[gpsd-dev] New gpsmon Problem

[Fred Wright]

Just a heads-up that I'm looking into a recently introduced problem with
gpsmon on some platforms.  I've tracked it down to a conspiracy among at
least two bugs in gpsmon itself and the *removal* of a bug in the
SConstruct script, as follows:

1) There's code in SConstruct to force pps=no when neither RFC2783 nor
TIOCMIWAIT support is available.  But, prior to 1b0eb63a, this code was
*after* PPS_ENABLE was written to gpsd_config.h, so it wasn't really
effective (which explains why I've seen varous pointless PPS-related log
messages on my PPS-incapable Mac).  Now that the code has been fixed to
have the intended effect, systems forcing pps=no no longer incorrectly
define PPS_ENABLE.

2) There's host vs. target confusion in gpsmon.  Since it can report data
from remote gpsds, the lack of PPS support on the *host* shouldn't prevent
it from reporting remote PPS info.  But currently, PPS_ENABLE is required
just to *report* PPS info, regardless of source.

3) When gpsmon's reporting of PPS info is disabled, the display handling
is screwed up, in a driver-dependent fashion.

At this point, #1 and #2 are fully understood, but #3 needs further
investigation.  I expect that fixing #2 would go back to sweeping #3 under
the rug, but it's probably worth at least some effort to track it down.

This isn't observable in the default Linux build, but it is if you build
with pps=no.

Fred Wright