| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
All regression tests pass.
|
|
|
|
|
|
|
| |
free_device is OK, but it's not safe to do FD_SET() checks on file
descriptors in unallocated devices.
All regressioon tests pass.
|
|
|
|
|
|
|
|
|
| |
This began producing buffer overflows out of the blue in the
regression tests. More weirdly, it was retroactive - revisions after
my commit of 2014-08-24T11:27:11 that introduced this change started
failing in the same way, even though they had passed regressions
before. This may reveal an underlying bug in free_device() that is
phase-of-moon dependent - to be investigated.
|
| |
|
| |
|
| |
|
|
|
|
| |
All regression tests pass.
|
|
|
|
|
|
| |
The factoring on this could use some cleanup.
All regression tests pass. PPS is live in both gpsd and gpsmon.
|
| |
|
|
|
|
|
|
|
|
| |
It came in way, way out of sequence ansd screwed something up. He may
have been fixing some real problem, however. This requures
investigation.
All regression tests pass.
|
| |
|
| |
|
|
|
|
| |
All regression tests pass.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This was broken by commit 575444. The check if the PPS time isn't
referenced with an old message time used PPS offset instead of the
difference between local PPS time and local time of the last message.
The PPS sample was reported via SHM or chrony socket only if the offset
was between -1 and 1000001 seconds.
Keep the local timestamp of the last fix in struct gps_device_t in
addition to the real timestamp and use it to check the PPS delay.
Rename the l_offset variable to delay to avoid further confusion.
Signed-off-by: Eric S. Raymond <esr@thyrsus.com>
|
|
|
|
|
|
|
| |
GNU/Hurd is not following Linux' numbers,
but BSD's numbers. In this case, IPV6_TCLASS is 61 there, not 67.
See Debian Bug #738646
|
|
|
|
| |
bzed had a better fix for this in place.
|
| |
|
|
|
|
|
|
|
| |
Otherwise, FD_ISSET(sub->fd, ...) might crash since sub->fd might have
set to -1 in detach_client() called by ppsthread or main loop
Signed-off-by: Wolfgang Ocker <weo@reccoware.de>
|
|
|
|
| |
IPV6_TCLASS is not defined in socket.h on hurd.
|
|
|
|
|
|
|
| |
As this code only gets called for tcp apparently, trying
to enable ECN here is futile.
Signed-off-by: Eric S. Raymond <esr@thyrsus.com>
|
|
|
|
|
|
|
|
|
|
|
| |
This attempts to enable marking gpsd packets to use a lowdelay
TOS marking in the packet headers. It also enables the ECN bit,
to lower dropping probability for systems that respect ECN.
It doesn't actually DO anything with the ECN bit as at a rate
of 1 per second, there isn't any rate reduction to be had.
Signed-off-by: Eric S. Raymond <esr@thyrsus.com>
|
|
|
|
|
|
|
|
| |
The control socket should support a "?devices" string sent to the
control socket. Unfortunately, this appears to be broken; so fix it.
Signed-off-by: Simon Dawson <spdawson@gmail.com>
Signed-off-by: Eric S. Raymond <esr@thyrsus.com>
|
|
|
|
|
|
|
|
| |
...actually revealed a bug - device-shutdown messages getting lost on the
way out to the test clients. This set of changes mostly fixes it. Some
glitches remain; this state of things passes all regression tests but
attempting to get rid of what now ought to be unnecessary code in fake.py
does not pass. To be continued...
|
| |
|
|
|
|
|
|
|
| |
It's not necessary to keep the capability since time_pps_setparams is
called before root privileges are dropped.
Signed-off-by: Eric S. Raymond <esr@thyrsus.com>
|
|
|
|
| |
Signed-off-by: Eric S. Raymond <esr@thyrsus.com>
|
| |
|
| |
|
| |
|
|
|
|
| |
scons pps=false check hangs with this change.
|
|
|
|
| |
All regression tests pass. PPS is live.
|
|
|
|
| |
gpsmon can't see this yet. That's next.
|
|
|
|
| |
All regression tests pass. PPS is live in both gpsd and gpsmon.
|
|
|
|
| |
All regression tests pass. PPS is live in both gpsd and gpsmon.
|
|
|
|
| |
This logging level is intended to directly explain accompanying I/O messages.
|
|
|
|
|
|
|
|
| |
This involved moving some out of the AIS driver. There is a related
small change in behavior; now, if split24 is on, the Type B half of
a matched pair will be shipped with type 'both'.
All regression tests pass. PPS is live.
|
| |
|
| |
|
|
|
|
| |
All regression tests pass. PPS is live.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
On OS X, /usr/include/sys/termios.h typedefs speed_t as unsigned long.
On Linux, it is an unsigned int. This causes printf() format string warnings.
Fix this by down-casting to an unsigned int and using "%u" as format.
Also discard a close() result.
Signed-off-by: Beat Bolli <bbolli@ewanet.ch>
Signed-off-by: Eric S. Raymond <esr@thyrsus.com>
|
|
|
|
| |
Signed-off-by: Eric S. Raymond <esr@thyrsus.com>
|
|
|
|
| |
Signed-off-by: Eric S. Raymond <esr@thyrsus.com>
|
|
|
|
| |
Signed-off-by: Eric S. Raymond <esr@thyrsus.com>
|
|
|
|
|
| |
Should be no functional change.
Thanks to: Michael Tatarinov <kukabu@gmail.com>
|
| |
|
|
|
|
|
|
|
|
|
| |
No logic changes in this commit. We push naosecond precision closer to
where time hints are shipped. Changes to ppsthread.c are kept minimal;
it needs a cleanup but that will be done in another change that can be
separately reviewed.
All regression tests pass. PPS is live in gpsmon and reported by gpsd.
|
| |
|
|
|
|
| |
All regression tests pass. PPS is live.
|