[Gpsd-commit-watch] [SCM] GPSD annotated tag, release-2.29, created. release-2.29

[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-2.29 has been created
        at  a6781bdea642a5e1b8c4b75adb8fc25c41c2f458 (tag)
   tagging  f3ac336f1b645fde93a173df0674618e59041fba (commit)
  replaces  release-2.28
 tagged by  Eric S. Raymond
        on  Fri Aug 19 12:23:53 2005 -0400

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

Chris Kuethe (39):
      ESR pointed out that I wasn't actually calling serialConfig() anywhere...
      Typo in a comment.
      Turn off BOFH mode.
      typo
      Sign mismatch.
      For operating systems that don't have round/roundf, fake it with rint/rinf
      This defines NAN for us where needed.
      Fixing a few little details
      At least on OpenBSD, endian.h is in /usr/include/sys
      Better OS-specific endianness detection
      A perl implementation of a gpx writer...
      Better comments
      Try to sync up a little quicker.
      A bunch o' tweaks
      Fix/workaround the effects of gpsd_open and friends forgetting the baud 
rate...
      Reconnect if the server goes away.
      OK, back out the 4 second wait.
      spacing
      Apparently 200 samples wasn't quite enough. 256 seems to work better.
      Chomp() wasn't doing the right thing, so use s///.
      More eyecandy.
      Allow for possibility that someone may hook up a military PPS-capable GPS.
      Time is not an effective way of determining when a track starts or ends.
      A whole bunch of rejiggering...
      >From Josh Triplett in #gpsd on irc.freenode.net:
      more fiddling to better handle some errors and bad data
      A new app and some receivers.
      Fixing the name of SiRF's chipset: SiRFstarII or SiRFstarIII.
      New receiver (GP-27) from Tobias Minich
      minor nit
      Some tune ups:
      Little HTML nits found by Dillo's built-in html checker.
      "atcher" mode? What is this "atcher" of which you speak (other than a 
typo).
      - Adjusted the heading sizes to try clarify which items were FAQ items...
      >From Jeff Francis:
      Until I figure out how to embed an SVN token containing the version of a 
file...
      A small bit of googling says that this is the magic
      Use unifdef(1) to remove the WITH_CURSES conditionals.
      cgps shouldn't be listening for inbound connections.

Eric S. Raymond (312):
      First cut at Evermore support.  Not yet tested.
      Fix state machine logic to handle both TSIP and Evermore in the same 
runtime.
      Make the state machine more strict.
      Added Sony support.
      Back out Evermore packet parsing.
      A step back towards parsing EverMore packets.
      Freaking one-character typos...all regression tests pass again.
      Make sure the Evermore driver gets included in the strategy array.
      More typo fixes.
      Fail on unsupported baud rates.
      Instrument so we can see the Evermore message cycle.
      Tighten up the packet state machine.
      Various documentation fixes.
      Add gpxlogger.c to the repo, and more documentation.
      Petr Slansky's evermore_write() routine.
      USED_SET needs to be a separate mask from SATELLITES_SET...
      Big driver structure refactoring.
      Report uniform 95% confidence intervals.
      We know how to scale error estimates now.
      Autodetect DBUS support (from Petter Reinhold).
      It is now possible to configure out NMEA support.
      Reorganize the Hacking Guide.
      Add a driver-prototype file.  Clean up and fix RTCM passthrough.
      Try to fix Oleg's automake problem.  More doc fixes.
      First cut at iTrax and iTalk support.
      Splint-clean the iTalk opacket getter.
      Architecture extension; handle variable sample rates.
      Cleanup patches from Petter Reinhold.
      Try to elicit an Evermore binary response.
      Configure DBUS correctly.
      Oleg Gusev's fixes.
      Add literal_send for Oleg Gusev.
      Try to prevent false lock onto TSIP.
      Add support for the 2-field return of C to client libraries.
      Make sure ITALK_ENABLE implies BINARY_ENABLE.
      Bimary member cleanup.
      Recognize Evermore before TSIP.
      Fix some documentation.
      First cut at merging Wolfgang's RTCM code.
      Initialization fix from Petr Slansky.
      rtcmdecode only works with -O1.
      Instrument the RTCM decoder so we can show that the parity computation is 
hosed.
      Wolfgang's attempt to fix the -O2 problem.
      Set ac_dbus to no if we don't find the relevant library.
      Show states by name rather than enum value.
      rtcmdecode.c is now splint-clean, but still doesn't work at -O2 under gcc 
3.4.2.
      Add some documentation.
      Cosmetic fixes.
      Export the RTCM control block.
      Inline some code as a step towords rurning rtcm_decode inside out.
      Use gpsd's reporting functions.
      Hide some magic constants that don't need to be visible.
      Oops...buffer wadn't saved.
      rtcm_decode now returns a pointer to the parsed structure.
      Refactored rtcmdecode.c now splints clean.
      Added output_mag, conditioned out, just in case we ever want to issue
      Found the workaround for the compiler bug that was fooing up parity 
computation.
      Next steop towards integrating the RTCM stuff.
      Split RTCM parsing into its own module.
      This is the new RTCM module.
      This produces an RTCM decoder that works.
      This produces a working decoder, too.
      Pull -W out of the options...
      Tighten up a buffer size.
      Document the weird known bug.
      With gcc4 we can integrate rtcm.c into the library...
      Document more gcc3 wackiness.
      Merge working Evermore state-machine from Petr Slansky.
      Another step towards RTCM104 integration.
      Correct tdop computation.
      Fix minor bug reported by Oleg Gusev.
      Fix minor bugs reported by Oleg Gusev.
      In theory we now have an RTCM104 driver...
      Rebuild regression tests.
      Don't add .log, it just confuses matters.
      More refactoring, RTCM dump function is now an ordinary dump hook.
      Clean up RTCM error messages.
      This is ready to be documented.
      Create a regression test for the RTCM decoder.  Add a useful header 
comment.
      Add documentation for the RTCM dump format.
      Register the RTCM documentation.
      Add unix_to_gpstime.
      Update some documentation.
      Typo fix.
      Another Evermore patch from Petr Slansky.
      Don't emit binary characters when logging messages with raw binary in 
them.
      Documentation impriovenents.
      Website fixes.
      Solo-user check is implemented for B, C, N.
      Document a minor bug rather than trying to fix it.
      Added srec(5) manual page.
      First checkin of Chris Kuethe's firmware uploader.
      These changes enable Chris Kuethe's firmware loader to compile.
      Add a man page for Chris's firmware uploader, remove the README it 
replaces.
      Rename Chris's firmware loader files.
      Firmware loader rename -- internal filename and #ifdef changes.
      Remove SiRF-specific assumptions from srecord.c and add it to libgps.c
      Splint-clean the srecord code.
      Splint-clean the RTCM code.
      Arrange to splint gpsflash.
      Fix up the splint productions.
      The big gpsflash refactoring.
      Rename the utils.c file from gpsflash to sirfflash.c.
      Splint-cleaned sirfflash.c.
      Handle the new HTML pages.
      rtcm.c is properly refactored, with the bitfield layouts hidden.
      gpsfake can feed the daemon an RTCM stream...
      Fix a small but nasty bug.
      Break the packet tester code out into a separate program.
      Do consistent naming and some splint cleaning.
      Restore some lost splint cleanups.  Stronger regression tests.
      Add gpsflash to the RPMs.
      Previous commit needed a buffer flush first.
      Documentation improvements.
      Stage one of making the RTCM decoder an ordinary driver.  Merge the RTCM
      Step two in RTCM merge: it no longer has a separate header.
      Switch to using C99 initializers everywhere.
      Check in Wolfgang's enhancements as a baseline before hacking them.
      We got message 3.
      Add usage message.
      This goes with parsing RTCM message 3.
      Another Evermore test from Petr Slansky.
      We can unpack type 4 and 5 messages now.
      Wolfgang corrects for a typo in the ITU standard.
      We now unpack type 7 messages.
      splint cleanup.
      In theory we now have a complete decoder.
      I missed Wolfgang's #if 0; fix bugs inside that.
      Comment improvements,  Explain the history better.
      This is a checkpoint for Wolfgang to look at.
      TSIP patch to add wrapup hook that restores old serial-line parameters.
      Back out the automatic DBUS check, it didn't work.
      Add little-endian macros to bits.h.  More fixes for the DBUS problem.
      Another EverMore patch from the maintainer.
      One more step in teaching the packet getter to grok RTCM.
      Use gpsd_hexdump().
      Yet another incremental patch.
      RTCM structure is going to have to be user-visible...
      Turn RTCM processing into a real driver.
      Add progress messages to gpsflash.  splint cleanup.
      More steps towards getting the packet state machine to recognize RTCM.
      Timestamp on both branches of the guard in gpsd_poll.
      Fix a bug in the hexdumper.
      More debugging machinery for RTCM tests.
      We can now see byte offsets when checking a decode.
      More EverMore fixes from Petr Slansky.
      Factor out subframe interpretation so EverMore can use it.
      Changes responding to Wolfgang's long response about RTCM, and factoring
      Solve a port problem for Wolfgang.
      Make HAVE_ROUND available.
      Ugh. Manual pages have to be in the repo in order for xmlto not to be 
required.
      Get rid of statetable.el...
      Force packet_names.h to be rebuilt.
      Fix some bugs in the RTCM104 interface.
      Move RTCM initialization to the right spot.  Passes all regression tests.
      Add References page.
      Documentation fixes and another EverMore patch.
      Wordsmithing.
      Switch drivers properly if TSIP driver grabs an EverMore by mistake
      More instrumentation from Petr.
      First version of wrapinvert processor.
      Add ability to list macros.
      Added the bit-reversal machinery for big-endian machines.
      Write the big-endian macros in terms of bitreverse().
      Avoid a splint problem.
      First (incomplete) cut at repack function.
      RTCM health fields aren't boolean.
      More decoder cleanup and repack progress.
      First step towards eliminating casrs in the decoder.
      Eliminate casts.
      OK, substructures are all gathered.  Now move to one big union.
      Message structure is one big union; grotty casts are gone.
      Splint cleanup after union merge.
      Scrap more unneeded intermediate variables.
      Support, and document, dumping packets of unknown type.
      Prevent a core dump.
      Add -i option to gpsfake.
      First-packet-skipped bug report is spurious.
      More work on the repack() function.  Avoid a core dump when processing 
DGPSIP.
      More work on repack().
      In theory we now have a repack function
      Documentation improvements.
      Add a section on mission and design philosophy.
      Calculate the number of almanac entries correctly.
      First step towards sepaeating out satlink protocol.
      Merged in Petr's stuff on undocumented EverMore control sequences.
      More work at separating the low-level decoder from the RTCM structure 
stuff.
      More separation of the low-level decoder.
      This change might keep us from losing on a 16-bit machine.
      More decoder separation.
      Fix Gary's make problem.
      Petr Slanky's latest Evermore patch and some splint cleanups.
      A simpler big-endianism test using autotools.
      Successfully broke adhesion between RTCM bitfield structures and the 
decoder.
      Break isgps.c apart from rtcm.c.
      Fix some documentation.
      Avoid spurious unknown-sentence messages.
      Narrow an interface.
      More interface narrowing.
      Try to avoid the nasty config problem.
      A large step towards automatically syncing with RTCM streams.
      Discarding entire packets on return to ground state makes RTCM work...
      Support compilers that can't do anonymous unions.
      splint cleanup.
      Return error status if any tests fail.
      Repartition code so we can run individual tests.
      New -t option lets us perform individual regressions.
      OK, this version both passes packet regression and gets partway
      Big-endian support in rtcm.c.
      Move Wolfgang's Emacs command.
      Save Wolfgang's generation script.
      Comment fixups.
      Sepatate character from retry counters.
      Packetizing RTCM revealed a bug deep in the framework code.
      With the bug in the framework code fixed, some delays are not necessary.
      Add the small RTCM check to the regression tests.
      RTCM104 integration.
      More RTCM integration steps.
      Update documentation.
      More cinsequences of renaming.
      This version passes all regression tests and should serve RTCM-104 on 
port 2101.
      Spelling fix.
      Small patch from Wolfgang.
      Back out a change that screws up testing.
      Correction to Wolfgang's fix.
      Fix a core dump induced by the extremely broken GSA in the i.Trek M3.
      Correction.
      Time to truncate some logs for faster regressions.
      Back out the attempts to do port 2101 service.
      First cut at RTCM undumper, and some splint cleanup.
      First steps towards a working RTCM encoder.
      Move towards supporting an RTCM encoder.
      Another step (passing regression tests) towards an encoder.
      Upper level of RTCM encoder works.
      T message undumps work now.
      Encoder is all working, except for floating-point precision lossage.
      The client library can niw read RTCM data.
      Added -n option and probe method to gpsflash.
      Define fields for pseudo-range data.
      Backout the 'Back out the attempts to do port 2101 service.'
      Condition out the service stuff again.
      Update todo list.
      More visible hunting as Wolfgang Rupprecht suggested.
      Wolfgang's attempted fix for the repack() code.
      Clean the test data.
      Finally get the encoder stable by adding some round() calls.
      Parity computation and splint cleanup.
      Add and document -d and -e options for rtcmdecode.
      Try to solve the missing-dependency problem.
      Prevent a core dump when we get a device assigned but don't know its type 
yet.
      Fix minor bug in D response.
      Minor tweak to EverMore driver by Petr Slansky.
      Fix Petr Slansky's NMEA-switching bug.
      Add -r option to gpsfake.
      Add regression test for R=2 mode.
      Add a note about Kismet.
      Enable a probe option in gpsflash.
      Wokaround for Kismet.
      The retry change was not smart.
      Make r=2 mode work again.
      Complain rather than try to write to a close fd silently.
      _Wall cleanup.
      Added expect(), so we can get a version string from the GPS.
      More gpsflash infrastructure.
      Not having I_READ_THE_WARNING set forces probe mode after giving the 
warning.
      Implement -F properly, and fix a bug reported by Wolfgang.
      Harden the code against high data rates.
      Don't use nmea_send() to ship binary!  Duh!
      Comment fixes.
      Document a gpsdrive bug that comes up as a FAQ.
      Typo fix.
      Make sure sirf.c contains documentation for message 27.
      Splint cleanup.
      More splint cleanup.
      Back out the input buffer size change.
      More attempts to get the EverMore to work.
      Unbreak an RTCM regression test.
      Typo fix.
      True North support integrated.
      Finish the True North integration.
      Call close last during wrapup.
      Merge Petr Slanky's gpsd_write() patch.
      Fix some bad naming.  Number of TrueNorth channels is 0.
      gpsprof seems rather broken.
      Initializers to C99 form.
      Fixed the last EverMore test.
      Major change to packet parser -- it now returns the *first* packet 
recognized.
      More fixes for the short-packet logic.
      Fix from Petr Slansky.
      We don't need dummy handlers to ignore sentences.
      Integrate gxlogger.
      splint and error-message cleanup.
      Check in the gpxlogger docs.
      Eliminate duplicated code between sirf.c/sirfflash.c/nmea_parse.c.
      Instrument the offline message.
      Minor changes to SiRF initialization.
      splint cleanup.
      Belgabor's fix for DBUD configuration.
      Make the build work again.
      gpspipe is no longer in contrib.
      Move gpsdspewcurses into the main directory as cgps.
      Teach cgps to use Gary's function for deducing unit defaults.
      splint cleanup after integrating cgps.
      contrib no longer has a Makefile.
      Use a select rather than spinning.
      Splint cleanup before release.
      Make sure cgps terminates on SIGHUP.  Document this.
      Enable gpsflash to compile without SiRF enabled...
      Version bump, preparing for release.
      There is no longer a contrib/Makefile.
      With these changes the RPMs will build...
      Solve the gpxlogger problem.

Gary E. Miller (29):
      Mostly fix garmin to work with new packet parse structure.
      Minor touch-ups.
      gpsprof works again on garmin GPS 18USB.  NaN not handled yet.
      gpsprof with altitude works again on garmin GPS 18USB.
      Undo bad change.  1/0 is not a math operation, it is a plot command.
      Last patch worked with Python 2.3, but not 2.4.
      Fix up Garmin GPS-18 info.
      First pass at getting rid of 'unknown sentence' logs for known but 
ignored NMEA.
      Typo, capital O, not zero in PGRMO.
      Add detection of Garmin Serial GPS.
      Improve PPS logging a tad.  Widen the max lock range a tad.  Now 
GPS-18LVC works w/PPS.
      Tweak the PPS doc.
      Add version number
      Steal ntpshm_pps() computation of ntp precision, use it in ntpshm_put().
      Back out 'precision' addition to ntpshm_put().
      Remove stupid typo.
      Fixes to packet engine from Curt and Josh.
      My apologies.  I let a private change leak in to SVN.  This reverses that.
      Only send garmin init sentences after detecting a garmin.
      Fix typo that clobbered the 5th sat data line.
      fix ^c and (q)uit.
      Oops, private fix leaked.  reverting.
      (q)uit works as well as ^C now
      GPSD_UNITS now sets proper units in cgps.
      LC_MEASUREMENT and LANG now set proper units in cgps.
      Fix screen paint overwrite.
      Garmin driver should not hide the WAAS satellite.  The other drivers do 
not.
      gpspipe now supports [server[:port]], instead of just local only.  Add 
svn ID.
      Update gpspipe doc for the new [server[:port]] option.

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


hooks/post-receive
-- 
GPSD