Teemu Ikonen authored 1 year ago and Clayton Craft committed 1 year ago

Modemmanager versions >= 1.23.2 allow GPS to be used also when the modem
is not enabled (i.e. does not have a SIM). Remove the modem enabled
check from the mm driver so that SIMless modems work with gnss-share.

Teemu Ikonen authored 1 year ago and Clayton Craft committed 1 year ago

The driver is called 'mm' in config.

Requires the dbus module.

[ci:skip-build] already built successfully in CI

Note that a successful pause() isn't necessary and is not fatal, we
still want it to call close anyways (even though that might end up
failing too I guess... but worth trying anyways!)

fixes #15
fixes #17

Apparently this doesn't work with g_socket_client_connect_to_host_async
(used by geoclue), it's unable to connect to the socket if it doesn't
have write access...

This reverts commit a4d9e5d4.

Crude implementation of RO-access to the socket. The "group" config
option will be able to read from, but not write to, the socket

For Debian.

This was added in Go 1.16, but we need to target 1.15 since that's the
latest version in...<drumroll> Debian.

fixes a goroutine leak

This is basically unneeded now that gnss-share can accept writes to its
socket.

Using time as a uuid is problematic when we want to generate multiple
topics, which is something coming later.

Copy should be made at the source, else there's a race condition between
readines and the anonymous func.

This wasn't accounting for GPS leap seconds, and could even set an
entirely wrong time if system time was incorrect. External tools could
set the time before gnss-share starts, or could do it through gnss-share
when it supports writing to the socket

readlines can also grab a lock, but since it's run as a new goroutine I
don't think this would deadlock. Anyways, this makes the block between
the lock/unlock hopefully slightly easier to read/understand.

That seems to be the only one that can get hung, since /dev/gnss*
devices can be powered off when not used, so they wouldn't be hung
anyways.

This was here somewhere before the last refactor.. oops

Fixes an issue where subscribe() fails, and the connection is kept open
forever.

This should help with debug later, since silently ignoring some of these
can make debugging harder

Note, a change in "internal/nmea/nmea_test.go" was also suggested, but
ignored since the ordering of objects in the test struct is used when
initializing the struct.

/home/clayton/src/gnss-share/internal/config/config.go:13:13: struct with 80 pointer bytes could be 72
/home/clayton/src/gnss-share/internal/pubsub/pubsub.go:25:13: struct with 40 pointer bytes could be 8
/home/clayton/src/gnss-share/internal/pubsub/pubsub.go:31:19: struct with 32 pointer bytes could be 24
/home/clayton/src/gnss-share/internal/gnss/gnss.go:26:11: struct with 56 pointer bytes could be 40
/home/clayton/src/gnss-share/internal/gnss/stm.go:30:16: struct with 88 pointer bytes could be 48
/home/clayton/src/gnss-share/internal/server/server.go:19:13: struct with 48 pointer bytes could be 40

No longer used

No longer used

The idea is that device output should not be modified by the function
that brings it into the app. the save* functions and relevant printf
calls have been updated to exclude spaces from output that uses the data
from the device so that the file generated can be imported again
(preserving newlines here was problematic) and that messages to console
don't have extra newlines.

No longer used