list of skill issues encountered this year, updated live, for future reference

[celery6]

• did not allocate any time for board peripheral bringup which shouldve been done using bare-metal stm32 cuz its stupid to debug rtos code + soldring issues simultaneously
• no time for practical coding workshops rip where onboarding
• underestimated how much time it takes to test on target... especially for those new to testing on embedded target
• FUCK .IOC CONFIG AND CUBEMX i did not learn any lessons from last year fml. configure everything before coding anything and never touch it again.
• upgrade freertos pls we are over 5 years out of date :Skullemoticon
• everyone apparently forgot things that don't happen instantaneously: harnessing, encoder shipping
• despite attempting to plan for like 5months, many things were missed. too abstract, not concrete enough :pensiewbwobbbleble: shouldve created CENTRALIZED doc or smth defining interfaces between things that may convert units, like imu->estimator, and proc->mcb.
• shouldve tested div by 0 earlier skull time crunch... except it cost more time debugging.
• very worth testing "trivial" case of hardcoding all inputs to 0 and checking no crashes. there were a lot of crashes.
• like the entirety of testflight integration.
• before flight, should OVERWRITE sd card when formatting instead of using "quick format" so that sd card is all zeroed out for post-analysis in case of corruption
• we decided against creating a global constants file because i thought constants belong in each controller/estimator unit. In hindsight it would probably be less error prone to put all constants in a .h file corresponding directly to model_params.m kms
• forgot to config peripherals in ioc better like i2c use fast mode, uart baud rate, etc
• didnt think too hard about imu saturation detection... apparently theres better methods like 95% of max, etc
• re-evaluate fatfs and use smth like journaling because our system has frequent unpredictable power-off events which fatfs is not designed to handle.
• get rid of cubeide by using... platformio?????????????
• avoid need to devcontainer, propose use wsl or smth else without shitty file Io on windows

[celery6]

im so going to regret over-focusing on the estimator implementation and lowkey neglecting everything else...

[SAT-oO]

to add on:

• consider automating unit test processes for the math implementations like MATLAB coder (dunno if this is sus), etc. This will safe so much more time