Persist bank + sample state and calibration data to flash across restarts (ectocore only)

[Copilot]

• Explore repository structure and understand the codebase
• Identify key files: main.c, globals.h, flashmem.h, sdcard_startup.h
• Understand that sel_bank_cur and sel_sample_cur are the current bank/sample
• Understand that playback_stopped flag tracks playback state
• Understand that flashmem.h already provides flash read/write functions
• Create a persistent storage structure for bank+sample state
• Add code to save bank+sample when playback stops
• Add code to restore bank+sample on startup
• Add validation checks for corrupted data
• Mark flash access functions with __not_in_flash_func() to ensure they run from RAM
• Fix FLASH_SECTOR_SIZE constant references
• Mark PersistentState_load with __not_in_flash_func() for safety
• Complete code review and address issues
• Refactor to combine calibration and bank/sample data structures
• Fix file loading issue: Trigger file change after savefile_do_load
• Synchronize sel_bank_next and sel_sample_next with restored values
• Restrict feature to ectocore only
• Fix ectocore loop to preserve restored bank state
  • Changed initialization of sel_bank_next_new from 0 to sel_bank_cur
  • Changed initialization of sel_sample_next_new from 0 to sel_sample_cur
  • These local variables in the ectocore loop now start with current state instead of always 0
  • Prevents the loop from immediately resetting bank to 0 when checking for changes

Root Cause:

The ectocore input handling loop has local variables sel_bank_next_new and sel_sample_next_new that were always initialized to 0. When the loop checked if the bank changed (sel_bank_cur != sel_bank_next_new), it would always detect a "change" from the restored bank (e.g., 1) to 0, then set sel_bank_next = 0, effectively resetting the bank.

Solution:

Initialize these local tracking variables with the current bank/sample values so they start in sync with the global state, allowing proper change detection.

Original prompt

---

This section details on the original issue you should resolve

<issue_title>Save bank + sample and restore on load</issue_title>
<issue_description> I'd like to add a new feature to the ectocore. Whenever playback is STOPPED (zeptocore or ectocore), save
the current bank and sample so that when it starts up again it starts from that bank+sample. There are two
choices for saving - internal flash or sd card. I'd try going with the internal flash first, which is fine.
It just means that when a new firmware is uploaded the dat amight be corrupted so there should be some basic
checks to make sure the data is not corrupted and that the bank/sample do exist before using them on startup.

</issue_description>

Comments on the Issue (you are @copilot in this section)

• Fixes Save bank + sample and restore on load #783

---

💬 We'd love your input! Share your thoughts on Copilot coding agent in our 2 minute survey.

[schollz]

@copilot A problem: The persistent save is corrupting the save data from the calibration. Probably better to combine the data structures for calibration + the bank/sample to load from because both need to be loaded on startup. You may need to refactor the calibration/saving loading so it works with your PersistentStorage

[Copilot]

@copilot A problem: The persistent save is corrupting the save data from the calibration. Probably better to combine the data structures for calibration + the bank/sample to load from because both nee...

Fixed in commit 750b27e.

I've refactored the code to use a unified PersistentState structure that combines both calibration data and bank/sample data in a single flash storage location. Key changes:

1. Expanded PersistentState: Now includes center_calibration[8] array alongside bank/sample fields
2. Mutual preservation:
   • PersistentState_save() (bank/sample) reads existing data first to preserve calibration
   • PersistentState_save_calibration() reads existing data first to preserve bank/sample
3. Replaced EctocoreFlash: Removed the separate structure and updated ectocore.h to use the new unified functions
4. Updated checksum: Now includes calibration data for complete validation

Both features now share the same flash sector without conflicts.

[schollz]

@copilot Perfect! It is actually getting the stored bank/sample information. However it is NOT loading the new bank/file on startup. Here's the startup debug info:

sdcard_startup] checking settings/grimoire/rune7
[sdcard_startup] bank 0 has 2 samples
[sdcard_startup] bank 1 has 2 samples
[PersistentState] Successfully loaded bank=1 sample=1 fh
[sdcard_startup] Restoring to bank 1, sample 1 from flah
memory usage: 64.4% (136580/212116)
[FV_Reverb_malloc] heap_size: 5484 (0 combs, 3 allpasse)
[FV_Reverb_malloc] heap_size: 9976 (1 combs, 3 allpasse)
[FV_Reverb_malloc] heap_size: 16232 (2 combs, 3 allpass)
[FV_Reverb_malloc] heap_size: 21012 (3 combs, 3 allpass)
[FV_Reverb_malloc] heap_size: 27004 (4 combs, 3 allpass)
[FV_Reverb_malloc] heap_size: 32456 (5 combs, 3 allpass)
[FV_Reverb_malloc] heap_size: 37592 (6 combs, 3 allpass)
[FV_Reverb_malloc] num_combs: 4, num_allpasses: 3
memory usage: 77.2% (163668/212116)
updating bpm timer: 185-> 170 (172)
[ectocore] startup
[PersistentState] Successfully loaded calibration from h
[ectocore] calibration data loaded from flash
[ectocore] calibrate 7=511,[ectocore] calibrate 6=570,[,
CTRL-A Z for help | 115200 8N1 | NOR | Minicom 2.9 | V

It says its loading, but the actual file is not being loaded for playback.

[Copilot]

@copilot Perfect! It is actually getting the stored bank/sample information. However it is NOT loading the new bank/file on startup. Here's the startup debug info:

sdcard_startup] checking settings...

Fixed in commit 4027bda.

The issue was that the restored bank/sample values weren't triggering the file change mechanism. The code was setting sel_bank_cur and sel_sample_cur correctly, but the file loading system uses sel_bank_next/sel_sample_next and fil_current_change to trigger changes.

Now after savefile_do_load() completes, if no savefile was loaded from SD card, the code triggers a file change with the flash-restored values. This gives savefiles precedence over flash restoration, which makes sense since savefiles are explicit user configurations.

[schollz]

@copilot Perfect! It is now loading the sample correclty! But one more thing - it is not setting the current bank such that when the ectocore "sample" knob is turned, it switches to a sample from bank 0, not from the saved bank which was set (which in this case was bank 1)

[Copilot]

@copilot Perfect! It is now loading the sample correclty! But one more thing - it is not setting the current bank such that when the ectocore "sample" knob is turned, it switches to a sample from bank...

Fixed in commit d3cc4d4.

The issue was that only sel_bank_cur and sel_sample_cur were being set during restoration, while sel_bank_next and sel_sample_next remained at their initialized value of 0. This caused an inconsistency where the sample knob code (which uses banks[sel_bank_cur]->num_samples) would see the correct bank, but other parts of the system checking the _next variables would see bank 0.

Now all four variables are synchronized when restoring from flash, ensuring consistent bank selection throughout the system.

[schollz]

@copilot One more thing, make this feature available to the ectocore only for now

[Copilot]

@copilot One more thing, make this feature available to the ectocore only for now

Done in commit 6104976.

Added #ifdef INCLUDE_ECTOCORE guards around all bank+sample persistence code in main.c and sdcard_startup.h. The feature now only compiles and runs for ectocore builds, not zeptocore.

[schollz]

@copilot Hm. There is still an issue, where it loads the bank/sample correctly and plays it from the right file and All that is perfect. However, when changing the sample knob it reverts to bank 0 instead of staying on the bank loaded. Somehow the bank loading persists to load the sample, but not persisting into the main ectocore loop when sample knob is changed.

[Copilot]

@copilot Hm. There is still an issue, where it loads the bank/sample correctly and plays it from the right file and All that is perfect. However, when changing the sample knob it reverts to bank 0 ins...

Fixed in commit 1c41960.

The issue was in the ectocore input handling loop. The local variables sel_bank_next_new and sel_sample_next_new were always initialized to 0 at the start of the loop. When the loop checked for bank changes (sel_bank_cur != sel_bank_next_new), it detected a false "change" from the restored bank (e.g., 1) to 0, then set sel_bank_next = 0, resetting the bank.

Now these variables are initialized with the current bank/sample values (sel_bank_cur and sel_sample_cur), so they start in sync with the global state and only detect actual changes.