openssh unix dev - Dec 2009 - [openssh-portable] utmpx and ut_name

Hello OpenSSH hackers,

The last couple of weeks I've been figuring out how hard it is to
replace FreeBSD's <utmp.h> with <utmpx.h>. I don't think
utmpx is
perfect, but at least it's better than what we have now and at least it
has gone through some form of standardisation. I noticed POSIX says the
following [1]:

| The <utmpx.h> header shall define the utmpx structure that shall
| include at least the following members:
|
| char  ut_user[]  User login name.
| ...

Now ut_user deviates a bit from <utmp.h>, where this field is normally
called ut_name. I've noticed most (all?) implementations just #define
ut_name to ut_user to keep things happy, but I'd rather not put this
obfuscation in any new implementation.

I've noticed OpenSSH also uses ut_name, even when using utmpx. I've come
up with the following patch:

%%%
--- loginrec.c
+++ loginrec.c
@@ -758,8 +758,8 @@
 	utx->ut_pid = li->pid;
 
 	/* strncpy(): Don't necessarily want null termination */
-	strncpy(utx->ut_name, li->username,
-	    MIN_SIZEOF(utx->ut_name, li->username));
+	strncpy(utx->ut_user, li->username,
+	    MIN_SIZEOF(utx->ut_user, li->username));
 
 	if (li->type == LTYPE_LOGOUT)
 		return;
@@ -1316,8 +1316,8 @@
 static int
 wtmpx_islogin(struct logininfo *li, struct utmpx *utx)
 {
-	if (strncmp(li->username, utx->ut_name,
-	    MIN_SIZEOF(li->username, utx->ut_name)) == 0 ) {
+	if (strncmp(li->username, utx->ut_user,
+	    MIN_SIZEOF(li->username, utx->ut_user)) == 0 ) {
 # ifdef HAVE_TYPE_IN_UTMPX
 		if (utx->ut_type == USER_PROCESS)
 			return (1);
%%%

It changes all uses of ut_name on utmpx structures to ut_user. Any
chance this can go into CVS?

Merry christmas,
-- 
 Ed Schouten <ed at 80386.nl>
 WWW: http://80386.nl/

[1] http://www.opengroup.org/onlinepubs/9699919799/basedefs/utmpx.h.html
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
URL:
<http://lists.mindrot.org/pipermail/openssh-unix-dev/attachments/20091225/eaa1565b/attachment.bin>

Hi all,

* Ed Schouten <ed at 80386.nl> wrote:
> The last couple of weeks I've been figuring out how hard it is to
> replace FreeBSD's <utmp.h> with <utmpx.h>. <snip>
I've also noticed something else that's a bit strange. From defines.h:

| #  if defined(UTMPX_FILE) && !defined(DISABLE_UTMPX)
| #    define USE_UTMPX
| #  endif

I've noticed some operating systems have UTMPX_FILE in <utmpx.h>,
pointing to the file name used by the utmpx implementation. I'd rather
not implement this, because the idea behind the utmpx API is to get rid
of direct file access. Is there a reason why OpenSSH uses this
definition?

In my experimental utmpx-branch in the FreeBSD SVN repository I've
changed the code to:

| #  if !defined(DISABLE_UTMPX)
| #    define USE_UTMPX
| #  endif

Cheers,
-- 
 Ed Schouten <ed at 80386.nl>
 WWW: http://80386.nl/
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 196 bytes
Desc: not available
URL:
<http://lists.mindrot.org/pipermail/openssh-unix-dev/attachments/20100107/6702a422/attachment.bin>

Ed Schouten wrote:
[...]
> Now ut_user deviates a bit from <utmp.h>, where this field is
normally
> called ut_name. I've noticed most (all?) implementations just #define
> ut_name to ut_user to keep things happy, but I'd rather not put this
> obfuscation in any new implementation.
Sounds reasonable, your patch has been applied and will be in the 5.4p1 
release.

Thanks.

-- 
Darren Tucker (dtucker at zip.com.au)
GPG key 8FF4FA69 / D9A3 86E9 7EEE AF4B B2D4  37C9 C982 80C7 8FF4 FA69
     Good judgement comes with experience. Unfortunately, the experience
usually comes from bad judgement.