-#ifdef WITH_UTMP
-
-/****************************************************************************
-Reflect connection status in utmp/wtmp files.
- T.D.Lee@durham.ac.uk September 1999
-
- With grateful thanks since then to many who have helped port it to
- different operating systems. The variety of OS quirks thereby
- uncovered is amazing...
-
-Hints for porting:
- o Always attempt to use programmatic interface (pututline() etc.)
- Indeed, at present only programmatic use is supported.
- o The only currently supported programmatic interface to "wtmp{,x}"
- is through "updwtmp*()" routines.
- o The "x" (utmpx/wtmpx; HAVE_UTMPX_H) seems preferable.
- o The HAVE_* items should identify supported features.
- o If at all possible, avoid "if defined(MY-OS)" constructions.
-
-OS observations and status:
- Almost every OS seems to have its own quirks.
-
- Solaris 2.x:
- Tested on 2.6 and 2.7; should be OK on other flavours.
- AIX:
- Apparently has utmpx.h but doesn't implement.
- OSF:
- Has utmpx.h, but (e.g.) no "getutmpx()". (Is this like AIX ?)
- Redhat 6:
- utmpx.h seems not to set default filenames. non-x better.
- IRIX 6.5:
- Not tested. Appears to have "x".
- HP-UX 9.x:
- Not tested. Appears to lack "x".
- HP-UX 10.x:
- Not tested.
- "updwtmp*()" routines seem absent, so no current wtmp* support.
- Has "ut_addr": probably trivial to implement (although remember
- that IPv6 is coming...).
-
- FreeBSD:
- No "putut*()" type of interface.
- No "ut_type" and associated defines.
- Write files directly. Alternatively use its login(3)/logout(3).
- SunOS 4:
- Not tested. Resembles FreeBSD, but no login()/logout().
-
-lastlog:
- Should "lastlog" files, if any, be updated?
- BSD systems (SunOS 4, FreeBSD):
- o Prominent mention on man pages.
- System-V (e.g. Solaris 2):
- o No mention on man pages, even under "man -k".
- o Has a "/var/adm/lastlog" file, but pututxline() etc. seem
- not to touch it.
- o Despite downplaying (above), nevertheless has <lastlog.h>.
- So perhaps UN*X "lastlog" facility is intended for tty/terminal only?
-
-Notes:
- Each connection requires a small number (starting at 0, working up)
- to represent the line (unum). This must be unique within and across
- all smbd processes.
-
- The 4 byte 'ut_id' component is vital to distinguish connections,
- of which there could be several hundered or even thousand.
- Entries seem to be printable characters, with optional NULL pads.
-
- We need to be distinct from other entries in utmp/wtmp.
-
- Observed things: therefore avoid them. Add to this list please.
- From Solaris 2.x (because that's what I have):
- 'sN' : run-levels; N: [0-9]
- 'co' : console
- 'CC' : arbitrary things; C: [a-z]
- 'rXNN' : rlogin; N: [0-9]; X: [0-9a-z]
- 'tXNN' : rlogin; N: [0-9]; X: [0-9a-z]
- '/NNN' : Solaris CDE
- 'ftpZ' : ftp (Z is the number 255, aka 0377, aka 0xff)
- Mostly a record uses the same 'ut_id' in both "utmp" and "wtmp",
- but differences have been seen.
-
- Arbitrarily I have chosen to use a distinctive 'SM' for the
- first two bytes.
-
- The remaining two encode the "unum" (see above).
-
- For "utmp consolidate" the suggestion was made to encode the pid into
- those remaining two bytes (16 bits). But recent UNIX (e.g Solaris 8)
- is migrating to pids > 16 bits, so we ought not to do this.
-
-****************************************************************************/
-
-#include <utmp.h>
-
-#ifdef HAVE_UTMPX_H
-#include <utmpx.h>
-#endif
-
-/* BSD systems: some may need lastlog.h (SunOS 4), some may not (FreeBSD) */
-/* Some System-V systems (e.g. Solaris 2) declare this too. */
-#ifdef HAVE_LASTLOG_H
-#include <lastlog.h>
-#endif