Technical notes on version 3.11.17
                        Adam Kropelin

General:

02Jan06:
- doc: Update version to 3.12.x and remove ChangeLog and ReleaseNotes
  sections since it is better to fetch them directly from SourceForge
  or from the source tarball.
01Jan06:
- snmp: Remove misuse of NETSNMP_CALLBACK_OP_RECEIVED_MESSAGE. Also use
  snmp_{session,pdu} instead of netsmp_{session,pdu} to retain 
  compatibility with the ucd-snmp library.
- Fix compliation of examples/ files: client.c, megaclient.c, 
  newslave.c, upsapm.c  
31Dec05:
- conf: Update EEPROM section comments.
19Dec05:
- win32: Convert Readme.TXT to DOS text format so it will be readble 
  in Notepad (default *.txt handler on Windows). Remove version number 
  since I'll never remember to update it. Point users to manual for 
  detailed install instructions.
- m/s: Fix sense of onbatt/online, which was reversed.
16Dec05:
- doc: Fix latex errors that crept in during SNMP and USB updates.
- configure: Force NIS to enabled since libnis is required by 
  apcaccess. Users may still disable NIS at runtime in apcupsd.conf.
- doc: Remove references to --enable-nis and --disable-nis.
- doc: Update Win32 section to cover nifty new NSIS installer.
15Dec05:
- win32: More installer fixes: Create start/stop links regardless of 
  service or not. Remove apcupsd.events on uninstall, if user wants to.
  Fix compile error in installation directory message text.
- win32: A few last fixes (hopefully): WinNT and 9x need different
  service start methods. Remove extra double-quotes around Win9x start 
  handling because they seem to result in a bogus Exec commandline.
14Dec05:
- win32: Fix an uninstaller bug where apcupsd.conf was left behind 
  even if the user chose to remove it.
- win32: Fix uninstall so it doesn't miss any files. Change destination
  path instructions to help prevent inoperable installations. Fix 
  documentation and start link creation.
24Nov05:
- Get rid of unsupported -c (configure), -u (update battery date), 
  and -n (rename) options.
22Nov05:
- configure: -mwindows is only needed for the link step, so kill
  $WCFLAGS and let all compile steps be driven by the .c.o rule.
- configure: Don't add ncurses libs to $LIBS, only to $POWERLIBS.
  That prevents libncurses from being linked with every executable.
- configure: Kill apcnisd. It requires sysv shmem and since we've 
  eliminated that, so too must apcnisd be eliminated.
19Nov05:
- snmp: Initialize nfds to zero before calling snmp_select_info().
  Fixes crash on Win32 and potential crash on other platforms.
- win32: Build fixes: apc_defines.h -> defines.h, remove use of 
  astr* pending future cleanup, C++ casting fixes.
- win32: Switch to pthreads for the Main_Msg_Loop thread. This
  seems to fix a bug where Main_Msg_Loop called do_status() which
  attempted to take the UPS read_lock. Apparently taking pthreads 
  locks from CreateThread threads is bad. Switching to pthreads
  seems to fix it.
12Nov05:
- No need to specify mode parameter to open() when O_CREAT is not set.
- configure: Simplify SNMP rules. Prefer net-snmp, but fall back on 
  ucd-snmp if it is not found. Look for snmp libs in /usr/local. 
  --enable-snmp and --enable-net-snmp act identically now.
11Nov05:
- Remove lingering references to getNextUps() in Cygwin build.
06Nov05:
- snmp: Add support for SNMP traps in powernet_snmp_ups_check_state().
- snmp: Accept "APC_NOTRAP" as a valid DeviceVendor string, providing
  users a way to disable SNMP trap catching.
- docs: Document SNMP trap catching.
- docs: Fix formatting of master/slave network errors section.
05Nov05:
- apctest: Remove useless deletion and recreation of lock file.
- snmp: SNMP strings are not NUL terminated. Treat them accordingly.
- snmp: Don't claim to support CIs that we don't populate.
- snmp: Regenerate *-mib-mgr.c and *-mib.h files with smidump-0.4.3.
- snmp: Fix heap corruption bug in powernet_check_comm_lost().
- snmp: Generate events during commlost handling.
29Oct05:
- m/s: Convert is_ups_foo() to ups->is_foo().
- m/s: C++ compatibility fixes.
- Add a few set_foo(bool) methods plus a few missing is_foo() methods.
26Oct05:
- docs: Fix page numbers in the index.
- docs: Fix multimon section and a few other details in the surrounding area.
18Oct05:
- bsd-usb: Workaround for compile failure on FreeBSD <= 4.7.
- bsd-usb: Fix a subtle bug in pusb_ups_check_state() which was pointed out by
  a compiler warning.
- usb: Make all pusb_* functions return int and thus match their prototypes in
  usb.c. Previously some functions were defined to return bool but declared
  to return int in usb.c. This led to bizarre return value corruption on
  FreeBSD. Eventually all boolean type returns should be bool, but that
  change is better done all at once.
16Oct05:
- docs: USB updates suggested by Mark Martinec.
- docs: Fix typo in web target.
- docs: Fix first round of major LaTeX bugs. There are many more such as 
  undefined references and overfull hboxes, but this is a step in the right 
  direction.
- docs: Second run of latex should be on apcupsd.tex not manual.tex. On its 
  own, manual.tex is just a massive stream of syntax errors.
- docs: Add missing \\.
- docs: Fix undefined and redefined symbols.
- docs: Kill blank lines after \begin{verbatim}.
12Oct05:
- docs: Move common latex2html flags into a variable so html and web targets do
  not accidentally differ. (Previously -t was not specified for html target).
  Also run apcupsd.html thru a sed script to remove references to
  "apcupsd.html" and thus allow the file to be renamed at will. Copy
  final html output into manual.html to maintain an archived copy of the
  generated output.
10Oct05:
- Rewrite the UPSCABLE description to hopefully eliminate the common mistake
  of thinking the cable model numbers were listed in columns by type. Also
  clean up some of the config directive descriptions and add some blank lines
  to improve readability.
- docs: Fix up some links to the Win32 section.
24Sep05:
- Update examples/snoopdecode.c to latest version and add to Makefile.
- Add SmartUPS serial protocol simulator 'smartsim' to examples/ directory.
- Do not clear_onbatt_msg() when shutting down. This just causes a redundant
  "On Battery" event.
- apcsmart: Convert do/while loop to a regular while loop in UPSlinkCheck().
  The do/while form is harder to read and doesn't add any value.
- apcsmart: Remove pointless check in apcsmart_ups_get_capabilities().
- apcsmart: Fix commlost deadlock caused when locking was added to NIS code.
  We have to drop the UPS lock in UPSlinkCheck(), otherwise NIS is blocked 
  until the link comes back up. Also fix timing bug which caused commlost 
  events every 20 seconds instead of every 10 minutes. To do that, rewrite
  the event loop to not depend on the timing behavior of getline().
- Fix TV_DIFF_MS macro.
- apcsmart: Fix excessive startup delay: Don't probe for commands where
  ups->UPS_Cmd[i] == 0.
18Sep05:
- Be careful not to close trace_fd while daemonizing.
- Add example {on,off}battery scripts that scale CPU frequency down to save
  power when UPS is running on battery. Requires linux-2.6.x kernel with
  cpufreq support and compatible CPU. Inspired by Paul Devriendt's talk
  at OLS 2004.
17Sep05:
- Use fast polling when battery is disconnected. This minimizes the battery
  reattach delay.
- Process actions immediately after volatile poll in addition to after a
  state check. This allows transitions caught by polling to be acted upon
  without waiting thru another state check interval.
- Remove debounce logic. The USB debounce code was vestigial since the big
  USB changes went in.  The SmartUPS code actually interfered with proper
  operation of ONBATTERYDELAY, causing powerfail and onbattery messages to be
  generated ~2 seconds apart instead of whatever the user configured via
  ONBATTERYDELAY.
- USB: Fix selftest detection on RS 1500 models. Delay slightly before reading
  APCLineFailCause because this usage apparently is not valid immediately
  after power failure. Also disable CI_ACPresent when CI_Discharging is
  available because the RS 1500 reports (Discharging=1 && ACPresent=1) during
  self test, which confuses Apcupsd.
15Sep05:
- USB: Work around spurious BatteryPresent=1 bug in RS 1500 firmware.
13Sep05:
- manpage: Default STATUSPORT is 3551. Manually editing cgi is no 
  longer required when changing the port.
7Sep05:
- Fix set_online(int) modifier to change both UPS_online and UPS_onbatt
  like the regular set_online(void) does. Eliminate future ocurrences
  of this bug by writing all set_foo(int) modifiers in terms of 
  set_foo(void) and clear_foo(void).
- Reorder device loop contents so do_action() is invoked early to handle
  preexisting conditions instead of waiting thru a device_check_state()
  period first. The overall order of operations is unchanged; only the
  loop entrypoint has shifted.
- Add BATTDETACH and BATTATTACH events which are triggered by changes in
  the UPS_battpresent flag.
- USB: Fix the battlow FIXME by performing status bit updates the way
  the USB driver used to, namely by clearing the APC part of the status
  word and then setting bits based on CIs. Note that this technique 
  requires a locking change in apcstatus.c to solve the NIS race 
  condition seen with the old USB code.
- NIS: Lock the UPS structure for reading in output_status(). Otherwise
  a client can get corrupt NIS data if the driver is in the middle of 
  updating a field. This is particularly troublesome for the net driver.
6Sep05:
- USB: Update APCLineFailCause interpretation in USB driver with info
  from APC.
- Add XFER_FREQ to represent transfers due to input frequency out of
  range.
- Remove individual UPS_prev_onbatt and UPS_prev_battlow bits in favor
  of a generic status bit change detection mechanism.
3Sep05:
- CGI: Teach multimon about NOBATT and make buffer length tracking in
  parsefield() less prone to errors in the future.
- USB: Make volatile data debug output consistent.
- USB: Set status bits directly from USB data so UPSINFO follows the
  hw state.
2Sep05:
- CGI: Fix a few instances of "sizeof(pointer)" caught by g++ on OpenBSD.
1Sep05
- SNMP: UPSINFO::G[] and UPSINFO::X[] to enum conversion.
- CGI: Fix C++ compilation.
- Kill unused (or nearly unused) fields from class UPSINFO. Also remove
  unused structs while we're in the neighborhood.
31Aug05
- USB: Make older gcc happy by not introducing a local variable in
  the middle of a switch statement.
- USB: My RS 1500 models seem to return APCLineFailCause=8 for
  transitions due to utility power failure. Update return value
  decoding to account for this.
- hid-ups: Name usage 0xFF860052 as APCLineFailCause.
- Convert UPSINFO::G[] to LastXferCause enumeration and eliminate
  the need for core code to know about last xfer CIs.
- Kill off UPSINFO::X[], replacing it with enum SelfTestResult.
- Add support for tracking battery presence via new NOBATT status
  flag.
30Aug05
- Rephrase comm{ok,failure} messages to not assume a serial port
  connection.
- USB: Add CI_APCLineFailStatus support. This allows apcupsd to
  distinguish battery transitions due to self test from those 
  due to utility power failure.
- Increase self test grace period to 20 seconds (from 12) to
  accomodate lengthy USB event hysteresis imposed by the UPS.
- USB: Set self test result string.
- USB: Fix excessive unlocking bug in pusb_ups_check_state().
- NET: Fix minor copy-n-paste error in net_ups_read_volatile_data.
23Aug05
- USB: Use CI_RETPCT instead of CI_APCDelayBeforeStartup which is not
  only cleaner but also causes RETPCT to be properly reported.
18Aug05
- Fix powerflute SIGALRM timer on Solaris by reinstalling signal
  handler whenever the timer expires.
15Aug05
- Increase inter-URB delay to 20 msec as per 3.10.18.
- Fix powerflute compilation on Solaris by preferring
  ncurses/*.h headers over include/*.h when both
  exist.
- C++ fixes for compilation on Solaris.
- BSD and Linux USB rework to report events via callback.
08Jul05
- Bump version number and date
- Back-port inter-URB delay lockup fix from Branch-3_10_17
- Significantly rework USB driver event and poll handling.
  More cleanup to come, and BSD driver still needs update.
22Jun05
- bsd-usb: compile fix
- linux-usb: Filter duplicate reports in pusb_ups_check_state().
- linux-usb: Strengthen CI <-> usage association 
- linux-usb: Only exit event read() loop when an 'important' CI
  is processed
- Add a parameterized version of set_battlow() and make use of it.
- linux-usb: Add some more 'important'-looking events
08Jun05
- Retroactively create my own techlog
07Jun05
- Rename mainsback.in to offbattery.in so default power loss/return
  actions are symmetrical.
- Fix bsd-usb build on FreeBSD-4.11 by allowing Makefiles to add 
  their own include flags via EXTRAINCS.
- Exclude HPUX from signal recursion
31May05
- Code style normalization.
19May05
- C++ cleanliness fixes in libusbhid
27Apr05
- Merge hidu_set_report bug fix from Branch-3_10_17.
24Apr05
- Make apctest work with empty DEVICE setting for USB driver.
22Apr05
- Merge BSD UHCI lockup fixes from Branch-3_10_17.
04Apr05
- Remove trailing commas in enum declarations to make AIX 5.1 +
  VACPP 6.0 happy.
29Mar05
- Make SNMP missing SNMP libs cause ./configure fail
- Remove references to serial port in comms lost/restored messages
  in apccontrol.in since in UPSes may be connected by other means
  (USB, net, etc.).
- More unsafe string conversions. strcpy -> astrncpy and
  sprintf -> asnprintf.
22Mar05
- Convert some potentially unsafe string functions to apcupsd safe
  versions. None of the usages were actually unsafe, but this change
  pacifies OpenBSD and prevents future bugs.