Repository navigation
Follow-ups left open by the StopRxLoop-during-Init fix (#503) #504
Description
Left open when #503 (merged as 2c31471) landed. None blocks the fix.
-
src/IRadio.h: theStopRxLoopnote names a "window" and what is "out of scope" instead of stating the contract. Proposed wording in the PR's review thread: a stop issued whileInit()is still bringing the chip up is honoured (Init returns without starting its loop); a stop issued beforeInit()or before a standaloneStartRxLoop()is stale and cleared by them, which includes a stop that lands betweenInitWrite()and aStartRxLoop()on another thread (DEVOURER_TX_WITH_RX=thread). - The
g_devourer_should_stopcheck at the entry of everystart_rx_loopis redundant with the loop's own stop predicate and uncommented. Either drop it or say why it is there. - Hardware coverage: the reproducer (Init on a thread,
StopRxLoopat 100/400/1000/2500 ms, restart-after-stop, stale-stop-before-Init) passed 18/18 on 8822BU, 8812EU and 8832CU. Jaguar1, RTL8733B and MT7612U received the same guard but were reviewed by reading only; run the probe on them when they are on the bench. - The probe itself lives only in a session scratchpad. Promote it to
tests/(it is a third RX-lifecycle probe next tobusy_window_probe.cpp/kestrel_tbtt_probe.cpp) so the matrix is reproducible.
Activity
josephnef commentedon Oct 29, 2026
Three of the four items are on branch
pr503-followups(84a7458): the IRadio.h contract wording, the why-comment on the entry check in all six backends, and the probe promoted totests/stop_during_init_probe.cpp+tests/stop_during_init.sh(targetStopDuringInitProbe). Re-run through the promoted script at that commit: 18/18 OK on 8822BU (ch6), 8812EU and 8832CU (ch36). Still open: running it on the 8812AU, RTL8733B and MT7612U, which are not on the bench.StopRxLoop: state the contract on IRadio, promote the stop-during-Ini…
josephnef commentedon Oct 29, 2026
Landed in #505 (27f3066), with the Qodo round folded in: private mktemp log dir, pid-tracked cleanup, initstop now requires Init to still be running at the stop and a working receiver afterwards, README references the IRadio.h contract, and every documented command uses
sudo env. Remaining item: runtests/stop_during_init.shon the 8812AU, RTL8733B and MT7612U when they are plugged.josephnef commentedon Oct 29, 2026
The three remaining backends are plugged now and measured through
tests/stop_during_init.sh(branchstop-during-init-mt7612u, which adds theFW_DIRknob the MT7612U needs and reports a throwing bring-up asINIT_THREWinstead of a core dump):18/18 on each, verify loop seeing frames after every honoured stop. That closes the last item; all six backends are measured.
stop_during_init: MT7612U firmware dir, report a bring-up that throws (…
josephnef commentedon Oct 29, 2026
All four items landed: #505 (27f3066) and #506 (09c2ff3). The stop-during-Init matrix is 18/18 on all six backends.