kleidos/test/hal/ at main · jmrp.io/kleidos · Tangled

jmrp.io / kleidos

0

[READ-ONLY] Mirror of https://github.com/jmrplens/kleidos. Kleidos — Hardware BLE password manager for ESP32-S3 (M5StickC Plus2, StickS3, M5Stack Gray, T-Deck, Cardputer) jmrplens.github.io/kleidos/

0

kleidos / test / hal /

at main 11 folders

José M. Requena Plens Merge Phase 3.5 Workstream C: on-demand GPS time sync 9d ago

Add display-only ClockSource record (NVS) and surface it in RTC? Record WHO set the clock last (Manual / WebAdmin / Gps; u8 clk_src in the kleidos NVS namespace) so the console and diagnostics can show the provenance of the running time. The two existing writers now tag themselves: the set-clock commit records Manual and the admin portal's time-sync handler records WebAdmin; the GPS writer lands with the sync task. RTC? appends src= as the LAST field on both the ok=0 and ok=1 lines so existing field-parsing HIL assertions stay valid. SECURITY: the tag is write-and-display only — it is never read by rtc::internal::evaluateTrust, rtc::clockTrust or brute_force_guard (grep-verified), and the running trust after any sync stays Volatile. The shared native NvsStore fake (test/support/fake_storage.h) grows from a no-op into a functional in-memory u8 model with a raw corruption seam, reset alongside the filesystem model; the set-clock commit suite's spy becomes per-key so both the tz_idx write and the Manual tag are assertable in one commit. Claude-Session: https://claude.ai/code/session_01SV58JXhfxhhc9DC6vdjBo4

1 week ago

test_clip_window

Test the software clip and drawTextBox Add host unit coverage for the new clip feature: - test_clip_window: the integer rect intersection and ClipWindow (full / partial / disjoint overlap, inactive pass-through, per-pixel membership). - test_display_graphics: a ClipSurface fake that honors a ClipWindow exactly like the drivers, asserting an armed clip drops every fill / single-pixel / text-glyph pixel outside the window (paintedOutside == 0) and that clearing the clip restores full-surface drawing. - test_text_box_component: drawTextBox wraps long text into multiple rows, arms one clip equal to the rect (then clipReset), draws every row under the active clip, cuts vertical overflow at the rect bottom, and the no-wrap / empty-input guards.

3 weeks ago

test_clock_source

Add display-only ClockSource record (NVS) and surface it in RTC? Record WHO set the clock last (Manual / WebAdmin / Gps; u8 clk_src in the kleidos NVS namespace) so the console and diagnostics can show the provenance of the running time. The two existing writers now tag themselves: the set-clock commit records Manual and the admin portal's time-sync handler records WebAdmin; the GPS writer lands with the sync task. RTC? appends src= as the LAST field on both the ok=0 and ok=1 lines so existing field-parsing HIL assertions stay valid. SECURITY: the tag is write-and-display only — it is never read by rtc::internal::evaluateTrust, rtc::clockTrust or brute_force_guard (grep-verified), and the running trust after any sync stays Volatile. The shared native NvsStore fake (test/support/fake_storage.h) grows from a no-op into a functional in-memory u8 model with a raw corruption seam, reset alongside the filesystem model; the set-clock commit suite's spy becomes per-key so both the tz_idx write and the Manual tag are assertable in one commit. Claude-Session: https://claude.ai/code/session_01SV58JXhfxhhc9DC6vdjBo4

1 week ago

test_dirty_region_set

perf(display): multi-region dirty blit + queued interrupt-DMA staging Structural color-TFT render win from the graphics research (chunk R). Three independently-correct parts in the shadow-framebuffer blit; the readRect/writePixels/markFrameDirty contract is unchanged (no visual regression), and the e-paper/mono backends are untouched. 1. Multi-region dirty tracking (research v1 §2.3). Replace the single growing dirty bounding box with DirtyRegionSet<8>: a fixed array of disjoint dirty rects with an LVGL-style per-axis area-join (overlap/abut/within-gap merge, cascading). commitFrame() blits each disjoint region instead of the union span, so a status-bar + far list-row change streams only the two rects. On slot overflow it degrades to today's single union bbox — correctness never depends on the slot count. Factored into the header-only, framework-agnostic dirty_region_set.h and covered by a native host test (area-join + a brute-force pixel-coverage invariant). 2. Async/queued interrupt-DMA blit + double-buffered staging (research v2 NEW-3). New platform SPI facade calls queueTransmit()/awaitQueued() wrap spi_device_queue_trans/get_trans_result. blitShadowRect() now ping-pongs two 256-byte staging buffers: queue chunk N's DMA, pack chunk N+1 while N is in flight, reclaim N in FIFO order. Single-chunk rects keep the polling path (lower one-shot latency). The display device queue depth is bumped to 2. 3. Buffer-age damage accumulation (research v2 NEW-7): N/A for the single shadow FB (no aged back buffer); documented in dirty_region_set.h as the natural home for it once a back buffer / e-paper A/B waveform lands. 4. Render instrumentation gated behind KLEIDOS_RENDER_STATS (debug builds with the live framebuffer): per-commit blit µs, regions, bytes, frame µs, overflow and async counters, plus totals. Read/zeroed over serial via the new `RENDERSTATS [RESET]` debug command (display::formatRenderStats). Classic-ESP32 feasibility: the queued interrupt-DMA path uses only generic spi_master queue APIs (no S3-only peripheral), so it works on classic ESP32 as well as the S3 — no gating needed for the drop-or-gate decision. sticks3 flash +2356 B (1,611,699 -> 1,614,055). Native suite 955/955 green.

3 weeks ago

test_display_graphics

Test the software clip and drawTextBox Add host unit coverage for the new clip feature: - test_clip_window: the integer rect intersection and ClipWindow (full / partial / disjoint overlap, inactive pass-through, per-pixel membership). - test_display_graphics: a ClipSurface fake that honors a ClipWindow exactly like the drivers, asserting an armed clip drops every fill / single-pixel / text-glyph pixel outside the window (paintedOutside == 0) and that clearing the clip restores full-surface drawing. - test_text_box_component: drawTextBox wraps long text into multiple rows, arms one clip equal to the rect (then clipReset), draws every row under the active clip, cuts vertical overflow at the rect bottom, and the no-wrap / empty-input guards.

3 weeks ago

test_keyboard_input

fix(test): configure clangd and clang-tidy for native test tree (#3) * fix(test): configure clangd and clang-tidy for native test tree - Generate compile_commands.json from pio run -e native -t compiledb so clangd resolves src/ headers in test files without errors. - Fix .clangd PathMatch regex (relative path, not absolute) and add -I.pio/build/native/unity_config so unity_config.h resolves. - Remove overly broad Remove: -I* that was stripping all added paths. - Add -Wno-format-security and UnusedIncludes: None to test block. - Add test/.clang-tidy that inherits the root config and: - Allows test_* functions to use snake_case (FunctionIgnoredRegexp) - Disables cert-err33-c, cert-dcl50-cpp, cppcoreguidelines-pro-type-cstyle-cast, bugprone-argument-comment, bugprone-misplaced-widening-cast, bugprone-implicit-widening-of-multiplication-result, misc-unused-using-decls — all false positives in test context. - Add /* intentionally empty */ comment to all empty setUp()/tearDown() bodies (28 files) to suppress SonarLint c:S1186. - Fix test_vault_task: change relative include ../../src/vault/vault_task.h to vault/vault_task.h (consistent with all other test files; -Isrc already in compile flags). - Reorganise test/ into a module-mirroring subtree layout: test/ble/, test/crypto/, test/hal/, test/ota/, test/platform/, test/states/, test/totp/, test/ui/, test/vault/, test/web/ (old flat directories removed). - Add sonar-project.properties exclusion e4: suppress c:S1186 on test/**/*.cpp at SonarCloud level. All 468 native unit tests pass (pio test -e native). * fix(clangd): add -DESP_PLATFORM to global CompileFlags All Kleidos targets are ESP32/ESP32-S3 (framework = espidf), so ESP_PLATFORM is always defined at compile time. Without this flag in the clangd config, the language server was activating #else branches inside #ifdef ESP_PLATFORM blocks, hiding device code and showing false diagnostics for src/ files. The test/** fragment already has Remove: -DESP_PLATFORM, so native unit-test files continue to compile without the flag. * revert(clangd): remove incorrect -DESP_PLATFORM from global CompileFlags Adding -DESP_PLATFORM globally caused 'too many errors' in src/ files because compile_commands.json was generated from the native env and lacked all ESP-IDF/Xtensa include paths. The correct fix is to generate compile_commands.json from a device env: ./scripts/generate_compiledb.sh sticks3 This populates the DB with proper -DESP_PLATFORM and all IDF includes for every src/ translation unit. The symlink at the project root then points clangd to the right compilation context.

1 month ago

test_rgb565_blend

feat(ui): complete AA primitive family and apply across line-art Add strokeLineAA (coverage/point-to-segment, round caps) and AA rounded-rect (fillRoundRectAA / strokeRoundRectAA) to aa_shapes, and route the UI's jaggy line-art through them on color panels: - strokeLineAA: integer/fixed-point point-to-segment coverage reusing distanceSub/coverageInside; round caps fall out of the distance method. Pure detail::segmentCoverage255 helper is host-testable (aa_shapes.cpp pure section split out of the ESP_PLATFORM guard). - Canvas gains supportsAA()/strokeLineAA()/fillRingSectorAA(); base defaults stay crisp (stacked stroke / no-op) so host + e-paper/mono keep non-AA render. RasterCanvas overrides route fillRoundRect, drawRoundRect, fillCircle, drawCircle, strokeLineAA and the ring sector through coverage-AA when display::canReadRect(). - Eye glyph: lid curve sampled into AA segments + AA slash; pupil now via the self-upgrading Canvas circle (drops the inline aa:: branch). - icon_component: chevrons/check/back/gear/key/wifi/clock/lock strokes use strokeLineAA; circles self-AA. Battery body left crisp (axis-aligned). - Round-rect corners AA on cards/buttons/fields/popup; straight edges crisp. - Progress arc routes through fillRingSectorAA on color (e-paper keeps the spoke fan). blendRgb565 (item 2): the LVGL single-multiply magic-mask needs a 5-bit alpha mix that drifts up to 2 LSB on green vs the exact divide-by-255 blend, exceeding the <= 1 LSB no-drift bar, so the exact form is kept (deferred). A round-trip test pins old-vs-current within <= 1 LSB. Tests: new hal/test_rgb565_blend (round-trip) + ui/test_aa_shapes (line coverage). aa_shapes.cpp added to the native build filter.

4 weeks ago

test_rtc_clock_trust

Extend native rtc:: test fake to the full facade surface Add test/support/fake_rtc.h: one header-only, full-surface fake of the non-inline rtc:: facade (clock/trust primitives plus the Phase-3 alarm/timer/CLKOUT/alarm-wake surface), driven by a single settable FakeRtcState with a reset() helper. It deliberately omits getValidTime so tests still exercise the real header-inline trust gate. This gives any native test a deterministic, canonical rtc:: fake and stops per-suite fakes drifting behind the facade. Migrate the two standalone rtc-only suites (test_rtc_clock_trust, test_totp_epoch) onto the shared fake with identical case behavior. The two entangled vault suites (test_brute_force_guard, test_audit_log_vault_ops) keep their in-TU fakes, which already match the shared surface, to avoid destabilizing the security lockout and audit timing tests. Also register the previously-unfiltered states/test_totp_epoch suite in the native test_filter so the migrated suite runs in the default gate. Full native suite: 1607/1607 passing. Claude-Session: https://claude.ai/code/session_01U878bNJtg9bCvAMM8CFKqS

1 week ago

test_rtc_hal_floor_cache

Close the trust-cache races and tighten the zone validator Consolidated-review remediation for the R1..R4 trust/clock work: - I-1: replace the internal seed-floor cache's boolean valid flag with a generation counter. applyEpoch bumps the generation after persisting the new floor; readSeedFloor snapshots it before the NVS read and publishes only under an unchanged generation, and a cache hit requires the cached generation to be current — a reader that raced a set can no longer latch a stale floor and pin clockTrust at None until reboot. Covered by a new native suite (test/hal/test_rtc_hal_floor_cache) that scripts the exact interleaving through an NvsStore fake hook; the device-only currentEpoch definition is now ESP_PLATFORM-gated so the backend TU links on the host against the injectable epoch provider. - M-1: apply the same generation pattern to the BM8563 VL/integrity cache so an in-flight pre-write chip read cannot overwrite a set's invalidation. - M-2: make s_clockSetThisSession a std::atomic<bool> (relaxed) — it crosses the services/UI cores. - M-3: zoneEntryValid now rejects active DST rules with startMonth == endMonth, which would silently classify via the southern-wraparound branch. - M-4: add a drawClockUnsetChip overload taking the precomputed reserve and use it from the Stick header band, so trust + chip width are evaluated once per paint (no one-frame reserve/draw mismatch on a trust flip). Claude-Session: https://claude.ai/code/session_01SV58JXhfxhhc9DC6vdjBo4

1 week ago

test_rtc_internal

Add ESP32 internal-RTC rtc:: backend for the no-I2C-RTC boards Replace the no-op rtc:: fallback on the four boards without an I2C calendar RTC (sticks3/cardputer/tdeck on ESP32-S3, m5stack_gray on classic ESP32) with a real backend over the SoC internal RTC. - Time base is gettimeofday/settimeofday (RTC-timer-backed: survives deep sleep and non-power-on resets), not a monotonic esp_timer count. - clockTrust() reports Volatile once the clock is set and no cold-boot/ backward-jump is detected, else None (the internal equivalent of the BM8563 voltage-low flag): power-on reset (wakeCause Undefined), a backward time jump beyond a drift margin, or a never-set clock all fail closed to None. - The NVS seed (namespace "kleidos", key "rtc_epoch", written by the admin portal) is read as the floor on boot and written on every set. - setCountdownTimer/armAlarmWake compute delta = target - now, clamp to the LP-timer range, and arm sleep::enableTimerWakeup (never 0/underflow); the internal wake presents as WakeCause::Timer. Functions with no internal equivalent (setAlarm/readAlarm/alarmTriggered/CLKOUT/...) return false. Pure conversion/trust/wake logic lives in a host-testable header (rtc_internal_clock.h) with a native-only injectable epoch seam mirroring platform/Clock.h; the backend TU wires it to gettimeofday/settimeofday, NvsStore and the sleep facade. rtc_hal.cpp is converted in place so exactly one !HAS_BM8563_RTC definition of the rtc:: symbols exists (no ODR clash with the BM8563 backend). Claude-Session: https://claude.ai/code/session_01U878bNJtg9bCvAMM8CFKqS

1 week ago

test_shake_detector

fix(test): configure clangd and clang-tidy for native test tree (#3) * fix(test): configure clangd and clang-tidy for native test tree - Generate compile_commands.json from pio run -e native -t compiledb so clangd resolves src/ headers in test files without errors. - Fix .clangd PathMatch regex (relative path, not absolute) and add -I.pio/build/native/unity_config so unity_config.h resolves. - Remove overly broad Remove: -I* that was stripping all added paths. - Add -Wno-format-security and UnusedIncludes: None to test block. - Add test/.clang-tidy that inherits the root config and: - Allows test_* functions to use snake_case (FunctionIgnoredRegexp) - Disables cert-err33-c, cert-dcl50-cpp, cppcoreguidelines-pro-type-cstyle-cast, bugprone-argument-comment, bugprone-misplaced-widening-cast, bugprone-implicit-widening-of-multiplication-result, misc-unused-using-decls — all false positives in test context. - Add /* intentionally empty */ comment to all empty setUp()/tearDown() bodies (28 files) to suppress SonarLint c:S1186. - Fix test_vault_task: change relative include ../../src/vault/vault_task.h to vault/vault_task.h (consistent with all other test files; -Isrc already in compile flags). - Reorganise test/ into a module-mirroring subtree layout: test/ble/, test/crypto/, test/hal/, test/ota/, test/platform/, test/states/, test/totp/, test/ui/, test/vault/, test/web/ (old flat directories removed). - Add sonar-project.properties exclusion e4: suppress c:S1186 on test/**/*.cpp at SonarCloud level. All 468 native unit tests pass (pio test -e native). * fix(clangd): add -DESP_PLATFORM to global CompileFlags All Kleidos targets are ESP32/ESP32-S3 (framework = espidf), so ESP_PLATFORM is always defined at compile time. Without this flag in the clangd config, the language server was activating #else branches inside #ifdef ESP_PLATFORM blocks, hiding device code and showing false diagnostics for src/ files. The test/** fragment already has Remove: -DESP_PLATFORM, so native unit-test files continue to compile without the flag. * revert(clangd): remove incorrect -DESP_PLATFORM from global CompileFlags Adding -DESP_PLATFORM globally caused 'too many errors' in src/ files because compile_commands.json was generated from the native env and lacked all ESP-IDF/Xtensa include paths. The correct fix is to generate compile_commands.json from a device env: ./scripts/generate_compiledb.sh sticks3 This populates the DB with proper -DESP_PLATFORM and all IDF includes for every src/ translation unit. The symlink at the project root then points clangd to the right compilation context.

1 month ago