[gpsd-commit-watch] [SCM] GPSD annotated tag, release-2.34, created. release-2.34

[Eric S. Raymond]

This is an automated email from the git hooks/post-receive script. It was
generated because a ref change was pushed to the repository containing
the project "GPSD".

The annotated tag, release-2.34 has been created
        at  442da6b11b2bad6114f95deb583f6a5571ae3aa9 (tag)
   tagging  f98817ddeebd7e5f82e8cc8c6ecb02649cb8b9af (commit)
  replaces  release-2.33
 tagged by  Eric S. Raymond
        on  Thu Dec 14 18:25:35 2006 -0500

- Log -----------------------------------------------------------------
Tag for public release.

Bill Marr (4):
      Update description of OpenMotif problem...
      Fix long-standing bug in speed display when knots are the selected speed 
unit.
      Remove reference to obsolete smoothing ('-s') option...
      Fix 2 identical, minor typos.

Chris Kuethe (290):
      Prevent cgpxlogger from logging non-fixes. From Robin L Darroch on 
gpsd-dev.
      Fix missing semicolon. From Robin L Darroch on gpsd-dev
      We don't have a HACKING file any more
      We don't have a HACKING file any more
      Allow this to compile when only the SiRF protocol is enabled
      True North requires NMEA support, and fix the erroneous iTalk description
      SiRF name cleanup.
      avoid a segfault if strchr returns NULL.
      Version bump. I don't think we're in Kansas^W 2.33 any more, Toto.
      revert. this breaks the satellite list.
      Better fix to strchr returning NULL
      It makes no sense to compile a gpsd with no protocols defined.
      strchr returns char *, so test it against NULL (a pointer) not 0 (an 
integer)
      strchr returns char *, and NULL if it can't find what was sought.
      strchr returns char *, so test it against NULL (a pointer) not 0 (an 
integer)
      strchr returns char *, so test it against NULL (a pointer)
      Disable RTCM-104 unless a protocol with half a chance of understanding it
      There is no --enable-zodiac configure option...
      A gpsd with no protocols is an invalid configuration; make configure fail.
      Use OpenBSD's nmea(4) line discipline.
      Obey the warnings of lint and cc because they know C better than you do.
      String safety, courtesy of snprintf, strlcat and strlcpy.
      Including config.h shouldn't break the build of gpxlogger...
      Remove unused library. From Stelian Pop on gpsd-dev
      allow this to compile again.
      having "char *poll" causes runtime linker warnings. rename to "pollstr"
      Extend copyright year.
      Don't let cgps spin if read fails. Prompted by Petr Slansky on gpsd-dev.
      The -o option was removed.
      Don't try to do strchr against the NULL pointer.
      Missed this when doing single-protocol tests. From Stelian Pop on 
gpsd-dev.
      adding "--include-deps" creates makefiles compatible with bsd and gnu make
      One more try at fixing the non-kitchen-sink build.
      Remove last traces of --enable-wired-policy. Prompted by Stelian Pop
      ESR removed the rest of the HEADING_FIX code, we'll "disappear" this bit 
too.
      Fix the description of message 27... finally.
      forgot to update the message size
      Fix the documentation of the SBAS corrections message...
      Move the rtcm message structure to its own header.
      Type length meditations.
      fix the $Id$ keywords here
      fix the $Id$ keyword substitution here
      Add subversion ID tags near the beginning of every file.
      matching quotation marks
      Put all the X bits together...
      I broke it, now I fixed it. Problem reported and fix confirmed by Gary 
Miller
      License change.
      Dual-licensing, with permission. Thank you!
      Add --disable-python, for build environments where python is not wanted.
      Make the summary screen show whether X11 and Python support is enabled.
      Document a few more message types
      Document two more messages, and adjust the debug level.
      off-by-one in statistics channel printer.
      Format of PGRME output was wrong. Reported by Petr Slansky.
      Thinking about a better parser for PGRME.
      Return the proper error code. More validation to come
      Prevent the nmea parser from exploding...
      fix the format strings of the timestamps in O and Y messages
      remove redundant toupper() calls...
      Initialize "mask" sooner, and do it only once.
      Clear displayed fields a little better on loss of fix. From Petr Slansky
      Correctly account for the satellites actually used in the positioning
      Eye candy: format the PRN numbers on skyplot as "%-3d".
      Ignore the '$' command unless we're in profiling mode.
      Use the RTCM104 conditional as a hint...
      Set the PRN tag to grey text if it wasn't used in the position solution.
      Forgot to update the test when I fixed the printout of dgps status
      Adjust cursor position properly when writing Almanac Received flag
      properly toggle the navigation paramters display
      I am not a patient man... 5 seconds is too long to wait.
      Invert the sense of the test.
      More stringent firmware checks.
      Fix off-by-one in number of mailing lists.
      mention our new mailing list archives
      Display improvements for DGPS status
      rename sbasvec to dgpsvec which more accurately describes what it's used 
for.
      Prevent gpsd from spinning if the dgpsip stream goes away.
      Make the callback conform to the prototype
      fix argument to perror
      Make the hexdump-to-binary converter part of libgpsd
      use the hex2bin function that libgpsd now provides
      Add the | command.
      Protect the | command by adding a -x command line option.
      Undoing my |-protection change of 28 Sep 2006.
      Undoing my | change of 28 Sep 2006.
      Petr Slansky pointed out that somehow, he was getting fix quality 3 - PPS!
      When generating the GPGGA message, look into the gpsdata.status...
      Update the driver based on newer documentation. From Petr Slansky
      Introducing a new configuration option "--disable-reconfigure"...
      add the sirf-specific reconfiguration inhibitor
      fix pasto.
      rename some of the messages to be more descriptive than "msg8f[]"
      GPGSV sent at 5 second intervals is usually enough.
      evermore now honours --disable-reconfigure
      remove mostly-useless variable
      tsip now honours --disable-reconfigure
      zodiac now honours --disable-reconfigure
      Eliminate compiler warnings:
      quell compiler warning in --disable-reconfigure case
      quell compiler warning.
      quell compiler warning in the case of --disable-reconfigure
      correct function name in debug output
      rename garmin_parse_input to garmin_usb_parse...
      missed one... print the correct function name in debug output
      gcc thinks the size argument is coming from the wrong place.
      italk now honours --disable-reconfigure
      mark the (obvious) end of the ALLOW_RECONFIGURE conditional
      don't try switch itrax receivers if reconfiguration is inhibited
      wrap more configuration lines in #ifdef ALLOW_RECONFIGURE
      avoid buffer overflows.
      allow compilation on systems where stdint.h doesn't exist...
      fix unbounded sscanf("%s")
      Forgot to include config.h...
      fix a couple of sign warnings
      iTalk message types as used in iTrax receivers
      A small tool for switching the speed and protocol of a SiRF receiver.
      More magic numbers. More to come.
      Truenorth tries to play the autobaud game too.
      The TRACK_SET flag was redundantly set, and STATUS_SET wasn't set at all. 
Fix.
      Props to Timo Ylhainen of Fastrax.
      Since we're overloading the altitude field anyway...
      scanf("%f",float) can't turn "?" into a float.
      Don't print floating point values without checking them for NaN
      I don't like fixed width fields.
      More magic numbers. Some from Timo, some from me doing protocol decodes.
      Add itraxtool...
      Add $Id$ svn keyword
      fix signed-vs-unsigned comparison warning
      documentation for itraxtool
      Allowing the fd to drain helps my WBT-200 be more reliable...
      Cleanups:
      On some systems not all of these B______ speeds are defined.
      Remove unused variables.
      Spam the receiver a little harder when sending reset messages
      Not all systems have strtonum.
      this is wrong, but it at least allows the code to compile...
      Proper test for strtonum if I'm going to continue to use it.
      use the right test, duh.
      Allow for gpsd_report and gpsd_hexdump to be suppressed to save cpu 
cycles.
      avoid using snprintf with "%02x" for a format. it's faster to do 2 
lookups.
      squelch compiler warning about this being defined but ununsed.
      This can be done without having to do arithmetic. From Petr Slansky
      Stop using a hand-rolled hexdump.
      This hand-rolled hexdump is debug code that will go away soon.
      remove useless code
      Remove unnecessary memset. Prompted by gprof and Petr Slansky.
      Symbolic constants for minimum and maximum allowable device speeds.
      Rename itraxtool to itraxctl in preparation for a generic gps tool 
(gpsctl)
      rename sirfproto to sirfctl in preparation for a generic gps tool (gpsctl)
      cleanup
      also rename the itrax manpage
      2400 spm is only for Revolution 2X devices.
      Eliminate compiler warnings about unused variables.
      Missing define
      Hush more compiler warnings
      Rename config.h so it can be installed as gpsd_config.h...
      libgps provides strlcat on systems where it is not provided by libc...
      the iTrax leader is "<!"
      "gosd"? What is this "gosd" of which you speak.
      The second character if the iTalk leader is '!'
      iTrax supports a "ping"
      typo
      Conditional claimed too much code. My bad.
      missing newline
      No unbounded string functions.
      make usage a little more helpful
      unused variable compiler warning
      Make this compile again
      fix syntax error
      Enable raw tracker data if the port speed is high enough.
      document message 29
      making italk have a better chance of working. at least it detects packets 
now.
      Avoid printing NaN where floats are expected.
      This is supposed to be 300ms, not 300us.
      changed the wrong delay.
      Document a few more messages
      Consistency fix to E message.
      disable itrax and italk...
      got the nybbles backwards
      clean up merge conflict
      Make sirfctl go away - it has/should be/will be subsumed by gpsctl.
      Updated cgps with support for device dependent behaviour.
      iTalk can come from at least uNav and Sony chipsets.
      Work on the website.
      Avoid a crash (NULL-dereference) if gpsd is pointed at a serial port...
      your axiom is false. :)
      Add my Wintec WBT-200 (itrax3)
      More little nits
      fix pasto in comment. thx2esr.
      This is floattest...
      Add a note about testing your compiler...
      Eliminate a compiler warning. And I got the -o/-O flags in the wrong 
order.
      Bounded string functions.
      Somewhat more involved floating point exerciser.
      Floating point math test integrated into configure.
      floattest is now part of the build process.
      squelch a couple of compiler warnings
      Make sure section 5 manpages are generated too.
      Updates to Evermore driver, from Petr Slansky
      Change  "Evermore" to "EverMore" ... from Petr Slansky
      make this compile again
      fix typo
      update the driver skeleton. from petr slansky
      Make this compile (Mick Durkin)
      Make it compile without sirf.
      gpsctl needs gpsd_id which isn't defined in the nmea-only case.
      squelch compiler warnings in nmea-only compile
      squelch unused variables in TSIP-only compiles
      move the braces inside the ifdef to make it easier to count brackets
      fix single-protocol build for garmin
      FV18 is just funky NMEA, so we require generic NMEA support when using 
FV18.
      last of the single-protocol builds is complete. fixed a goof i made
      add the ANTARIS-based FV-25
      Add the Trimble Copernicus, and fix the URLs
      Signal safety
      add -D to usage
      -t is device type, not speed
      Add the WAAS and SPS specs
      Stop blaming Tri-M for San Jose's broken firmware.
      Add my new FV25 receivers
      Small fixes:
      update GM-38 URL
      update Lassen SK II URL
      The product name is "ANTARIS" - all capitals
      The Axiom (SiRFstar 1) uses NMEA 2.2
      Pass fix data in url to the image generator...
      MFC: handle gpsd connect failure
      TSIP is broken. That's a showstopper bug.
      I don't like $advertise
      if we can't open the device read-write, try to open it read-only. gpsd is
      add some very fluffy eyecandy to gpspipe:
      use an escape sequence rather than a control character to generate 
backspace
      octal character code seems to be a more robust way of generating the 
backspace
      TSIP works again
      Update my Trimble entry, now that TSIP works again
      add a warning comment
      I'm not convinced that DOPs need more than one decimal place...
      Tidy up the satellite info parsing a little
      fully-activate a receiver in no-wait mode.
      RTCM104_SERVICE isn't defined by configure
      make this compile in the RTCM104_ENABLE case
      RTCM104_ENABLE doesn't seem to do the right thing.
      Recognize a few more packet types and log when we reject a packet
      Slightly more informative debug output
      adjust this format a little more
      tidy up the --disable-reconfigure case
      move a log statement so that there is a clear indication of whether a
      lower the log level required to print these messages...
      comment out unnecessary log - a similar log is now emitted by tsip.c
      xgps display tuning.
      Sample TSIP data for regression testing
      update comments in the header
      this time i have a 3d fix
      allow TSIP "super-packets"
      Quirk fixes for the TSIP driver.
      I think LOG_IO is the right level for logging TSIP messages.
      Update my FV25 (ANTARIS) entry...
      fix detection of packet 0x46
      This has been bugging me for a while now
      Finish the OpenBSD kernel timestamping hooks.
      Working timestamps arrived after the NMEA line discipline.
      Because configure allows the selection of PPS on DCD or CTS...
      Document support for PPS on OpenBSD and how it integrates with the OS.
      explain why these ioctls might fail
      We temporarily need root privs to set up the nmea line discipline on 
OpenBSD.
      Wrong size for packet 0x42. Found while testing regression tests.
      Stop using special colours for SBAS PRNs. Instead, use a distinctive 
shape.
      Finally, after all this time, add a legend explaining the color scheme
      Clarify the legend a bit. Prompted by Robin
      Explain that setting up timestamping for ntpd on OpenBSD needs root.
      Grumble.
      Garmin GPS 60 from Diego Berge
      Prevent signal race. One symptom is deadlocks in gpsd_close
      Allow device speed to be fetched with the B command even if it can't be 
changed
      fix confusing comment
      Move some includes around and protect them properly.
      Add a Nemerix board manual. Only documents the NMEA extensions.
      Improve the DBUS test. From Greg Troxel's NetBSD port.
      Add $Id$ tag
      lsof is nasty. fstat is the way to go.
      Quell compiler warning. I'll revisit this when someone sends me a
      Select() on OS X really is broken
      Explain why I'm using a combination query
      Sync documentation with reality
      A bit more about the ANTARIS receivers
      The GM38 is a Furuno chip. $PFEC = Proprietary Furuno Electric Company.
      Getting rid of a few question marks...
      Forgot to put gpsd.php in the release
      fix uninitialized mutex. from greg troxel.
      make this actually work

Eric S. Raymond (443):
      Make gpsfake use a port that doesn't collide with a production gpsd.
      Work around the bugginess of fuser -s prior to version 23 of psmisc.
      Better documentation of some bugs.
      Fix a silly off-by-one error.
      Til's recipe for smashing his hunt bug, and a long explanatory comment.
      Backed out bad multipliers at Rob Janssen's request.
      Remove inadvertent clearing of timestamp.
      Arrange for Python packages to be installed.
      Documentation updates.
      Eliminate -v option.
      Add a needed library depensency.
      Gracefully degrade testing on non-Linux systems...
      Rebuild regression tests; GSA has the current timestamp now.
      Disable iTalk by default.
      Gary Miller says the Garmin bug is gone.  Add a to-do item.
      Implement Owen Wallace's idea to track maximum fd...
      Second stage: lower maximum fd, when possible, on closes.
      Third stage: make *all* changes of all_fds have apprpriate adjust calls 
nearby.
      Fixes by Mick Durkin...
      Fix the NMEA altitude compilation.
      Documentation improvements.
      More documentation fixes.
      Yet more documentation and comment fixes.
      The -d option is no longer necessary.
      This version of gpsfake is OK with the -p option.
      Do a better job of catching invocation errors.
      Pipe (-p) option split into -1 and -p options.  New test added.
      Fix breakage in a test, handle faking a port for regression
      Avoid a nasty bug when reclaiming fds.
      Debian init files from Tilman Koschnick <address@hidden>.
      Documentation fixes.  Add Tilman's udev stuff.
      Add udev support.
      Merge Eivind Tagseth's patch for limiting memory utilization on SBCs.
      Fix typo pointed out my Marshall Crocker.
      Note a compiler bug.  Fix typos.
      Eliminate use of fuser(1) from gpsfake.
      HTMLize the HACKING file.
      Added buffer policy switch.
      Kluge to make pre-2003 Earthmates with EARTHA handshake work again.
      Remove bogus requirement of command-line argument.
      Note fix to Gentoo startup problem.
      Reduce gpsd's default footprint by about 10MB.
      Robin Johnson's fixes for udev support.
      Jerome Cornet's cleaner fix for the Earthmate.
      Various small documentation fixes.
      Introduce NTRIP configuration option.
      Add a note on the regression-drift problem.
      Fix the J option support.
      Enable 'j' command in cgps.
      Support code for querying validity of fix fields.
      splint cleanup after CK's strcat()/strlcopy() cleanup.
      Fixed reversed sense on some tests (only used for debugging).
      Properly clear out latitude and longitude.
      J patch withdrawn, all regression tests pass.
      Correct the setting of some status bits.
      Hmmm...this change may belong in libgpsd_core.c.
      Begin to remove use of the channels limit.
      This version also passes all regressions.
      Channels member is no longer used by the NMEA parser.
      Increase the maximum number of satellites we can handle...
      Document the fact that the channels member is no longer used for parsing.
      Try to unbreak the gpxlogger build.
      Declare prototypes for strlcpy.
      Move to lesstif is needed.
      Clarify a log.
      Fix spurious emission of PGRME when there is no fix.
      Design decision: Trie North data gets mapped into existing fields (track, 
etc)
      Clean up more attempts to set the fix structure directly.
      Make the error-modeling code visible.
      It's not necessary for the error-modeling code to set the flag bits for 
fields.
      Properly clear the DOP members in each sentence.
      Have the error modeller key off whether members are NAN...
      Reduce the size of the J patch still further.  All regression tests pass.
      Shrink the J patch still further.  All regression tests pass.
      Nailed a data-management bug.
      Yet another minor data-management bug.
      Put the error-modeller call in a place where it won't cause a segfault.
      Refactor the way stashing oldfixes is handled.
      Finally, a version of the J-command patch that passes regression tests.
      First stage of cleanup after the J patch; remove unused storage.
      More cleanup.
      Implement no-CASOC policy in a more explicit (and correct!) way.
      Ooops...bad capitalization.
      Stelian Pop's cleanups for the J command.
      Stelian Pop's rewrite of cgpxlogger.c to use libgps and have a J switch.
      Fix reversed test in GPGLL processing that was hosing regressions.
      A gallimaufry of small fixes:
      Avoid issuing spurious 3D error estimates in 2D mode.
      Conditioning DOP calculation on the state of new data actually sees to 
improve
      Possible fix for NTP breakage.
      Do mode checks relative to the user fixbuffer.
      Rebuild for 64-bit machine.
      splint cleanup.
      New version that copes with >12 channels better, by Jeff Francis.
      Declare static some variables that don't need to be exported.
      Fix installed-file coverage in the specfile.
      Another step in eliminating dead storage.  All regression tests pass.
      Eliminate another sneak path that was carrying stale data between cycles.
      Clean up EPE calculation.
      The big dead-storage elimination: we get rid of the newdata member.
      Don't report bad fixes to a DGPS server.
      Factor core library to reduce footprint when SiRF isn't compiled in.
      Refactor core library some more. All regression tests pass.
      Remove unnecessary inclusions.
      Move the logic for poking the NTP daemomn out of the NMEA driver.
      Don't pretend we offer more accuracy than the  GPS system supports.
      Update some documentation.
      Stelian Pop's fix for the J command handling.
      Make the description of how to submit logs more visible.
      Enhance the upload script in the www directory...
      Minor corrections.
      Correct scaling in PGRME output.
      The change in PGRME output scaling requires that we rebuild the 
regressions.
      Restore reporting time to millisecond resolution.
      More tweaks and minor updates for the FAQ.
      Make it clearer that gpsprof is importing a non-system module.
      Documentation and comment tweaks.
      Add a new capability to the generic driver structure: wakeup...
      Eliminate duplicative code in the Truenorth driver.
      Slim the TNT driver down still more.
      Comment fixes for the True North driver.
      Recognize use of @ as a private packet leader character...
      OK, I'm slow today.
      Partial splint cleanup.
      splint cleanup.
      More splint cleanup.
      splint cleanup of gpsflash code.
      Update the to-do list.
      Add a needed header.
      The link to the Hacking Guide was missing.
      Fix some markup errors in the Hacker;'s Guide.
      Switch us to using lesstif rather than OpenMotif, as OpenMotif won't ship
      Fix another bug in the wakeup logic.  Document the move to lesstif better.
      This temporary reversion was needed to regression-test the next change.
      The ugly code for recognizing DLE-led packets...
      Code and regression tests are back in sync.
      Note that Evermore is now supported.
      Move some things off the to-do list that no longer belong there.
      Correct an erroneous comment.
      New section on design-ahead; non-PVT data and the Version 4 protocol.
      splint cleanup.
      Eyvind Tagseth's change to the way channel fds are handled.
      Add undocumented -f option for J flicker checking.
      This version of gpsfake seems to show that the J command works fine.
      Spec file comment now up to date with respect to new features.
      cgps.c no longer needs to declare UNUSED. it's in gpsd_config.h now.
      Minor documentation tweaks.
      Packet cleanup from Petr Slansky.
      Minor documentation tweaks.
      Partial splint cleanup.
      First version of gpsctrl, untested.
      Enhance gps_query so it's sprintf(3)-like, for more conventient assembly
      Keep the Python library in sync.
      Simplify the gpsctrl code...
      This is the proof-of-concept for a version of gpsctrl...
      Andreas Stricke's namespace cleanup.
      When to use C, when to use Python.
      Add gpsctrl.1 to the set of manual pages exported to the website.
      Add a TODO item.
      Oops, I forgot this isn't HTML.
      Oops -- forgot to check in the manual.
      Add some guard code for devices that don't have mode and speed switches.
      More guard code to prevent redundant control operations.
      Clean up some unfortunate nomenclature discovered while working on 
gpscntrl.
      Fix libgpsd(3) so it's usable when ntpshm_init() has not been called.
      More work on gpsctrl.
      Split the "initializer" driver method in two.
      Rename some methods for clarity.
      Cleanup after the method-splitting patch.
      This temporary reversion was needed to regression-test the next change.
      Make it possible to disable reconfiguration.
      Add a to-do item.
      Client code cleanup and simplification.
      Fix and document the implementation of the J switch in gpspipe.
      Typo fix.
      Implement -j option in xgps.  Document it in cgps.
      Split the initializer method for FV18 and Evermore.
      First stage of gpspipe cleanup.
      Get rid of several flag variables...
      Attempt to transplant the methods ftrom garminctl...
      Document things in the bug queue.
      Typo and documentation fixes.
      Get rid of an annoying compiler warning.
      splint cleanup.
      Commit a righteous featurectomy -- smoothing option is gone from xgps.
      Shut up splint some more.
      gpsd now ships an I response when it assigns a device to a user.
      Enable clients to display the GPS type.
      Clean up the notification logic in the device-assignment code.
      Make the subtype member long enough for a Garmin.
      Improved version of code for shipping a firmware version.
      Make gpsctrl emit a usage message if you give it nothing to do.
      More action logging in gpsctrl.
      Back down to a version of the True North code that was working...
      Fix a typo.
      Fix a stupid bug in the firmware tpe notification code.
      Encapsulate the logic for waiting on packets in gpsctrl.
      Drivers have to explicitly set the driver_mode field when they do a 
switch...
      When gpsctrl is given no arguments, it IDs the selected device and exits.
      Clean up a saved_baud use.
      Pure refactoring.
      More pure refactoring.
      Another small refactoring step.
      Add a comment and initializer (so far not used)...
      Another strncpy/strlcpy fix.
      Get rid of a stray bracket.
      Define a function to do ID reports with subtypes...
      Define uniform log level macros.
      Sigh...avoid harmless but *annoying* name collision...
      TSIP driver can report a subtype.
      splint cleanup.
      Untested (but harmless if buggy) code to get subtype info from a Zodiac.
      Convert some log levels I missed first time around.
      Insert delay before close to give the device time to settle.
      Use xmlto or xsltproc conditionally to rebuild the man pages.
      Mick Durkin's fix.
      Integrate Mick Durkin's notes on writing a GPSD driver.
      Cosmetic changes to go with Mick Durkin's document.
      Interleave NMEA subtype probes with received packets.  This seems to
      Track the change I just made in the driver API.
      Document the reason for splitting up the probe sends.
      Minor markup fixes.
      Binary turnoff code removed at Gary's request.
      Make sure none of the trigger strings for NMEA drivers are identical to 
probes.
      Go through SiRF NMEA on tbe way to SiRF binary.  This doesn't break the
      At the moment it doesn't look like supporting iTalk binary is a win.
      Removed on Chris Kuethe's say-so.
      Chris fixed one of our to-dos.  (The fix needs some mod, though.)
      Fixed 'E' reporting to emit '?' for no data...
      libgps change needed to go with improved error reporting from the daemon.
      Change the name of gpsctrl to gpsctl.
      Toss out overlong NMEA packets.
      Added the Hall of Shame page.
      More blunders for the Hall of Shame.
      Add links to Hall of Shame.
      More for the Hall Of Shame, and some minor fixes.
      Oops...this shouldn't have had my copyright in it.
      Various small markup fixes.
      Typo and rename fixes.
      Add the page for hardware vendors.
      Follow up on a name change.
      Minor fixes to the page for vendors.
      Minor corrections.
      Fixed the title.
      Axiom Sandpiper II (SiRF-I) works.
      Fix the site-upload script to cope correctly with the axiom directory.
      Move the vendor docs into their own directory.
      Fix a broken link.
      Changes by Mick Durkin.
      Added documentation for NavMan and Motorola devices.
      Move the axiom stuff under vendor-docs.
      Get rid of a mistaken upload command that could stomp my personal site (!)
      Banish font sizes fixed in points...
      Begin adding documents from Chris Kuethe's pile.
      More references from Chris's pile, and a typo fix or two.
      Fix an error in generation of the ID message at startup...
      Minor improvements to the FAQ.
      Modify floattest.c slightly so it returns 1 if the test fails.
      Rebuild regressions...
      Quiet down the compiler.
      Compiler-warning and splint cleanup.
      Revise the bug list.
      Add an item to the to-do list.
      A minor correction.
      Bug report modified at Gary's suggestion.
      Patch to support configuration on Debian Sarge.
      Cleaned-up, de-Adsensed version of a site search box goes on References 
page.
      Fix Andreas's automake conditiional.
      Turn the Google box into xhtml.
      Move the Gentoo ebuild into the repo.
      Petry Slansky's patch to widen the subtype window a bit.
      Typo fix pointed out by Bill Marr.
      Big cgps front end rewrite from Jeff Francis.
      Several small fixes:
      Another gallimaufry of small fixes:
      Zero the private drive storage at device-activate time...
      Type-forcing option now works in gpsctl and is documented.
      Conditionalization cleanup patch from Petr Slansky.
      Follow through on a rename.
      First step towards restoring parameters on close:
      Back-to-NMEA-on-exit logic for the Evermore, too.
      Split TSIP initialization into a subtype probe and a configurator.
      Make sure TSIP probing only happens once.
      Stop trying to reconfigure Sony CXD2951 and Motorola OnCore chips.
      Remove an unnecessary make.
      Add to-do item about restoring driver state on exit.
      Split wrapup method into revert and wrapup.  The revert method is onlt 
called
      Changes from Mick Durkin.
      Minore corrections so the state-restore command plays well with 'B'.
      Give trimble its own subdirectory.
      New manuals, and a checksum file, linked from the References page.
      Notice to vendor reps.
      Incorporate good suggestion by Chris.
      Avoid a confusing message in the log.  Bug report by Petr Slansky.
      Fix for EverMore by Petr Slansky.  He reports it working.
      Add verification tag for Google AdSense and analytics.
      Added gpscat to the distribution toolset.
      Add cross-references to new tools.
      Evermore fixes and move of a close message, from Petr Slansky.
      Count packets, not fragments.
      Note that expecting $Ack105 only works on a SiRF-II.
      Added AdSense ads (say *that* six times quickly!)
      Remove sentences field.
      Group hardware entries by vendor.
      Apply the normal stylesheet to this page.
      Typo fix.
      Move the checking of probe functions...
      Integrate Chris's Wish List page.
      Menu tag was too wide, fix it.
      EverMore patch by Petr Slansky, and a link for the project page.
      Do a better job of telling gpsd_wrap() that a device has been removed.
      Third revision.
      Change the initialization/deinitialization API slightly, and document it.
      Temporarily disable some code until I figure out what is going on.
      Explanatory comments by Mick Durkin, with a few corrections.
      New control flow works on SiRF.
      If we see an EverMore binary packet, we must reconfigure for EverMore 
*binary*.
      No need to configure the device if all we're doing is checking 
accessibility.
      Mick Durkin's second round.
      Folded in Chris Kuethe's suggestions.
      Pull the "configurable" bit out of the context structure & make it 
per-driver...
      Fix a glitch, apparently due to some race condition in gpsfake...
      Minor corrections.
      Make back_to_nmea a public flag (rather than a per-driver private flag)...
      Remove some conditionalizations that don't make sense.
      New SiRF probe logic works with SiRF-II, should work with I and III as 
well.
      SiRF autodetect logic has been fixed to not require $Ack any more.
      Added Chris Kuethe's PHP test station script.
      Start the PHP script on the road to modern HTML.
      Another XHTML correction.  Arrange for SVN keyword expansion.
      More XHTMLization.
      Start on CSSing this.
      Fix a slightly broken test...
      Use watcher mode in the PHP script to avoid timing problems with the 
query.
      Get rid of some ugly CSS hacks that aren't actually needed...
      More CSS cleanup.
      PHP script is now fully CSSed.
      More CSSing.  Add an explanatory caption to the skyview.
      Conditionalize so the satellite-view caption is not displayed...
      FAQ update.
      Make gpscat able to set parity and stopbits.
      Add an item to the list of bugs exposed in other software.
      Better recovery from a bad pid file.
      Fix a long-standing bug in NMEA processing.
      Remove Evermore NMEA, it was pointless.
      Fix for a bug turned up by Mick Durkin.
      Fix bug reported by Chris Kuethe.
      Revert the results of Chris hacking too fast...
      Fix a minor bug: spurious "no probe matched" report.
      Change the configurator method to take a packet sequence number 
argument...
      Poll for the Navigation Parameters in SiRF message 0x13...
      Fixes to SiRF reversion logic...
      splint cleanup.
      Stop trying to defer device activation, it breaks the regression tests.
      Minor website update.
      gpsfake can now recognize and gather TSIP packets.
      Fix a comment.
      Make the -l option of gpsfake more useful.
      Avoid strange gpsfake -l bug at end of input.
      Document hexdumping in gpsfake.
      Make our assumptions about the units in Garmin error estimates
      Gary reports PPS is working...
      Split the packet-sniffer internals out of the session structure.
      Check in regress-builder, after having used it to detect some minor build
      Gary dropped a stitch.  No big deal.
      Separate the packet-level stuff from the rest of the library declarations.
      Correct a minor error in the production for a regression test.
      A user figured out how Garmin encodes signal strength.
      Some name changes in formal arguments,  No logic changes,
      Put the old code for computing Garmin SNRs back in, conditioned out.
      Teach Garmin how to return a subtype ID.
      Updated START_SUBFRAME from current IERS bulletin.
      Halfway to new sky-view conventions in xgps.
      In xgps, WAAS satellites are now represented by diamonds.
      Document the sky-view conventions.
      Additions to the history file.
      Fix ALLOW_RECONFIGURE compilation.
      Chris's build broke after my attempt to clean up gpsd.h on 30 Nov 2006.
      This is a try at re-merging my changes.
      First cut at making gpsfake.py work using lsof(1) rather than
      Remove a magic number.
      Hide an RTCM detail better.  No logic changes.  All regression tests pass.
      Change the way the RTCM decoding is layered...
      Improved documentation for the ISGPS layer.
      Make the code for Magnavox-format output work again...
      No demand for Magnavox output yet, so we'll condition out that code...
      RTCM decoder doesn't need to know about the full session structure any 
more.
      Add an interesting documntation link.
      Back out my last change but one to rtcmdecode, it broke a regression test.
      Found and fixed the bug that was breaking an RTCM regression test.
      Move strl functions into their own module...
      Break up some modules a little finer, in support of some Python bindings.
      First cut at a Python binding for the packet.c module.
      Another step toward a Python wrapper for packet.c.
      Fix a comment.
      Eliminate redundant setting of CS7/CS8.  All regression tests pass.
      Evermore patches and one documentation tweak from Petr Slansky.
      Correct bad comment arising from a misunderstanding.
      Modified the gpsfake.py code to set serial-device modes...
      Have gpsfake set VMIN the same way serial.c does.
      Work around a potential race condition...
      Attempt to avoid an occasional race condition in gpsfake.
      Fix a bug that occasionally put float data where no float data should be.
      Insurance against flaky drivers.
      Fix some indentation in gps.py.  Cosmetic change only.
      Checkpoint of the incomplete and half-tested support for setting serial...
      Eliminate threading from gpsfake.py.
      Reintroduce line buffering in gps.py.
      Patch suggested by Andreas Stricker.
      Fix a (generated) misleading comment.
      The TSIP problem is solved.
      Followed some advice from Andreas Stricker.
      Bug fixes from Robert Berger.
      Fix a bug in the processing of comment packets.
      Another patch from Andreas.
      This version works of gpspacket.c works and allows us to set a callback.
      Give gpspacket.c a real doc string.
      Name change.
      Fix a packet-sniffer bug in comment processing.  With this, gpscat 
acquires
      Change gpsfake to flush data written from gpsd before each feed.
      This patch solve the OpenBSD TIOCDRAIN problem.
      Splint cleanup.
      Now that TSIP packet analysis is working again, check in the test loads.
      Our formula for computing estimated track error was computationally 
unstable.
      Update the HOWTO.
      Add a needed warning in the gpscat documentation.
      Try to solve a site-rendering problem by tweaking the CSS.
      Don't rely on code previous to gpsd_error_model() to set error estimates 
to NAN.
      Rebuild the regression tests...
      To-do item from Hamish.
      Bug pointed out by Mick Durkin.
      Fix from Petr Slansky.
      Note Chris's Mac OS/X select(2) bug.
      Create the upstream-bugs page from part of the TO-DO file.
      Add new regression test.
      Add a new regression test, speed up all existing regression tests.
      Yell about the Mac OS/X select(2) bug.
      Toss a couple of superannuated scripts, document the rest.
      Tweak the build machinery so it can build RPMs again.
      First cut at tools for building and fetching RPMs from remote sites.
      Another hack at getting getrpm/putrpm to work right.
      More polish on the getrpm/putrpm pair.
      Add a newer version of the Garmin Binary spec and some new material to 
NMEA.txt.

Gary E. Miller (64):
      Fix problem that causes gpsd to hang in tcsetattr() when being run from 
gpsfake.
      Fix typo in comment.
      Sync gps.1 with gps.xml
      Add a bit more logging to ntpshm_init.
      Revert change.
      Add in regression for Dave Highland's uBlox LEA-4H module.
      Add in regression for Jeff Fisher's Pharos GPS-360.
      Tweak comments for uBlox LEA-4H.
      Only allow PPS after four good fixes.
      Fix comment.  Code is now tested on BE/LE and 32/64 bit CPUs.
      Add Hummingbird test
      Fix uBlox regressions for PGRME fix..
      Fix bu303 regressions for PGRME fix..
      Fix bu303 regressions for PGRME fix..
      Fix typo, from Carl Brown
      Tweak the NMEA fudge time
      Document the $ command.
      Add packet length debugging messages.  Fix endianess problem.
      Add super-raw mode to gpspipe for ckuethe
      Document super-raw mode in gpspipe.
      ckuethe points out that super-raw is not just SiRF binary.
      Start to add support for Garmin binary over serial port
      More steps on the way to Garmin binary over serial.
      Garmin serial binary is very similar to TSIP binary.
      Now usually detects Garmin Serial binary and calls proper driver.
      Decodes Garmin Serial binary packets OK...
      NTP, PPS and cgps work now.  Init is still problematic.
      Fix Garmin USB as broken by recent Garmin Serial Binary changes
      Fix GPGSV output, tweak the file comments
      Turn off Garmin position averaging
      Don't init Garmin NMEA if we are going to init Garmin Serial binary
      Patch to allow PPS on CTS instead of DCD.
      Fix typo in PPS patch. Disable automatic garmin binary init.
      Move all packet creation into one function.
      More packet handler consolidation.
      Previous commit broke the build; fix it.  From:  Petr Slansky.
      Add more PPS debugging
      Add a bit more verbosity to configure.  From Dave C.
      Improve Garmin Serial binary detection by reducing TSIP false positives.
      Adjust PPS to also work with a 2Hz square wave.
      Garmin now respects the ALLOW_RECONFIGURE option.
      Fix typo in sirfmon.1, sirfmon.xml
      Fix typo in the typo fix in sirfmon.1, sirfmon.xml
      Add new program: garmintool.
      Make germin_ser_parse() confirm to the generic parse_packet() prototype.
      Fix SVN ID, reported by Petr Slansky.
      Second try at SVN Id.
      3rd try at versioning.
      Cleanup some warnings.
      The present working directory (pwd or '.') is not in many peoples paths.
      Add J command to gpspipe.  Thanks to JeffFrancis.
      Fix units error in Garmin speed.  Thanks to Jeff Frances.
      Change the name of garmintool to garminctl
      update garminctl.1 to match garminctl.xml for people without xmlto.
      Convert garmin_usb to run in serial emulation mode...
      Silence stupid splint message.
      Add some more PPS logging.
      Revert an overzealous search/replace.
      Fix bad check for garmin_gps in /prc.  Thanks to Eivind Tagseth.
      Fix sat display for Garmins.  Thanks to Jeff Francis.
      Fix bad packet logging.  Delay the ACK a bit.
      garmin_usb_configure() is NOT optional.
      Fix Garmin USB initialization.
      ACKing a garmin packet too soon may hang things.

Jeff Francis (3):
      When the display is 80x24, don't display the title line...
      Added error estimate fields back into cgps per ESR's request.
      Typo.  'dgps' is not the same as 'gpsd'...

-----------------------------------------------------------------------


hooks/post-receive
-- 
GPSD