chore: issue 57 remove duplicated components (#58) * feat: adopt emil's hal.cortex_m, deduplicating cortex/default_init Bump the emil pin to 321a369d (embedded-pro/embedded-infra-lib#95, "generic Cortex-M layer") and link hal.cortex_m instead of hal_st's own copy of InterruptCortex.hpp/cpp, across all 32 drivers that use InterruptHandler/ InterruptTable/DispatchedInterruptHandler/ImmediateInterruptHandler/ InterruptPriority/ActiveInterrupt (now hal::cortex::*). hal.cortex_m also brings FaultTracer, whose weak HardFault_Handler etc. collide with the vendor startup files' ".thumb_set HardFault_Handler, Default_Handler" alias. halst_target_bringup() now forces the link order of hal.cortex_m, hal_st.bringup, and hal.cortex_m.runtime explicitly so FaultTracer wins deterministically instead of by linker luck, and reuses emil's abort()/_sbrk instead of hal-st's own (removing another duplicate; _sbrk needs the `end` linker symbol, added to sections.ld). hal_st/cortex and hal_st/default_init are gone. What's left of the latter (Default_Handler_Forwarded, HAL_InitTick, __assert_func, assert_failed) has no emil equivalent -- it's ST HAL bring-up glue, not portable -- so it moves to hal_st/bringup/, named after emil's own term for this category of code. DataWatchpointAndTrace, the other tenant of the old cortex/ folder, moves to its own hal_st/dwt/ since nothing else needs it. Verified by direct arm-none-eabi-g++/ld compilation and linking against the real emil source (the full stm32g431 CMake preset needs a packaged host build this environment can't produce in reasonable time): all edited drivers compile clean, and `-Wl,-y` traces confirm HardFault_Handler, abort, _init, _sbrk, and assert_failed each resolve to exactly the intended definition with no multiple-definition conflicts. Co-Authored-By: Claude Sonnet 5 * chore: add Claude Code project config (AGENTS.md, CLAUDE.md, .claude/settings.json) Ignore .claude/scheduled_tasks.lock, which is session-local runtime state. Co-Authored-By: Claude Sonnet 5 * fix: only enable EMIL_BUILD_CORTEX_M for ST cross-compiles hal.cortex_m.runtime (Atomic/DefaultInit/SysCallStubs) contains ARM-only inline assembly (mrs/cpsid/msr, bkpt). Forcing EMIL_BUILD_CORTEX_M on unconditionally pulled it into the host x86_64 build too, since TARGET_MCU_VENDOR is only "st" for the stm32* cross-compile presets and unset for "host" -- CI failed assembling those instructions for the host toolchain. Gate the flag on TARGET_MCU_VENDOR so it's only forced on when actually cross-compiling for ST. Co-Authored-By: Claude Sonnet 5 * feat: remove DWT module, superseded by emil's implementation DataWatchPointAndTrace has been contributed upstream to embedded-infra-lib (emil); hal_st no longer needs its own copy. Co-Authored-By: Claude Sonnet 5 * fix: ignore .claude directory in ls-lint .claude was added for Claude Code project config but fails the snake_case .dir rule. Co-Authored-By: Claude Sonnet 5 * chore: bump embedded-infra-lib hash to pull in EMIL's cortex_m SystemTick/TracingReset/TimeKeeperGeneric Prerequisite for #57: the previously pinned EMIL commit doesn't yet have hal/cortex_m/SystemTick(TimerService), services/tracer/TracingReset, or hal/generic/TimeKeeperGeneric. Co-Authored-By: Claude Sonnet 5 * feat: remove hal-st components duplicated by EMIL (#57) - TracingResetStm: pure wrapper around services::TracingReset(hal::ResetStm&, Tracer&), which already exists identically in EMIL. Per the issue, the application should construct it directly instead. - SystemTickStm: unused internally; superseded by hal::cortex::SystemTick. - TimeKeeperStm (TimeKeeperSt): a pass-through to SystemTickTimerService::Now() with no ST-specific behavior; superseded by hal::TimeKeeperGeneric, which resolves via the same default systemTimerServiceId. - SystemTickTimerService: kept (load-bearing for StmEventInfrastructure, and still holds the ST-specific HAL_GetTick() bridge and sub-tick-precision Now() that EMIL's cortex_m version doesn't provide), but its raw SysTick register programming is now delegated to hal::cortex::SystemTick instead of duplicating it inline. Verified with a full stm32g431 cross-build (hal_st.stm32fxxx, hal_st.synchronous_stm32fxxx, hal_st.instantiations) against the newly fetched EMIL commit. Co-Authored-By: Claude Sonnet 5 * feat: remove ResetStm, duplicated by EMIL's hal::cortex::Reset Identical behavior (NVIC_SystemReset() vs. hal::cortex::Reset writing the same AIRCR reset key/bit directly) with no STM32-specific logic. Already orphaned: its only caller, TracingResetStm, was removed earlier in this branch. Co-Authored-By: Claude Sonnet 5 * docs: fix stale halst_target_default_init reference in README, drop restating comment Addresses PR #58 review comments: - README still documented the removed halst_target_default_init() CMake function; the bringup rework renamed it to halst_target_bringup(). - UartStmDuplexDma.hpp had a comment restating the overridden interface with no non-obvious "why", against repo comment style. Co-Authored-By: Claude Sonnet 5 --------- Co-authored-by: Claude Sonnet 5

Uh oh!

remove duplicated components #57

Description

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions