* feat: add debugging configurations for replay drive and LLDB attachment
* Add readme with video demo
* clean
* docs: update debugging safety documentation with demo link
* no need for mp4 then added on PR
* Update SConstruct
* bump opendbc
* updating readme
* updating readme
* updating readme
* is this better / worth it?
* final cleanups
* hacky. but does it work?
* Yep that worked!
---------
Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
* bump msgq
* add third_party/linux/include/media/msm_vidc.h
* add sde_rotator hw interface
* add msm_vidc hw decoder interface
* update SConscript to build qcom decoder and rotator
* use qcom decoder in replay framereader
* decode directly to NV12 with the correct stride without using the hw rotator
* bump msgq back to master
* don't compile rotator
* cleanup
* works now but much to simplify
* rm signals
* rm header
---------
Co-authored-by: Test User <test@example.com>
Co-authored-by: Adeeb Shihadeh <adeebshihadeh@gmail.com>
* torqued: apply latAccelOffset to torque control feed forward
* test learned latAccelOffset captures roll compensation bias on straight road driving, when the device is not flush in roll relative to the car
* test correct torqued latAccelOffset parameter convergence
* dont run by multiple threads at the same time!
* this doesn't work since we rely on is_connected
* Revert "this doesn't work since we rely on is_connected"
This reverts commit 7455b2fe83.
* use callback queue to make this thread safe and remove locks (which lag ui thread?)
* woah this works
* no more lock!
* always run signal handler and store callbacks, like qt
* debug
* more
* okay not for now
* combine _get_connections and _connection_by_ssid, closer to qt and not an explosion of GetSettings dbus calls
* debug
* try this
* skip
* len
* skip hidden networks
* actually slower
* stash
* back to 8929f37d49
* clean up
* stasj
* remove one of many classes
* clean up and fix
* clean up
* stash/draft: oh this is sick
* so epic
* some clean up
* what the fuck, it doesn't even use these
* more epic initializers + make it kind of work
* so simple, wonder if we should further 2x reduce line count
* i've never ever seen this pattern b4, rm
* remove bs add niceness
* minor organization
* set security type and support listing and rming conns
* forget and connect
* jeepney is actually pretty good, it's 2x faster to get wifi device (0.005s to 0.002s)
* temp
* do blocking add in worker thread
* add jeepney
* lets finish with python-dbus first then evaluate - revert jeepney
This reverts commit 7de04b11c2.
and
* safe wrap
* missing
* saved connections
* set rest of callbacks
* skip hidden APs, simplify _running
* add state management
* either wrong password or disconnected for now
* i can't believe we didn't check this...
* disable button if unsupported!!!
* hide/show event no lag hopefully yayay
* fix hide event
* remove old wifi manager
* cache wifi device path + some clean up
* more clean up
* more clean up
* temp disable blocking prime thread
* hackily get device path once
* ok
* debug
* fix open networks
* debug
* clean up
* all threads wait for device path, and user functions dont ever attempt to get, just skip
* same place
* helper
* Revert "helper"
This reverts commit e237d9a720.
* organize?
* Revert "organize?"
This reverts commit 3aca3e5d62.
* c word is a bad word
* rk monitor debug for now
* nothing crazy
* improve checkmark responsiveness
* when forgetting: this is correct, but feels unresponsive
* this feels good
* need these two to keep forgetting and activating responsive
* sort by connected, strength, then name
* handle non-critical race condition
* log more
* unused
* oh jubilee is sick you can block on signals!!
* proof of concept to see if works on device
whoiops
* so sucking fick
* ah this is not generic, it's a filter on the return vals
* flip around to not drop
* oh thank god
* fix
* stash
* atomic replace
* clean up
* add old to keep track of what's moved over
* these are already moved
* so much junk
* so much junk
* more
* tethering wasn't used so we can ignore that for now
* no params now
* rm duplicate imports
* not used anymore
* move get wifi device over to jeepney! ~no additional lines
* request scan w/ keepney
* get_conns
* _connection_by_ssid_jeepney is 2x faster (0.01 vs 0.02s)
* do forget and activate
* _update_networks matches!
* rm old update_networks
* replace connect_to_network, about same time (yes i removed thread call)
* no more python-dbus!k
* doesn't hurt
* AP.from_dbus: actually handle incorrect paths w/ jeep + more efficient single call
* properly handle errors
* it's jeepney now
* less state
* using the thread safe router passes a race condition test that conn failed!
* bad to copy from old wifimanager
* fix conn usage
* clean up
* curious if locks are lagging
* not for now
* Revert "curious if locks are lagging"
This reverts commit 085dd185b0.
* clean up _monitor_state
* remove tests
* clean up dataclasses
* sort
* lint: okay fine it can be non by virtue of exiting right at the perfect time
* some network clean up
* some wifi manager clean up
* this is handled
* stop can be called manually, from deleting wifimanager, or exiting python. some protection
* its not mutable anymore
* scan on enter
* clean up
* back
* lint
* catch dbus fail to connect
catch dbus fail to connect
* torqued: apply latAccelOffset to torque control feed forward
* test learned latAccelOffset captures roll compensation bias on straight road driving, when the device is not flush in roll relative to the car
* test correct torqued latAccelOffset parameter convergence
---------
Co-authored-by: felsager <d.felsager@gmail.com>
Shane Smiskol
DevTekVE
pencilpusher
Harald Schäfer
Trey Moen
commaci-public
YassineYousfi
Kacper Rączy
Adeeb Shihadeh
Maxime Desroches
ZwX1616
felsager
Jaume Balust
Comma Device