F-33: Task stack headroom unmeasured #36

Severity: MEDIUM · Effort: Small · Batch: 5 · Status: Instrumented in #21 (29b54d2); the bench measurements and any resizing are still open

Found in the second full review of the firmware. Recorded with the rest of the pass (F-19…F-44) in docs/REVIEW_REMEDIATION_PLAN_2.md, which is where its status is kept.

Description

rotary_enc (3 KB) and withrottle_rx (4 KB) run the whole updateAllThrottles LVGL repaint
on their own stacks; the receive task also ran JmriJsonClient::connect from the PW callback
(see F-34). A too-small stack fails as corruption, not as an error.

Plan

Log uxTaskGetStackHighWaterMark for every application task on the bench under load, record
the results in THREADING_MODEL.md, and resize with a margin.

Acceptance Criteria

• CONFIG_THROTTLE_STACK_REPORT (menuconfig → Diagnostics, off by default) logs every
  application task's high-water mark every 10 s, and the one-shot tasks log theirs as
  they exit. Compiled out when off; the option builds.
• THREADING_MODEL.md has a table for the measurements and the procedure.
• Bench: build with the option on, drive under load, fill in the table, and resize any
  task left with less than about a quarter of its stack. Nothing was resized blind.