OTA hardening follow-up to the patch_nimble.py hardcoded-"tdeck" bug that broke
tdeck-ota linking (undefined reference to nimble_host_reset_reason):
- Add _build_helpers.env_libdeps_dir(env, *parts): the single per-environment
libdeps path resolver (.pio/libdeps/<PIOENV>/...). Converted all five pre-scripts
(patch_nimble/msgpack/filestore/littlefs_paths, sync_file_libdeps) to use it, so
the env component can no longer be hand-rolled/hardcoded wrong per script.
- "OTA: Ready" log -> "OTA: wireless flash service started (pyxis-tdeck:3232)":
ArduinoOTA.begin() is void and can't confirm a ready state, so log the target
instead of claiming readiness we can't verify.
Verified: both tdeck and tdeck-ota build, and the NimBLE patch lands in EACH env's
own libdeps tree (nimble_host_reset_reason count 2 in both).
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01UWZuYkHBRqNb6BZHV8sTG5
patch_msgpack.py + patch_filestore.py: read PIOENV instead of
hardcoding "tdeck" in the libdeps path. The hardcode meant the
msgpack public-modifier patch silently no-op'd under tdeck-bluedroid
(and tdeck-ota), making microLXMF's packRawBytes / raw_data /
indices accesses fail to compile. Mirrors the pattern already used
in sync_file_libdeps.py. Resolves the failing tdeck-bluedroid build.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
Five build-config changes that unlock proper microReticulum behavior
on the T-Deck Plus:
1. PSRAM allocator (`-DRNS_DEFAULT_ALLOCATOR=RNS_PSRAM_ALLOCATOR`,
`-DRNS_CONTAINER_ALLOCATOR=RNS_PSRAM_POOL_ALLOCATOR`,
`-DRNS_PSRAM_POOL_BUFFER_SIZE=2048000`). Previously default-heap
on ESP32, which routed every microReticulum allocation (path
table, destinations, etc.) through internal SRAM. Under live
announce flood the internal heap dropped from 92KB → 40KB free,
max_block fragmented from 77KB → 19KB. With PSRAM allocator the
internal heap stays rock-steady at 137KB and microReticulum lives
in 2MB of dedicated TLSF pool in PSRAM.
2. LittleFS instead of SPIFFS (`-DUSTORE_USE_LITTLEFS`). microStore's
FileStore puts (path table, etc.) hit the filesystem several
times per second on a busy network. SPIFFS GC stalls for 100s of
ms during block erase, causing flush_buffer() to fail and every
put to bail silently. LittleFS handles sustained writes cleanly.
3. Filesystem-backed path persistence (`-DRNS_USE_FS`,
`-DRNS_PERSIST_PATHS`). Without these flags, Transport::start()
skips _path_store.init() entirely and every announce-driven put
fails at TypedStore::isValid() — surfacing as "Failed to add
destination to path table" spam (~3.4/sec) and an empty UI
announce list.
4. Switch lib_ldf_mode to chain+ (was deep+) — deep+ scans every
#include statement and auto-fetches matching libs from the
registry, which was pulling a parallel microReticulum copy and
bypassing our deps/microReticulum/ overlay. chain+ follows only
explicit dependencies declared in library.json files.
5. patch_msgpack.py PIO pre-script — promotes hideakitai/MsgPack's
`Packer::packRawBytes` and `Unpacker::indices` from private to
public so microLXMF can splice arbitrary msgpack values into
LXMessage's fields-map wire format. Mirrors the equivalent patch
in microLXMF/conformance-bridge/CMakeLists.txt:106-125.
Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
torlando-agent[bot]