| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
This logging level is intended to directly explain accompanying I/O messages.
|
|
|
|
|
|
|
|
|
|
| |
Way too much logging was happening at LOG_IO level, which is intended for
watching data traffic in and out of the daemon rather than all the minutiae
of data analysis - that's LOG_DATA.
Also, LOG_DATA gets pushed down two levels. The effect is that -D 5
means exactly what it did, but for purposes other than driver debugging
-D 4 now suffices.
|
| |
|
| |
|
|
|
|
|
|
|
| |
Because it might be going to chrony. Or something else.
No logic changes, just name changes and one new derived #define,
TIMEHINT_ENABLE.
|
|
|
|
| |
We'll have to sneak up on this change more slowly.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
ntp_offset becomes time_offset. There is a new config symbol CHRONY_ENABLE;
most time service code is npw controlled by TIMESERVICE_ENABLE. The
file ntpshm.c becomes timeexport.c
This change is not complete. More disentanglement has to be done inside
timeexport.c itself; at the moment enabling one but not both of
ntpshm or chrony will probably break its compile. The point of getting
this commit out is so Gary will see the new baseline code ASAP.
All regression tests pass.
|
| |
|
| |
|
| |
|
|
|
|
| |
...rather than setting it ad-hoc on a per-driver basis.
|
|
|
|
|
|
|
|
| |
This is a large, ugly change. But without it we can't troubleshoot the
ICP/IP-source initialization bug properly - colliding definitions of
gpsd_report() were interfering with error reporting early in gpsd runs.
More cleanup work remains to be done, but at least this is working.
|
|
|
|
|
|
| |
...and be more explicit about operand sizes.
All regressuion tests [ass, code splints clean.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This awkward construct seems to have triggered some sort of toolchain
bug. The symptom is that gpsd_hexdump_level has different values depending
on whether we test it inside or outside of hex.c.
As a first step towards eliminating this global, we throw out a bunch
of hex logging that is duplicative or not really needed for production
drivers. Experimental drivers can call the hex dumper unconditionally.
The goal is to reduce the number of uses of this variable to eliminate it
withoud doing violence to the internal APIs. This is a first step.
All regression tests pass.
|
| |
|
|
|
|
|
| |
...and guard them with debug level so they won't be done when the output would
never be shipped anywhere.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Required so clients looking at flag masks in data exported via the shared-memory
interface will see the right thing.
These were separated originally in order to avoid pushing the reqyuired width
of the gpsdata.set flag mask over 32 bits. It became 64 bits in the Version 5
API change, so that constraint went away.
All regression tests pass.
|
|
|
|
| |
All regression tests pass.
|
|
|
|
| |
All regression tests pass, splint passes.
|
| |
|
|
|
|
|
| |
Replace it wuth what it calls, gpsd_write(). Requiresd only a prototype
tweak and some casts, no logic changes. All regression tests pass.
|
| |
|
| |
|
|
|
|
| |
All regression tests pass.
|
| |
|
| |
|
| |
|
|
|
|
|
| |
ckuethe believes, and the manual seems to confirm, that we don't need to ship
the device a link init because its default is to stream sentences.
|
|
|
|
|
|
| |
With most machines being 64-bit now, the assumption that "word" = 16 bits and
"long" = 32 bits is increasingly archaic. This commit removes it from our
naming conventions.
|
| |
|
| |
|
|
|
|
| |
Doesn't require htons(3) and wacky header magic.
|
| |
|
|
|
|
|
|
|
| |
So we can grep for things that need to be cleaned up without a
lot of noise.
All regression tests pass.
|
|
|
|
| |
after discussion with sanooj
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This solves the disappearing epx/epy bug on SiRFs, but it was actually a
systemic problem that affected several drivers. Navigation solution messages
were clearing DOPs, making it impossible for the error modeller to compute
estimates. New logic: Clear DOPs only when we get a skyview report. They'll
be regenerated by our visibility-matrix calculation when the skyview sentence
is analyzed.
If a sentence from the device supplies a DOP between the time the visibility
calculation is done and when the next fix is reported, it will override
our computed value. This might change later!
This change required a regression-test rebuild.
|
|
|
|
| |
superstar2_msg_navsol_ecef is no more. never was. go about your business
|
| |
|
|
|
|
| |
All regression tests pass.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
These calls move out of scattered places in the drivers into one spot
in libgpsd_core.c where they can be guarded with consistent validity
checks. Time offset is now computed by a new driver method,
ntp_offset(), which has access through the session structure to the
baud rate, the tag of the sentence last received, etc. If the
ntp_offset() method returns NAN, no notification is shipped.
The logic of these methods replicates the sentence and baud-rate
specific computations that were embedded in individual sentence
methods before.
All regression tests pass.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Large patch, no actual executable code changes except in three debug dumpers.
Breaks up the *_SET status macros so the client side continues to use them,
but the daemon uses a similarly-named set with an _IS suffix.
This frees up two mask bits in both sets - the client side no longer needs
to have REPORT_SET and CLEAR_SET bits, and the daemon side no longer needs
to have VERSION_SET and POLICY_SET.
The only actual code change is that the maskdump.c module, generated by
maskaudit.py, splits in half - one child now dumps client-side flags, the
other daemon-side flags. One other function call in a debug dumper in
libgps_core.c changes.
All regression tests pass.
|