getaddrinfo/getnameinfo.

It already has been.  This is the fallback.

Andreas.

-- 
Andreas Schwab, schwab@linux-m68k.org
GPG Key fingerprint = 58CA 54C7 6D53 942B 1756  01D3 44D5 214B 8276 4ED5
"And now for something completely different."