Unnamed repository; edit this file 'description' to name the repository. http://cgit.235523.xyz/tinyusb.git/atom/?h=etmtrace-rp2350 2026-08-26T10:06:07Z 2026-08-26T10:06:07Z hathach thach@tinyusb.org 2026-08-26T10:06:07Z urn:sha1:b396e322633984b08d2d92eb0551ae445817ea8f 2026-08-26T10:02:18Z hathach thach@tinyusb.org 2026-08-26T10:02:18Z urn:sha1:e9bd924396c8a5d441e50d2dba1b1057daf814a7 2026-08-26T09:59:35Z hathach thach@tinyusb.org 2026-08-26T09:59:35Z urn:sha1:61740be39db510928b1d724e3f22b4d2b115eaba 2026-08-26T09:58:24Z hathach thach@tinyusb.org 2026-08-26T04:21:33Z urn:sha1:e316a13d3fe0668812f5cc2d20a20c899e9f53cf With the 12 x n MHz rule gone (previous commit), nothing needs a non-stock clock: every board runs whatever pico-sdk's platform_defs.h selects (125 MHz rp2040, 150 MHz rp2350), set from crt0 by the SDK's own runtime_init_clocks(). The per-platform PLL table this file carried only existed to force other rates, so it goes too. Overriding is still possible the way the SDK documents it - pass SYS_CLK_KHZ together with PLL_SYS_VCO_FREQ_HZ / PLL_SYS_POSTDIV1 / PLL_SYS_POSTDIV2 from scripts/vcocalc.py - which is what the etm-trace notes now tell you to do for the higher trace rates. The removed function and its measured PLL values stay in git history if a board ever needs a pinned clock again. 2026-08-26T09:58:24Z hathach thach@tinyusb.org 2026-08-26T04:57:34Z urn:sha1:35c27a4637480161e5f89d314147942159bc108e Adds board pico2_etm_trace - a Pico 2 seated on the pico2 trace motherboard carrier (MIPI-20 Cortex Debug+ETM, 27 Ohm source-terminated, GND-guarded) - and the trace reference that goes with it. Firmware needs almost no trace support: J-Link's built-in RP2350 device script declares the off-ROM-table trace components (funnel/TPIU/ETM) and re-arms the whole chip-side path via OnTraceStart at every resume, so TRACE_ETM=1 does one thing only - keep the UART console TX-only when the console RX would sit on a trace pin. The carrier moves the console to GP12/13 (GP1 = TRACECLK), so it keeps a full-duplex console while tracing. Never set a custom JLinkScript here: it replaces the built-in one and pin trace stops working. Validated on the carrier at the rp2350 stock 150 MHz (75 MHz TRACECLK, width 4): cdc_msc enumeration burst 3/3, zero overflow, data sampling +1 ns (idle eye -1000..+2000 ps), 33M fetches per 10 s capture. Higher rates measured too: 180 MHz (90 MHz TRACECLK, +4..+5 ns) and 240 MHz (120 MHz TRACECLK, +3.5 ns) - the latter is the J-Trace PRO V2 ceiling, which is the probe at its documented limit rather than a board limit. board_init() no longer switches clk_sys on either chip, and boards run the pico-sdk default (125 MHz rp2040, 150 MHz rp2350). The 12 x n MHz PIO-USB rule that motivated the old forced clock is not real - it is the Pico-PIO-USB author's day-one design note (4d952cb, Dec 2021), never enforced by an assert, and contradicted by measurement: MSC dd over PIO-USB at ten clocks from 120 to 240 MHz on rp2350 gives 62 KB/s at every one; rp2040 CDC bulk loopback is byte-exact at 120/125/133/156/180 MHz; adafruit_fruit_jam passes 19/19 HIL at its stock 150 MHz including usbtest 30/30 and PIO-USB host+dual. CircuitPython ships PIO-USB at these same defaults. Two earlier fly-wire-era workarounds are dropped after A/B on the PCB: the TIMER0/1 DBGPAUSE clear (captures profile healthy firmware without it) and 12 mA fast-slew trace pads (default pads pass with a wider eye). A runtime clk_sys switch does still silently truncate a capture at the switch, which is why nothing sets the clock at runtime. Also included: docs/superpowers/followup/pr0000-msc-host-tur-retry.md - an unbounded Test Unit Ready / Request Sense retry in the MSC host driver, found while sweeping clocks and deferred to its own PR. 2026-08-26T06:41:05Z Ha Thach thach@tinyusb.org 2026-08-26T06:41:05Z urn:sha1:b15c720ab0a36ee16c1cf7984c7c0f4ccf3a7015 Three readability changes with no behaviour change on the healthy path — every pre-existing test passes untouched. main() was 368 lines with try/finally three deep, its two abort paths near-identical 40-line blocks; _abort_report holds that shape once, and the controller-hint cache and pool construction move to their own helpers. 368 -> 279, test_board 180 -> 151, test_device_usbtest 164 -> 125. test_hil_bounded.py cost 78s on every commit under test/hil/, mostly one 3s post-flash settle paid by ten tests against a fake rig. Now 37s. Fixes two pre-existing defects the extraction exposed: _write_failed_spec was unguarded inside the abort path, so an OSError there replaced the caller's RuntimeError and no report was written at all; _save_controller_hints overlaid a startup snapshot onto the re-read cache, clobbering a concurrent job's newer values. Also five comments that stated the opposite of the code, and both table renderers measuring width with len() against two-column status marks. 2026-08-25T08:04:42Z Ha Thach thach@tinyusb.org 2026-08-25T08:04:42Z urn:sha1:375e20090a0a60fd2d47ba6553d189e05482dfea hil_report.json and hil_report.md were written independently. Four writers produced the markdown and three wrote no JSON at all -- and those three are the paths where a run died, so a JSON consumer saw nothing exactly when it mattered: the per-board verdicts an agent hands back reported the whole fleet as "no report row" while a human read the real story from the markdown. Every writer now goes through render_report(), so a table can never contain something the JSON does not. The document gains `scope` (a three-board PR run and a full run that lost 24 boards were indistinguishable) and `caveat` (how the run ended). `banner` carries rig health across an --accumulate retry; `caveat` records how a run ended and must not -- conflating them made a clean retry publish an abandonment that never happened. helper/hil_report.py owns the document end to end, dissolving the import cycle that forced write_timeout_report to compose its own markdown and removing a duplicate cell classifier kept in sync by hand. hil_summary.py is deleted; its CLI moves there. hil_ci.sh uploads the sidecar so a remote --accumulate has a merge bas 2026-08-25T03:50:36Z Ha Thach thach@tinyusb.org 2026-08-25T03:50:36Z urn:sha1:60194f8d10ebc9ec1aeba559e359ebe3ca9b756b 2026-08-25T03:35:45Z Ha Thach thach@tinyusb.org 2026-08-25T03:35:45Z urn:sha1:b610ff039bafa1040c19d6a11cb04adcb22936e5 test_the_upload_board_can_diverge_from_the_built_board called get_family_boards without ci=True, so it pinned the developer's set, not the runner's: the CI skip lists move the one-first pick on three families. It held locally and went red on its first CI run. Pass ci=True, as _prune_buildable already does, and pin the runner's twelve. Rule 2 is a bare test/hil/ prefix, so the harness's own unit tests booked the full 27-board rig for diffs that cannot reach it. Carve test/hil/test/** out to rule 1b, beside test/{fuzz,unit-test}/**; the harness itself is untouched. A test pins that directory's file list, so anything added there that the rig does read fails rather than silently skipping hardware. Rule table updated in the spec and its carbon in the docstring. 2026-08-25T02:46:42Z Ha Thach thach@tinyusb.org 2026-08-25T02:46:42Z urn:sha1:da255b1d2db10b8f31332a779b2a526f579acee1 ci: an empty selection must build nothing, plus selector follow-ups A PR whose build axis legitimately selected nothing rebuilt everything. build.yml reads .build.families twice - as a |-joined regex, and implicitly as "is anything selected" - but tested only -z "$FAMILY_REGEX", which an empty list and a charset-rejected one both satisfy while meaning opposite things. ci_set_matrix had already returned the correct all-empty matrix; the fall-open branch discarded it. #3842 and #3840 each spent 74 cmake legs on it. Branch on the two cases instead, rename FAM_* to FAMILY_*, and cover the block with a test that extracts it from build.yml and executes it - it had no test at all, which is how this shipped through two merges. Follow-ups to the same machinery: glob.escape the repo root at five sites, so a checkout path containing [ or * stops failing closed; drop the ci-full label, read after the matrix was already computed and so never functional; delete 13 mcu:MKL25ZXX / mcu:SAME5X skip tokens matching no board; carry the rule table in the module docstring, guarded against drift; and pin six selection behaviours a mutation pass proved untested. Cut the selector's cost 1.8x (26.0s -> 14.6s) with 0 divergences over 260 paths, and stop scoping the membrowse upload by the PR example filter.