| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
|\
| |
| |
| |
| | |
LuxoftSDL/hotfix/Fix_endpoints_data_copying_during_Policy_Table_Update
Fix endpoints data copying during Policy Table Update
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Up to now, when performing Policy Table Update, the module_config part
of the policy table was not copied. The change adds copying of this
part. The code is taken from CustomSDL/sdl_panasonic/develop (dd8ade70),
with just a small fix about copying exchange_after_x_ignition_cycles.
Fixes: APPLINK-20261
|
|/ |
|
|\
| |
| | |
Hotfix/remove redundant test main
|
| |
| |
| |
| |
| | |
Related:
APPLINK-20109
|
| |
| |
| |
| |
| |
| |
| |
| |
| | |
Add tests in 3rd-party-static/test/.
Add tests in tools/intergern/test/.
Add tests in tools/policy_table_validator/.
Related:
APPLINK-20109
|
|/
|
|
|
| |
Related:
APPLINK-20109
|
|\
| |
| |
| |
| | |
LuxoftSDL/hotfix/Data_race_in_ThreadedSocketConnection
Hotfix/data race in threaded socket connection
|
| |
| |
| |
| |
| |
| | |
Duplicated code moved to separate method
Renamed local variable frames_to_send to frames_to_send_local to
avoid confusion between framaes_to_send and frames_to_send_
|
| |
| |
| |
| |
| |
| | |
Added one more protection with mutex for usage of variable "frames_to_send_"
Relates: APPLINK-19408
|
| |
| |
| |
| |
| |
| | |
Edited indents according codinh style
Relates: APPLINK-19408
|
|/
|
|
|
|
| |
Added protection for usage of the frames_to_send_ with the mutex as it's done in the other places
Relates:APPLINK-19408
|
|\ |
|
| |\
| | |
| | |
| | |
| | | |
LuxoftSDL/hotfix/Change_description_of_greyOut_param_in_HMI_API_SDL_rel4_0
Changed description of parameter greyOut to correspond requirements
|
| | |
| | |
| | |
| | | |
Fixes: APPLINK-18998
|
| |\ \
| | | |
| | | | |
Hotfix/core dump if make master reset
|
| | | |
| | | |
| | | |
| | | |
| | | |
| | | | |
Increase of scope of lock in method Handle() in order to exclude problem with core crash.
Closes-bug: APPLINK-17367
|
| | |/
| | |
| | |
| | |
| | |
| | | |
Forbid processing message from mobile if Core begins stopping of work.
Closes-bug: APPLINK-17367
|
| |\ \
| | | |
| | | |
| | | |
| | | | |
LuxoftSDL/hotfix/SDL_writes_to_DB_incorrect_value_as_imageType_after_IGN_CYCLE
Add checking of SO field.
|
| | |/
| | |
| | |
| | |
| | |
| | |
| | | |
Add checking of SO field during creation of request to HMI.
If user calls to nonexistent field from SO, SO creates this field with default value.
Closes-bug: APPLINK-13006
|
| |\ \
| | |/
| |/|
| | |
| | | |
LuxoftSDL/hotfix/Fix_deadlocks_in_AM_and_log_socket_endless_blocking
Fix deadlocks in AM when stopping SDL and log socket endless blocking
|
| |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
When we stop AM we unregister apps and call StopSavePersistentDataTimer().
If the timer callback by chance is active at this moment we can have a
deadlock between AM and ResumeController. To prevent it limit the lifetime
of ApplicationListAccessor and allow ResumeController to finish his job.
Another option is to move resume_controller().OnSuspend() call before
unregitering apps.
Another deadlock problem arises when while destroying AM an app tries to
register (3 seconds before that actually). Then AM tries to destroy
ResumeController, but RC is in ApplicationResumptiOnTimer() callback
and tries to get AM instance, and blocks on AM singleton's lock. The
fix is to StopRestoreHmiLevelTimer() before destroyng AM.
A third reason for SDL not stopping normally is when a logger is writing
on a socket, but the client on the other side is not reading (and is not
closing the socket neither). Then write() blocks and logger, and therefore
SDL, cannot be stopped. The fix is to put timeout option to the socket.
|
|\ \
| | |
| | | |
Send policy app id during application registration
|
|/ /
| |
| |
| |
| |
| |
| |
| | |
There was the bug in notification parameter naming
the SmartObject has removed invalid parameter `policyAppId`
the proper name for the parameter is `policyAppID`.
Closes-Bug: [APPLINK-18660](https://adc.luxoft.com/jira/browse/APPLINK-18660)
|
|\ \
| |/
| |
| |
| | |
Conflicts:
README.md
|
| |\
| | |
| | |
| | |
| | | |
LuxoftSDL/hotfix/Fix_race_condition_in_Singleton_and_AppManager_failure_to_stop
Fix race condition in Singleton and Application Manager failure to stop
|
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When SDL is started, but the HMI is not, and we try to register a mobile
app (RegisterAppInterfaceRequest), then in this situation SDL can't be
stopped with ctrl+C - it enters in an endless cycle.
The reason for the problem is that we can't delete the AM, because the
RequestController thread of AM is still running (waiting in a cycle for
the HMI to respond). Also a second AM is created, because in the
Singleton we set to 0 the instance pointer before deleting it and
someone calls instance() before destroy() finishes, because there is no
common lock. The separate locks create a race condition.
The fix is to use a single mutex for the Singleton methods, introduce a
is_stopping_ flag in AM, set it to true in AM's Stop() method, and
also destroy RequestController's thread pool there.
Then check stop flag in RegisterAppInterfaceRequest::Run() and exit the
HMI waiting cycle there.
|
| |\ \
| | | |
| | | |
| | | |
| | | | |
LuxoftSDL/hotfix/OnVehicleData_send_to_app_that_unsubscribed_vehicleData
Fix OnVehidacleData sended to app with unsubscribed vehicleData
|
| | |/
| | |
| | |
| | |
| | |
| | |
| | | |
The vehicleData parameters in the notification massage (received
from hmi) are used for finding the subscribed applications i.e all
app. id`s are saved as well as every subscribed parameter. Then for
every app. specific notification is sent.
|
| |\ \
| | | |
| | | | |
Unit tests integration
|
| | | | |
|
| | | | |
|
| |\ \ \
| | |/ /
| |/| | |
Fixes core dump during snapshot generation.
|
| | |/
| | |
| | |
| | | |
Closes-bug: APPLINK-17257
|
| |/ |
|
| |\
| | |
| | | |
Fix crash when second instance of SDL is started
|
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When the app. is unabled to bind a socket return error, which caused
InitMessageSystem() to fail. The crash was occuring when exit is called.
Fixes: APPLINK-16303
|
| |\ \
| | |/
| |/| |
Fixes core dump while sending ResetGlobalProperties
|
| |/
| |
| |
| |
| |
| |
| | |
Due to response deletion by request controller there is possibility of
using object members being destructed.
Closes-bug: APPLINK-16029
|
| |\
| | |
| | | |
Fix crash when processing SystemRequest with QUERY_APPS
|
| | |
| | |
| | |
| | |
| | |
| | | |
Process tts_name as string, not as array. tts_name in QueryApps json file is "string" according to APPLINK-11731, tts_name in
UpdateApplist is "array of strings" according to API.xml
Fix: APPLINK-15783
|
| |\ \
| | | |
| | | |
| | | |
| | | | |
LuxoftSDL/hotfix/Fix_Loggers_TelnetAppender_port_reuse_on_fast_restart
Fix Logger's TelnetAppender socket reopen on fast restart
|
| | |/
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
When SDL is stopped, TelnetAppender's server socket listening on port
6676 will remain in TIME_WAIT state (if there was a client connected
to it) fr some time, depending on the operating system settings.
With the old code, if we restart SDL soon after stopping it,
TelnetApplender will fail to bind to its port, because 'it is in use'.
The fix is to set SO_REUSEADDR option to server socket when opening
it - this will allow the port to be reused when it is in TIME_WAIT
state.
|
| |\ \
| | | |
| | | |
| | | |
| | | | |
LuxoftSDL/hotfix/Remove_invalid_messages_from_MessageBrokers_buffer
Remove parsed JSON and invalid messages from MessageBroker's buffer
|
| | |/
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
With the previous code, messages that cannot be parsed (pure garbage,
invalid JSONs, JSONs that are not objects or does not have jsonrpc
member set correctly) remained in MesageBrocker's TcpServer receive
buffer for the corresponding file descriptor forever.
Because the garbage is in front of the buffer and next messages are
appended at the end, these next messages are not processed, since
parsing stops on first error.
Also if parsed message has different length than the original (this
is possible if there are floating point numbers for example), the
original message is not deleted properly and next message may be
broken.
The fix is to handle invalid messages and remove invalid or parsed
data from the buffer, without removing any potentially valid, but
incomplete data (masked websocket frames or SDL messages).
It is still possible to misinterpred garbage as a beginning of a
incomplete valid frame. In this case we are doing this: when we have
old data in the buffer and a new message arrives, if we can't parse
the buffer WITH the new message - try to parse only the new message
and clear the buffer on success, because if the new message can be
parsed alone, then the old must be garbage.
Fixed problem when deleting receiving buffer of disconnected client.
|
| |\ \
| | | |
| | | | |
Fixes setting of HMI level for consequtive permissions validation.
|
| | |/
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
HMI level must be set for just registred application since permissions
validation rely on HMI level.
Fixes: APPLINK-15635
Conflicts:
src/components/application_manager/src/commands/mobile/register_app_interface_request.cc
src/components/application_manager/src/commands/mobile/register_app_interface_response.cc
src/components/application_manager/src/resumption/resume_ctrl.cc
|
| |\ \
| | | |
| | | |
| | | |
| | | | |
LuxoftSDL/hotfix/Fix_crash_on_getting_connection_key
Fixes crash while trying to get app connection key
|