[Master] Diff All containing Profile/Mixer, not loading or corrupting some settings

[Jetrell]

The matter of DIFF/DUMP introduced settings errors, or locking up the compiler when loading, is not new, when the DIFF ALL has multiple PID and Battery Profiles applied.. This has been evident with F411's, since back in the 3.0 releases.

But this is also now experienced in 7.0, with F405's, since the addition of the Profile mixer switch. And now that the compiler has been upgraded for 8.0. Its even starting to effect F722 controllers when loading settings.

In my experience it does not occur every time. So it may not be seen by all users.. But for someone as myself who is often flashing fw and loading settings for testing purposes. Its has become far more noticeable. With the only workaround, being to load the DIFF ALL in two parts.

It would appear there is a fixed transfer rate between the Configurator and the flight control ? Due to it effecting processors in order of their clock speeds, with the introduction of more demanding background tasks being placed upon them.

[stronnag]

To add a bit more context, following some discussion with @Jetrell

• The hack in the configurator to use arbitrary delays is an inadequate solution, masking the real issue.
• The problem is serial overrun due to the configurator not "pacing" the delivery of data to that the FC can process.
• For example, profile X takes c. 1 second to return to a CLI prompt. The configurator can swamp the FC during this time.
• mwp's fcset tool does not use a delay, instead it awaits the FC sending the prompt "\n# " before sending the next line, so it can't overrun the FC's serial processing.
• @jetrell has reported that a further configurator hacked to increase the delays does alleviate the immediate issues.
• While increasing the extant delays from 50ms to 100ms for normal line delay and from 200ms to 1000ms for profile delay did alleviate problem, it would, in this reporter's view, be preferable to adopt an "await prompt" solution.
• The "await prompt" solution has the additional benefit of increasing throughput as it does not incur delay when it is not necessary.

[rts18]

I concur with your observations gentlemen. I too have experienced the matters. Both loading 2 or 3 profiles. And more so of late, since the GCC13.2 tool chain commit.

[stronnag]

Just for "fun", I instrumented mwptool's fcset to measure the time elapsed from writing a line command to receiving the new CLI prompt "\n# ". This is on a F405 bench test FC, recent INAV master (8.0), demonstrating that the current delays are an inadequate provision against overrun.

Index	Time(s)	Command
0001	1.207	defaults
0002	0.048	servo
0003	0.016	servo
0004	0.016	feature
0005	0.016	feature
0006	0.016	feature
0007	0.016	feature
0008	0.016	feature
0009	0.016	beeper
0010	0.016	blackbox
0011	0.016	blackbox
0012	0.016	blackbox
0013	0.016	blackbox
0014	0.016	blackbox
0015	0.016	blackbox
0016	0.016	blackbox
0017	0.016	blackbox
0018	0.016	blackbox
0019	0.016	blackbox
0020	0.016	blackbox
0021	0.016	blackbox
0022	0.016	blackbox
0023	0.016	serial
0024	0.016	serial
0025	0.016	serial
0026	0.016	serial
0027	0.016	serial
0028	0.016	serial
0029	0.016	serial
0030	0.016	aux
0031	0.016	aux
0032	0.016	aux
0033	0.016	aux
0034	0.016	aux
0035	0.016	aux
0036	0.016	aux
0037	0.016	aux
0038	0.016	aux
0039	0.016	aux
0040	0.016	aux
0041	0.016	set gyro_zero_x
0042	0.016	set gyro_zero_y
0043	0.032	set gyro_zero_z
0044	0.016	set ins_gravity_cmss
0045	0.016	set acc_hardware
0046	0.016	set acczero_x
0047	0.016	set acczero_y
0048	0.016	set acczero_z
0049	0.032	set accgain_x
0050	0.032	set accgain_y
0051	0.016	set accgain_z
0052	0.032	set mag_hardware
0053	0.032	set baro_hardware
0054	0.032	set serialrx_provider
0055	0.032	set rx_min_usec
0056	0.032	set blackbox_rate_denom
0057	0.032	set failsafe_procedure
0058	0.032	set align_board_roll
0059	0.032	set align_board_pitch
0060	0.064	set small_angle
0061	0.064	set applied_defaults
0062	0.064	set gps_provider
0063	0.064	set gps_sbas_mode
0064	0.064	set gps_ublox_use_glonass
0065	0.096	set nav_wp_radius
0066	0.096	set nav_wp_max_safe_distance
0067	0.112	set nav_rth_allow_landing
0068	0.112	set nav_rth_altitude
0069	0.112	set nav_rth_home_altitude
0070	0.112	set nav_fw_bank_angle
0071	0.112	set nav_fw_loiter_radius
0072	0.112	set fw_loiter_direction
0073	0.112	set nav_fw_launch_detect_time
0074	0.112	set nav_fw_launch_motor_delay
0075	0.112	set nav_fw_launch_max_altitude
0076	0.112	set nav_fw_launch_climb_angle
0077	0.112	set nav_fw_allow_manual_thr_increase
0078	0.112	set telemetry_halfduplex
0079	0.112	set smartport_fuel_unit
0080	0.160	set name
0081	0.016	mixer_profile 1
0082	0.048	set model_preview_type
0083	0.032	mmix
0084	0.016	mmix
0085	0.016	mmix
0086	0.016	smix
0087	0.016	smix
0088	0.016	smix
0089	0.016	smix
0090	0.016	smix
0091	1.266	mixer_profile 2
0092	0.016	profile 1
0093	0.080	set fw_p_pitch
0094	0.080	set fw_i_pitch
0095	0.080	set fw_ff_pitch
0096	0.064	set fw_p_roll
0097	0.080	set fw_i_roll
0098	0.080	set fw_ff_roll
0099	0.080	set fw_p_yaw
0100	0.064	set fw_i_yaw
0101	0.080	set fw_ff_yaw
0102	0.080	set max_angle_inclination_rll
0103	0.080	set max_angle_inclination_pit
0104	0.048	set tpa_rate
0105	0.064	set tpa_breakpoint
0106	0.048	set roll_rate
0107	0.064	set pitch_rate
0108	0.048	set manual_rc_expo
0109	1.264	profile 2
0110	1.260	profile 3
0111	0.016	battery_profile 1
0112	0.032	set bat_cells
0113	0.032	set vbat_min_cell_voltage
0114	0.048	set vbat_warning_cell_voltage
0115	0.048	set battery_capacity
0116	0.048	set battery_capacity_warning
0117	0.048	set battery_capacity_critical
0118	0.048	set throttle_idle
0119	0.048	set nav_fw_cruise_thr
0120	0.048	set nav_fw_max_thr
0121	0.064	set nav_fw_launch_thr
0122	1.257	battery_profile 2
0123	0.032	set bat_cells
0124	0.048	set vbat_min_cell_voltage
0125	0.048	set vbat_warning_cell_voltage
0126	0.048	set battery_capacity
0127	0.048	set battery_capacity_warning
0128	0.032	set battery_capacity_critical
0129	1.262	battery_profile 3
0130	1.266	mixer_profile 1
0131	1.264	profile 1
0132	1.267	battery_profile 2
0133	6.067	save

[rts18]

@DzikuVx Why was this closed? The issue @stronnag has described in the previous posts has not been resolved.

The problem is serial overrun due to the configurator not "pacing" the delivery of data to that the FC can process. For example, profile X takes c. 1 second to return to a CLI prompt. The configurator can swamp the FC during this time.

Using the latest master from both repositories and loading settings with more than one profile, the configurator still locks up. Requiring the 'diff all' to be painfully loaded in multiple parts.

[stronnag]

Reopening as multiple users report this issue, which does not appear to have been addressed.