Unnamed repository; edit this file 'description' to name the repository. http://cgit.235523.xyz/tinyusb.git/atom/?h=claude%2Fadd-systemview-debug 2026-09-03T22:11:02Z 2026-09-03T22:11:02Z hathach thach@tinyusb.org 2026-09-03T22:11:02Z urn:sha1:f9c0c1239250399bc41347b9160bd95721f86dca - sysview_ci.py: wait() after kill() on the OpenOCD/nc grace-period timeout, so a still-exiting process can't hold the probe when the recovery flash starts - SKILL.md: point the OpenOCD and post-mortem recipes at examples/device/cdc_msc_freertos, where build-sv actually lands - stm32f401blackpill: override SYSVIEW_BUFFER_SIZE_DEFAULT to 4096; the 64 KiB part overflows RAM by 12952 bytes at the family's 65536 default under SYSVIEW=4 - get_deps.py: align the SystemView entry's continuation lines 2026-09-03T21:55:33Z hathach thach@tinyusb.org 2026-09-03T21:55:33Z urn:sha1:78ec7f6abae94f59612f30cda4d24031f5e7b760 Third-party actions referenced by mutable tag are a supply-chain risk; pin to the v11 commit SHA per the repo's existing action-pinning convention. 2026-09-03T21:21:57Z hathach thach@tinyusb.org 2026-08-11T10:08:20Z urn:sha1:4df375487cd49312fb55e8a1908bcd4bb544e79a The HIL job captures the sysview-flagged boards after the functional tests; a compare job turns base-vs-PR reports into sysview-comment, and pr_comment.yml appends it to the HIL sticky comment. Every job is capped above the pool guard. 2026-09-03T21:21:57Z hathach thach@tinyusb.org 2026-09-03T10:14:39Z urn:sha1:9e681141bc24ff71081dfce683ae4371e7520124 sysview_ci.py builds, flashes and captures the sysview-flagged boards (stm32f407disco and raspberry_pi_pico in tinyusb.json) on the rig, taking each board's flock via hil_lock and restoring park firmware afterwards. Hard-won correctness baked in from the dogfood runs: firmware paths carry the flasher's extension via FLASHER_SUFFIX (JLinkExe and OpenOCD infer the image format from it, and the RTT control block still comes from the ELF, which is not always the flashed image); WCH boards are detected by their wch-riscv target config (is_wch_board(), the plain "openocd" flasher name since #3804 folded openocd_wch/openocd_adi away); build and flash failures are excerpted from BOTH ends (cmake prints the diagnosis first and the call stack last, ninja the reverse) and flash failures include the flasher's own output so rc=124 can distinguish a wedged probe from a dead target; a bad ELF no longer aborts the whole multi-board run; the enumeration wait sits after the capture session's own reset; each capture picks a free RTT server port. session_resets()/sysview.attach_only skips the in-session reset for boards that never come back from it (metro_m4_express: both campaign captures died with 'No control block found' while attach-without-reset streamed immediately) - the flasher's post-flash reset already supplied the fresh boot. The suites join the pre-commit hil-test hook and master's test/hil/test/ carve-out pin. 2026-09-03T21:21:57Z hathach thach@tinyusb.org 2026-08-11T10:08:05Z urn:sha1:dcb3fcfd30ac03b810d6de82674c008d5163400f The v2 post-mortem design and plan, the HIL performance-report design and plan, and the promotion criteria for when a debugging technique earns its own skill (added to the agents-workflows design). 2026-09-03T21:21:57Z hathach thach@tinyusb.org 2026-08-11T10:08:05Z urn:sha1:7474173b075ae656ed16dea5dcb3f830349f3450 RTOS/ISR scheduling, CPU load and latency-profile questions go to sysview first; target-debug keeps the failure-hunting techniques. CLAUDE.md points at the new skill. 2026-09-03T21:21:57Z hathach thach@tinyusb.org 2026-09-03T10:14:39Z urn:sha1:463ff1f3d2d97b53659364d7c4cde99693688100 SKILL.md leads with the capture routes ranked by evidence: OpenOCD RTT streaming through the rtt skill's tools/rtt.py (--channel 1 --reset-before-attach, since the Init record that carries the timestamp frequency and sync preamble is emitted once at boot; 238 KB captured without it decodes to zero events), the live GUI recorder under Xvfb for depth, and --from-raw post-mortem decode last. boards.md holds the per-board reference: probe routes, RAM bases, measured buffer sizes, the full-pool campaign table and the honest capture status of every board tried, including the gaps: nrf5340dk HardFaults in vTaskStartScheduler before any task runs; PIO-USB host ISRs are invisible to level-1 coverage (fruit_jam); metro_m4_express is attach-only (the SAMD5x DSU CPU Reset Extension holds the core after an in-session reset run); WCH is dump-only - the QingKe SDI has no autonomous memory access, so a persistent debug session corrupts the running core ~1.9 s into USB traffic, ruled out against lock, post-mortem mode and firmware; a stray TU_ASSERT still halts for inspection rather than being silenced. sysview_record.py drives the capture end to end and refuses to back up its own ini stub as the user's SystemView config (a SIGKILLed run had consumed the real backup that way). sysview_report.py turns the exported CSV into the JSON report: ISR load, function pairing and durations, workload-window CPU load from the scheduling events. Returns are classified on the bare 'Returns' detail - SystemView annotates a duration on only ~0.5% of them (measured on stm32f407disco and raspberry_pi_pico), and when the annotation is absent the duration comes from the recorded timestamps, verified the same quantity at 167.998 vs a declared 168.000 ticks/us over 264 annotated returns. Both loss markers count as overflow: explicit '*** Overflow ***' rows and 'Returns to *** OVERFLOW ***' exit contexts - only counting the first understated real loss by two orders of magnitude. 2026-09-03T21:21:57Z hathach thach@tinyusb.org 2026-08-11T10:07:44Z urn:sha1:d2de029fd9def460d66e87fd1088cc49ed0f4574 The dual-role dogfood measured the 4096 fallback losing 96.7% of ISR exit contexts on mimxrt1064_evk and 99.1% on metro_m4_express; a local-board run measured stm32u575nucleo losing ~19% of the whole stream under a 12 s cdc_msc workload. The default was too small for any real capture on fast parts and every user had to rediscover that per board. Families now declare SYSVIEW_BUFFER_SIZE_DEFAULT in family.cmake (the SYSVIEW_RAM_BASE_DEFAULT pattern): imxrt, nrf, rp2040, samd5x_e5x, stm32f4, stm32f7, stm32h7 and stm32u5 set 65536, the measured-safe value. -DSYSVIEW_BUFFER_SIZE still beats everything; small-RAM parts keep the 4096 fallback and their per-board overrides (stm32f072disco's 2048 verified still honoured). lpc55 measured 34 lost events on the same route, an order of magnitude lighter, so it keeps the fallback rather than inheriting a default on thin evidence. 2026-09-03T21:21:57Z hathach thach@tinyusb.org 2026-08-11T10:07:44Z urn:sha1:95ec96b2962ebe57a43902495fd76e40dfb5558d One sysview_rtt_conf_wch.h serves every WCH family instead of a per-family copy; ch583 keeps its own RTT lock header because its interrupt controller needs a different masking sequence than the ch32v parts. 2026-09-03T21:21:57Z hathach thach@tinyusb.org 2026-09-03T10:14:28Z urn:sha1:5a0d707dd96ae7821d97793b441b86aabb684855 Cortex-M0/M0+, the WCH RISC-V parts and MAX3266x have no usable DWT cycle counter, so their family.c provides SEGGER_SYSVIEW_X_GetTimestamp() from a free-running hardware timer: a 32-bit general-purpose timer on ST parts (sysview_stm32_tim2.h), SysTick-extended sources on the M0 families, the core's SysTick-equivalent on WCH with the documented ordered read of the split count registers (sysview_wch_timestamp.h), and a TMR on MAX3266x (sysview_max32_tmr.h). max32666fthr's DWT implements only watchpoint comparators (DWT_CTRL.NOCYCCNT=1 read live; UG6971 documents no DWT/trace at all), so SystemView's ARMv7-M default read 0xE0001004 as a frozen 0 and every duration decoded as zero. The TMR runs at 48 MHz (f_PCLK = f_SYS_CLK/2; Continuous mode, CMP=0xFFFFFFFF, prescaler 1 - UG6971 15.2 guarantees TMRn_CNT readable while counting); TMR0 by default - present on every MAX32 part, unused by the examples, FreeRTOS ticks on SysTick - overridable with -DSYSVIEW_MAX32_TMR=n. The family builds SystemView with SEGGER_SYSVIEW_CORE_OTHER and sets CFG_TUSB_SYSVIEW_TIMESTAMP_BSP, a tusb_sysview.c opt-in for ARMv7-M parts without CYCCNT where the BSP also reports the rate via SEGGER_SYSVIEW_X_GetTimestampFreq(); the fixed-1MHz microsecond contract is unreachable on a powers-of-two-only prescaler. SEGGER_SYSVIEW_X_GetInterruptId() gains the ICSR.VECTACTIVE read for CORE_OTHER builds. MAX32650/32690 use different GCR clock-gate names and stay unported. Measured on the rig: seven host-timed 1.000 s gaps at 1.00044-1.00091 (median +0.05%). stm32f0 refuses to configure SYSVIEW for stm32f070xb at CMake time: that variant has no TIM2 at all, and the alternative was an undefined-reference link error with no hint. The shared hw/bsp/sysview_*.h join ci_select rule 16's hw/bsp file list (cross-family build files, compiled only under -DSYSVIEW which no CI leg enables), mirrored in the spec's rule table, so master's classification guard resolves them on both axes.