dangernoodle-io / breadboard / 24943248230

100%

push

github

fix(ota_pull): streaming JSON scanner + chunked early-exit read (#84)

GitHub releases/latest JSON has grown past the prior 16 KB read buffer
(currently ~22 KB and trending up with each release/board added). cJSON
refused to parse the truncated input, returning -1 every time, leaving
devices unable to OTA on stable releases.

Replace cJSON_Parse with a targeted, escape-aware scanner over the response
that extracts only `tag_name` and the matching asset's `browser_download_url`.
Read the response in 2 KB chunks and re-parse after each, exiting as soon
as both fields are captured — typically well before any growth-prone fields
(body, large asset lists) are pulled. API_BUF_MAX kept as a 32 KB safety cap
but rarely reached.

Other improvements bundled with the parser overhaul:
- 5% progress milestones (was 10%) for finer visibility into the long tail
- explicit "OTA download complete, finalizing" + "esp_https_ota_finish took
  N ms" lines so finalize timing is observable
- diagnostic on parse failure logs board name, total bytes, head + tail
  snippets — debug aid for future regressions
- http timeout_ms 60s -> 20s so a stalled socket errors out cleanly into the
  perform loop rather than getting clipped by a tighter task WDT bound

3 new host tests cover the cases that justify the scanner over cJSON:
body field with embedded "tag_name" lookalikes, body with escaped quotes
and unicode escapes appearing after assets, and matching asset preceded by
multiple non-matching ones.

Tests: 159/159 passed; coverage 100% (130/130 lines, 12/12 functions, 81/81
branches).

81 of 81 branches covered (100.0%)

Branch coverage included in aggregate %.

113 of 113 relevant lines covered (100.0%)

278.07 hits per line