this fixes two issues:
The first issue that if we are missing a log file in the middle of the
list then it was not possible to download recent logs, as we get the
incorrect value for total number of logs. This happened for me with
107 logs, with log62 missing from the microSD. It would only show 45
available logs, so the most recent logs could not be downloaded.
The second issue is that get_num_logs() was very slow if there were a
lot of log files in a directory. This would cause EKF errors and ESC
resets. Using a opendir/readdir loop is much faster (approx 10x faster
in my testing with 107 logs on a MatekH743).
when nose-in or tail-in, if the aircraft has significant pitch
asymmetry in hover then we would spin around in no wind if we use
pitch as an input
this makes pitch input for nose-in and tail-in optional and off by
default to preserve existing behaviour
this zeros-watchdog was caused by a SPI DMA error on STM32F405:
https://discuss.ardupilot.org/t/crash-with-4-2-0-beta-and-4-3-0-daily-bdshot/83297
we had incorrectly left these internal errors enabled when asserts
were not enabled. That led to a osalSysHalt()
without these we get an spi_fail internal error, caught by the
SPIDevice code
we have seen errors where the BMI088 gets out of sync, so that the 3
axes are rotated. The data is shifted by 4 bytes, so that X=Z, Y=X
and Z=Y
this changes the BMI088 to "stop on full" mode, which is what Bosch
use in their example drivers, and also catches FIFO overrun events and
triggers a full FIFO reset. This should fix the problem with the FIFO
sync
this fixes the flash re-init problem when flash storage fills on
H7. It was caused by rejecting writes where one or more of the 32 byte
chunks was not all 0xff but was equal to the current data. That
happens when writing to the sector header in AP_FlashStorage
it also moves the interrupt disable inside the loop to allow for
other interrupts to run between blocks
this halves the number of flash writes needed, and makes flash storage
twice as space efficient on H7
On H7 we need to write 32 bytes at a time to flash, which corresponds
to 30 bytes of data in AP_FlashStorage. By using a 16 byte storage
line we don't waste as much space
we have now shown that interrupts being enabled during flash
operations can cause the infamous "68ms" bug, or watchdog when using a
32 bit timer on boards using flash for storage
The issue is quite repeatable with a load of a very large waypoint
file (over 500 waypoints) using "wp ftpload" in MAVProxy. This puts a
huge load on flash storage.
Our current working theory is that while doing flash writes for
storage on dual-bank we block access to only one bank, so if another
thread uses a timeout function with a short timeout while the flash
write is happening and chVTDoTickI calls code which crosses the flash
bank boundary then it can cause chVTDoTickI to violate the assumption
that no more than CH_CFG_ST_DELTA ticks pass while it is calculating
the value to set in the system timer. In that case we get a delay of a
full timer wrap, which is 68ms on boards with 16 bit timer and 70
minutes on boards with 32 bit timer
