fix(firmware): load STACKCHAN.RON via long-filename lookup

[andymai]

Summary

• STACKCHAN is a 9-character base name, which is not a valid FAT 8.3 short name. embedded-sdmmc's open_file_in_dir only accepts 8.3 names, so it rejected "STACKCHAN.RON" with FilenameError(NameTooLong) — which read_config masked as FileNotFound. Boot config silently never loaded, even with the file present on the card (stored as short name STACKC~1.RON + a long-filename entry, as any OS creates it).
• Resolve the short name by matching the long name via iterate_dir_lfn, then open by that short name. Preserves the documented /sd/STACKCHAN.RON path and existing cards.

Test plan

• just clippy-firmware (-D warnings) — clean.
• On-device (CoreS3, real STACKCHAN.RON on the card): flashed --release; boot log went from SD: read /sd/STACKCHAN.RON failed (FileNotFound) to net config: ssid=Interweb country=US hostname=stackchan.local — the 241-byte config now parses and applies. Clean boot, LCD renders.

Known follow-ups (same root cause, out of scope here)

• Writeback (PUT /settings, atomic STACKCHAN.NEW → STACKCHAN.RON): embedded-sdmmc cannot create long-named files, so writeback to the 9-char-base name is still impossible. Needs a separate decision (8.3 staging name, or rename).
• WAKE_WORD.tflite has the same long-name (and >3-char extension) issue on the read path.

Greptile: please review.

[greptile-apps]

Greptile Summary

This PR fixes the long-standing silent boot failure where STACKCHAN.RON (9-char base name) was rejected by open_file_in_dir with NameTooLong, masked as FileNotFound, causing config to never load even with the file present on the card.

• Adds iterate_dir_lfn in read_config to match the long filename case-insensitively and retrieve the FAT 8.3 alias (STACKC~1.RON), then opens the file by that alias — correct approach for embedded-sdmmc's API constraints.
• factory_reset still passes CONFIG_FILE (\"STACKCHAN.RON\") directly to delete_file_in_dir, which hits the same NameTooLong error and propagates StorageError::Write on every call — it needs the same LFN-resolve treatment.

Confidence Score: 3/5

The read-config fix is correct and on-device tested, but factory_reset always reports a write failure due to the same long-filename root cause.

The read_config fix is well-implemented and on-device verified. factory_reset passes the long filename directly to delete_file_in_dir, receives NameTooLong rather than NotFound, and unconditionally propagates StorageError::Write — meaning POST /factory-reset is broken on every call.

crates/stackchan-firmware/src/storage.rs — specifically the factory_reset method's use of CONFIG_FILE in delete_file_in_dir

Important Files Changed

Filename	Overview
crates/stackchan-firmware/src/storage.rs	Adds LFN directory scan in read_config to resolve STACKCHAN.RON's 9-char base name to its 8.3 alias before opening — correctly fixes the silent FileNotFound on boot. factory_reset still passes CONFIG_FILE directly to delete_file_in_dir, which returns NameTooLong (not NotFound) and always propagates StorageError::Write.

Comments Outside Diff (1)

1. crates/stackchan-firmware/src/storage.rs, line 766-789 (link)

   factory_reset always returns StorageError::Write due to long name

   delete_file_in_dir(CONFIG_FILE) receives "STACKCHAN.RON", which has the same 9-char base name that open_file_in_dir rejected before this fix. The driver returns NameTooLong before it even looks up the file on disk. Because the guard only suppresses embedded_sdmmc::Error::NotFound, a NameTooLong result sets first_error to Write unconditionally — so every POST /factory-reset call reports failure regardless of card state. factory_reset needs the same LFN-resolve-then-delete-by-short-name treatment applied to read_config in this PR.

   Prompt To Fix With AI

   This is a comment left during a code review.
   Path: crates/stackchan-firmware/src/storage.rs
   Line: 766-789
   
   Comment:
   **`factory_reset` always returns `StorageError::Write` due to long name**
   
   `delete_file_in_dir(CONFIG_FILE)` receives `"STACKCHAN.RON"`, which has the same 9-char base name that `open_file_in_dir` rejected before this fix. The driver returns `NameTooLong` before it even looks up the file on disk. Because the guard only suppresses `embedded_sdmmc::Error::NotFound`, a `NameTooLong` result sets `first_error` to `Write` unconditionally — so every `POST /factory-reset` call reports failure regardless of card state. `factory_reset` needs the same LFN-resolve-then-delete-by-short-name treatment applied to `read_config` in this PR.
   
   How can I resolve this? If you propose a fix, please make it concise.

   

Prompt To Fix All With AI

Fix the following 1 code review issue. Work through them one at a time, proposing concise fixes.

---

### Issue 1 of 1
crates/stackchan-firmware/src/storage.rs:766-789
**`factory_reset` always returns `StorageError::Write` due to long name**

`delete_file_in_dir(CONFIG_FILE)` receives `"STACKCHAN.RON"`, which has the same 9-char base name that `open_file_in_dir` rejected before this fix. The driver returns `NameTooLong` before it even looks up the file on disk. Because the guard only suppresses `embedded_sdmmc::Error::NotFound`, a `NameTooLong` result sets `first_error` to `Write` unconditionally — so every `POST /factory-reset` call reports failure regardless of card state. `factory_reset` needs the same LFN-resolve-then-delete-by-short-name treatment applied to `read_config` in this PR.

_{Reviews (1): Last reviewed commit: "fix(firmware): load STACKCHAN.RON via lo..." | Re-trigger Greptile}