| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
|
|
| |
These will be used for u-blox cycle detection. iTOW will change even
if the GPS has not determeined UTC yet. fTOW adjusts the extact time
of the fix, but is not to be used in cycle checking.
|
| |
|
|
|
|
| |
100% untested.
|
|
|
|
|
|
|
| |
The u-blox UBX-RXM-RAWX packet can be 8214 bytes long! Round up
to 1024 * 9. This should not have been failing silently...
Surprise, this fixes a problem in the NEO-M9P regression...
|
|
|
|
|
|
|
| |
This should force output when people program their u-blox
in strange ways. The obvious effect on the regressions is
just extra TPV output. Need to improve the cycle detection
a bit more to prevent that.
|
|
|
|
|
| |
This made the ublox-neo-m8n.log regression finally look good.
Needed for u-blox 9.
|
|
|
|
|
| |
The protocol version will be needed for future configuration
operations. The cycle enders will be used soon.
|
| |
|
|
|
|
| |
Oddly, no regressions changed...
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The definition of GPS_EPOCH in gpsd.h is correct, but the comment
specified the incorrect corresponding date, which is 06-Jan-1980, not
06-Jan-1981, which isn't even a Sunday.
The century hack in driver_nmea0183.c establishes a 100-year window
for interpreting two-digit years. For now, the range 1980-2079 is
reasonable, given that the GPS epoch is 06-Jan-1980, and the mention
of wrapping "at" 2080 is consistent with this, but the comment
incorrectly placed the GPS epoch in 1981 (perhaps due to believing the
wrong comment in gpsd.h), and the code was written to wrap *after*
2080, rather than *at* 2080. This would place the window at
1981-2080, which would screw up 1980 (admittedly an unlikely case).
This is mostly about not misleading people reading the code than
actual real-life trouble.
TESTED:
Existing tests pass. A test case for a year of "80" would be needed
to actually verify the fix, though.
|
|
|
|
|
|
|
|
|
|
| |
For some reason this was only causing trouble on OpenBSD.
TESTED:
(With subsequent fixes included) Ran "build-all check" on Mac Pro
10.9, Mac Pro 10.14, MacBook Pro 10.9, PowerBook 10.5, VMs for OSX
10.5-10.13, Ubuntu 14.04, CentOS 7, Fedora 25, FreeBSD 10.3, OpenBSD
5.6, OpenBSD 5.6 32-bit, NetBSD 6.1.5, and Beaglebone (arm) Debian 7.
|
|
|
|
|
|
| |
Long standing bug, that became obvious when more NMEA decodes
were added. A 32 bit mask used to index into a 40 member array.o
Also fix over agressive mode setting on $PSRFEPE.
|
|
|
|
| |
Oh, and fix missing decimal point.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The start of this overly large patch was to simply move the test
for MODE_2D/3D flipping, which only affect NMEA 183, back into
driver_nmea0813.c
But that was intertwined with how gpsd_error_model() computes
derived variables, which required major changes to how NMEA 183
mode_2D/3D are set.
This ultimatly led to major regression test results. Almost all for
the better.
I tried to break it up, but moving from one paradigm to another
needed one big jump...
|
|
|
|
| |
NMEA just keeps getting uglier and uglier...
|
|
|
|
|
| |
The SiRFstar step type init will get used on other drivers
that now have similar buffer stuffing issues.
|
|
|
|
| |
Any client can include timespec.h, clients do not include gpsd.h
|
|
|
|
|
| |
Also, remove sime header inclusions discovered to be unnecessary during
the change.
|
|
|
|
| |
Pseudo-range and carrier phase.
|
|
|
|
|
| |
Part of the move from timestamp_t to timespec_t. The double that
is timestamp_t will soon be too imprecise.
|
|
|
|
| |
Which bumps the JSON minor rev.
|
|
|
|
| |
Add notes as reminder for next change.
|
|
|
|
| |
A few JSON changes this cycle: gnssid:svid added to SAT. time added to ATT
|
|
|
|
|
|
| |
The navcom driver could overflow subtype. This is an incompatible
change to the size of devconfig_t. But API version already changed
this dev cycle.
|
|
|
|
|
|
|
|
|
|
|
|
| |
All functional changes inside "#ifdef GREIS_ENABLE"
Includes new regression tests. All regressions tests
pass.
Developed by Gregory Fong, with help and support from
Virgin Orbit.
Signed-off-by: Gary E. Miller <gem@rellim.com>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Now that the Galileo constellation is live, the NMEA 4.1 standard
appears to have standardized on the "$GA..." prefix for
Galileo-specific messages. The lexer currently filters these out; this
patch ensures they go through to e.g. gpspipe -r. (I tore my hair out
for days trying to figure out why these were not being passed through
even though I could see them using screen etc.)
Also added logic to the GSA and GSV message parsing to account for the
Galileo messages. It probably needs more work to match up satellite
numbers between the GSA and GSV messages and to account for the GNSS
type field in NMEA 4.1, but it's a start at least.
I also fixed a couple of situations where the 'GB' prefix was being
ignored even though 'BD' was not. This leads to a regression in
test/daemon/beidou-gb.log, but the "regression" is actually incorrect
old behavior (JSON messages omitting BeiDou satellites) exposed by the
patch.
Signed-off-by: Gary E. Miller <gem@rellim.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Commit 22c3faf added an IPC_RMID operation right after the shmat(), in
order to keep the segment from lingering after gpsd exits. But this
clears the key, making it impossible for a client to attach to the
segment.
This fix moves the IPC_RMID to shm_release(), so that it's not invoked
until gpsd is done with the segment. This required saving the shmid
in the gps_context structure.
TESTED:
On both OSX and Linux, used ipcs to verify that the segment is present
with the correct ID while gpsd is running, and disappears after gpsd
exits. Also verified that warning message for shmctl() appears when
two gpsds conflict over the SHM segment.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Add Windows versions for network functions and disable unused features.
Add detection of various headers and functions that aren't available in
Windows. Note that netlib_localsocket() has no functional implementation
on Windows, but it isn't to be used on that platform and it's not part of
libgps anyway.
Using send() rather than write() seems to work on Windows.
For Windows need to ensure networking is initialised on opening of sockets
(and then correspondingly shutdown upon closing).
Note that within gpsd.h the termios structures and serial related functions
are removed from the Windows build. These are only accessed in serial.c by
gpsd, so since the Windows build is only generating libgps there is
currently no need to modify serial.c.
And for os_compat.h, daemon() is simply disabled under Windows as it's not
used within libgps.
TESTED:
Confirmed compiles under a cross compiler.
Manual build and run of test_libgps on Windows which successfully connects
to a host running GPSD
Manual build DLL version and link with a Windows version of a program
(Viking) that then successfully connects and monitors positions from GPSD
Otherwise no effect on current supported systems.
'scons build-all check' - passes.
Signed-off-by: Fred Wright <fw@fwright.net>
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In compiler.h, it adds the missing "std::" namespace prefixes to the
memory_barrier() definition.
In gpsd.h it:
1) Moves the include of compiler.h outside the conditional 'extern
"C"', since the "atomic" stuff in the former is incompatible with the
latter.
2) Fixes DEVICEHOOKPATH for C++11 (C++11 requires spaces between
literals and string macros).
3) Cleans up some ordering of system includes left over from the
former head/tail setup.
TESTED:
Ran "scons build-all check" on OSX 10.5-10.12, Ubuntu 14, CentOS 7,
Fedora 25, FreeBSD 10.3, OpenBSD 5.6 (32- and 64-bit), and NetBSD
6.1.5. Also tested Qt builds with OSX 10.9 (Qt4 and Qt5), OSX 10.12
(Qt5), and Fedora 25 (Qt5).
Observed the correct DEVICEHOOKPATH in the log on OSX.
|
|
|
|
|
|
|
|
|
|
|
| |
The reason for directly including the contents of gpsd_config.h in
gpsd.h is long gone, so there's no longer any reason to create the
latter by concatenation rather than just making it a normal header
which #includes gpsd_config.h. This is a minimal rework of this form.
TESTED:
Ran "scons build-all check" on OSX 10.9 and Fedora 25 (without Qt in
the latter case, where the Qt build needs further work).
|
| |
|
| |
|
|
|
|
| |
...that my TN-200 with firmware 2.3.2 generates those on occasion.
|
|
|
|
|
|
| |
...is currenly emitting 84 character lines. The data is meaningful,
the checksum is correct, I think someone just got a format string
wrong. GPSD is throwing away usable data.
|
|
|
|
|
|
|
|
|
| |
Each fix updates the error estimates using the generic model unless
the driver provides a better error model. This means that we use the
old generic model for most of our drivers, but some devices provide
their own error estimates - if so, we'll use those instead.
Original diff from Diego Berge, with some corrections from me.
|
|
|
|
|
|
|
| |
now, all itrax features (including italk and $PFST) shall be
controlled with enable-itrax.
obviously i'm hacking on italk again, so make the italk driver compile again.
|
| |
|
|
|
|
| |
functional but committed to allow for in-tree development.
|
|
|
|
|
| |
if unspecified, gpsd will retain its current behaviour: dropping to
"nobody".
|
|
|
|
| |
...(notably the TSIP ones). There's a conflict...
|
|
|
|
| |
...though it's still not hooked up to anything in rtcmdecode().
|
|
|
|
|
|
|
| |
...so the packet sniffer no longer needs to take an argument that is
an rtcm structure. This is a step towards a new and better gpsfake.
No logic or protocol changes. All regression tests pass,
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
| |
The 'cosmetic rearrangement' appears to have busted something
on his system only. Restore the previous version, but with
COMMENT_PACKET added.
Later: yes, this worked around the problem. Now we get to figure out
what broke his toolchain.
|
|
|
|
| |
Teach the daemon to ignore comment packets led with # and ended with \n.
|
|
|
|
| |
Purely cosmetic change, shouldn't even compile different code.
|
|
|
|
| |
problems created by the last big refactoring patch.
|