Sophie

Sophie

distrib > Mandriva > 2006.0 > x86_64 > by-pkgid > 23fcd86b8f6bdf6d1463505ad3db7b3b > files > 391

apcupsd-3.10.18-3.1.20060mdk.x86_64.rpm

             Technical notes on version 3.10.8
                      28 December 2003
                        Kern Sibbald

General:

Changes submitted this submission:
28Dec03
- Added upsBaiscOntrolConverveBattery change to smtp requested by
  Kastus Shchuka (kastus@list.sourceforge.net) 
- Move wall in killpower of apccontrol to after killpower to avoid 
  errors stopping script.
- Install apctest along with apcupsd.
- Fix DISTVER for RH Enterprise
19Dec03
- Added patch by Adam Kropelin that ignores invalid samples. This 
  eliminates the need for the wait, and seems to calm the system. Apparently, 
  even though apcupsd did not use the bad samples, it would break from the
  wait loop, and request an update, which gave even more bad samples. With
  Adam's patch the samples are simply dropped and we continue waiting (if 
  appropriate).  Thanks Adam.
18Dec03
- Add Slackware updated patch submitted by Andrei Mamontov
- Corrected the missing library dependencies in src/Makefile.in
  Thanks to Reg. Clemens for pushing me a bit on this.
06Dec03
- Added usleep(10000) provided by Reg. Clemens, which resolves the
  apcuspd USB problems on 2.6 kernels. Now removed by 19Dec03 patch.
24Nov03
- Added a ONBATTERYDELAY to the conf file. This is the time in
  seconds that apcupsd will wait after detecting a power failure
  until it sets the onbattery state.  If you do not want to be
  annoyed by short power outages (usually less than 10 seconds),
  set the ONBATTRYDELAY appropriately and comment out the
  powerout) "echo ...|wall" statement.
- I finally got the no driver configured error. It turns out that the
  Makefile detection of changes is all botched up. Since I don't understand
  how the author expected it to work, I used the sledgehammer approach and
  do a "make clean" on every ./configure.
- Replaced apcnet.c with Howard Wilkinson's new master/slave code. If 
  anyone *really* needs the old driver, it is in src/apcnet.c.old
- Upgraded to Cygwin 1.5.5

Everything below this was done in 3.10.7
08Nov03
- Do automatic UPS detection in the USB code if no device is specified
  or if the specified device cannot be opened.
- Include patch provided by Howard Wilkinson, but put the new apcnet.c
  in apcnet.c.new until I have a chance to examine it in detail.

  His notes:
  I attach a complete rewrite of the Master side of the mast/slave code
  to support asynchronous connections to slaves.  I needed to do this as
  our network has a number of machines that tend to be powered off for
  long periods and the connect() calls were hanging around for more than
  nettime.
  I also, added a small patch for a new command line flag
  '-P|--pid-file' so that I can easily run multiple daemons on one
   machine.  I needed this as I have a machine that acts as a multiple
   master while getting its power from another UPS managed by another
   machine.

07Nov03
- Lib wrappers was not building correctly on FreeBSD due to the 
  library -lnls, so I copied over the Bacula code for configure.
  Thanks to Lars Köller for reporting this and for his patches.
- Attempt to get mkinstalldirs called by an absolute path rather
  than relative.
- Apply Lars' fixes to platforms/Makefile.in and platforms/etc/Makefile.in
  Install scripts with $(INSTALL_SCRIPT) instead of $(INSTALL_PROGRAM)
- Close all file descriptors except stdin/out/err.
- If there are no stdin/out/err (Windows), then create them and point them
  to /dev/null.  This prevents the serial port from suddenly getting
  debug output sent to it -- chaos.
- It seems that since the beginning of 3.10.x, the APCSMART_UPS never
  reported the proper status. This is because it was not in the case
  statement. Incredible since it is now the "standard" driver! Thanks
  to the user that brought this to my attention.
- Fix a segfault in linux-usb.c due to reference one beyond the command
  table. This happens only on newer USB UPSes. Thanks to the user who
  pointed this out.