Handling SIGINT in the freshly-started server leads to a race condition:
on Ctrl+C, Tuhi is killed before tuhi-live can call StopLive, leaving the
device in live mode. On the next start of tuhi-live nothing works because the
device will complain about invalid state.
Simplest solution here is to ignore SIGINT in the Tuhi instance and instead
rely on python's multiprocessing module to take us down when the main process
exits.
Fixes#206
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
There's just no point to that and if the tablet is currently in live mode, it
refuses the GET_WIDTH request anyway.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
If for some reason the device is still in live mode from an earlier
invocation, stop live mode first before we attempt to start a new one.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We have a dbus property for this so let's just use that as a backend for our
live property instead of emulating our own on top of it.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
0xb3 is the generic error code (or success, where applicable). Let's handle
those by default so that the rest of the messages only have to care about
replies with message-specific opcodes.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Add a new main() there so we have a single entry point for the gui instead of
a complex script that needs to be called. That start up script should really
just be the minimum bits possible.
It's still not perfect because we won't work without the gresources and they
require that the pkgdatadir is set. But at least it moves all the logic out of
the startup script.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We have implied inclusion orders here, specifically with gi. This works as
long as the files are included in the right order from other files but its not
generic enough - we really do need the gi.require_version() bit
everywhere to avoid issues.
Of course, once we do this flake8 complains about everything so let's
reshuffle things in an order it's happy with. This seems to be local modules
first, then ... whatever except when not?
application.py is excluded from this patch because it's about to get changed
in the next one and I'm too lazy to separate those out.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This sets up the config dir, loggers, etc. for us. Fixes regressions
introduced a while ago when Tuhi.run() was removed and then more obvious
breakage now that we require the Config path to be set up.
Fixes#206
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Since most of the signals are propagated by the bindings we largely just need
to hook up simple methods to log what is happening.
One nasty hack: we import some of the bits into the kete namespace so
tuhi-live continues to work. This can be fixed up in a follow-up commit.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The first argument in any signal is the object that emits it, so we don't have
to add it again.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Both of the signals here are connect to the device, not Tuhi. The only reason
this worked was because the signal sent by the dbus bindings had the device
twice in the arguments and 'tuhi' was actually the device.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This is used by the GUI and by kete and it's just an abstraction of the
handling anyway with little actual logic. Let's make this sharable.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This conflicts with trying to register the device on authentication errors and
making re-registration unreliable. Tuhi would keep connecting to the device
while holding the button down, so we'd get another failed connection and a
disconnect from the device. The Register signal would come in and get lost
somehow.
Fixes#195
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Check those modules that aren't part of the python installation and bail out
of meson where they're missing. This is technically wrong because we don't
need them at build-time and only at run-time but pragmatically sensible
because we waste too much time dealing with those bugs.
Fixes#200
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
flathub requires us to use specific versions, so let's do that here too so
we're consistent in testing.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
This was the one that introduced the Gtk.Template class that we use. And
better to have a meaningful exit than a crash.
Fixes#197
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
We used to drop all but the last 10 drawings on startup and as the comment
says so that users don't rely on tuhi for storage. That's mostly fine when
Tuhi is a separate daemon but now the primary interface is the GUI. So we can
run into this situation:
- user starts Tuhi (which in turn starts tuhi-gui and tuhi-server)
- tuhi-server downloads 20 fresh drawings from the device
- tuhi-gui crashes before it can fetch them from the tuhi-server
- restart Tuhi, the server now purges 10 out of those 20 drawings that the
user has never seen. They're gone for good.
Let's drop that code until we figure out a better solution to that (e.g. drop
drawings older than X months but that too is fraught with errors).
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Previously, the YAML output was basically:
- send: [0xb9, 0x01, 0x00]
- recv: [0xba, 0x02, 0x44, 0x00]
using a context manager where replies are expeced we can now group the
replies in the yaml file:
with logger as _:
self.send_nordic_command(...)
self.wait_nordic_data(...)
And that will prodice a grouped entry in the YAML file:
- send: [0xb9, 0x01, 0x00]
recv: [0xba, 0x02, 0x44, 0x00]
Which means processing those becomes a bit easier.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Both handlers need to be set up to log at debug level, the logger itself will
filter correctly anyway where verbose isn't given. And we want them both to
log in the same format, so let's re-use the formatter.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
The session id is a UUID, postfixed by an incrementing number. The UUID stays
the same for Tuhi's livetime, the number increments with every bluetooth
connect/disconnect. Result is e.g. 4e55c30035d043ce9f6f4914fb223820-2.
This session id is printed to the log file, added to the YAML raw logs and to
the JSON file, making it easier which log part resulted in which drawing, etc.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Reproduced on the Spark, first bytes in a new file recording were
- file header: 0x62, 0x38, 0x62, 0x74
- stroke point: 0xbf, 0xff, 0xff, 0x4a, 0x14, 0x29, 0x31, 0x6a
- stroke delta: 0xa8, 0x02, 0x04, 0xc3,
The initial point thus has a delta pressure value 0x6a which must be added to
*something*. And zero is the most sensible *something* that I can think of.
Fixes#111
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Deprecated, but still exists on the Slate so let's wrap the connect attempt
so we throw the right error.
Signed-off-by: Peter Hutterer <peter.hutterer@who-t.net>
Peter Hutterer