increase fiber_stack_init_size and task_stack_init_size from 16 to 64
to fix crashes on 32-bit msvc builds. git bisect identified commit
3246dd2 (which reduced sizes from 64 to 16) as causing the issue.
empirical testing shows 48 fails intermittently but 64 is stable on
32-bit msvc, likely due to different alignment or initialization
overhead on 32-bit platforms.
Co-authored-by: Claude <noreply@anthropic.com>
mingw provides posix compatibility for file i/o but not for signal
handling. hal-posix-task relies on SIGALRM, setitimer(), and
sigprocmask() which are not available on windows even through mingw.
changed hal selection for mruby-task to use hal-win-task for mingw,
while mruby-dir, mruby-io, and mruby-socket correctly use posix hals
for mingw since those features are supported.
Co-authored-by: Claude <noreply@anthropic.com>
when building with MSVC on Windows, RUBY_PLATFORM (from the Ruby
installation running rake) may indicate "mingw" if Ruby was installed
via RubyInstaller, causing incorrect selection of POSIX HALs instead
of Windows HALs.
fixed by checking spec.build.primary_toolchain first:
- if toolchain is "visualcpp", select Windows HALs
- otherwise fall through to existing platform checks
this ensures MSVC builds use hal-win-* gems even when Ruby itself
was installed with MinGW.
affected gems:
- mruby-dir
- mruby-io
- mruby-socket
- mruby-task
Co-authored-by: Claude <noreply@anthropic.com>
change sleep_us_impl and sleep_ms_impl parameters from mrb_int to uint32_t.
this makes the type requirement explicit and resolves msvc warning c4244.
all type conversions happen at ruby boundary functions after validation.
Co-authored-by: Claude <noreply@anthropic.com>
reduced TASK_STACK_INIT_SIZE from 64 to 16 and TASK_CI_INIT_SIZE from 8 to 4,
matching mruby-fiber's conservative allocations. this saves 56 bytes per task
(320 bytes down to 160 bytes for initial allocations). stacks grow dynamically
via mrb_stack_extend when needed.
Co-authored-by: Claude <noreply@anthropic.com>
use t->c.ci->proc directly with explicit null check instead of falling
back to t->proc (which was removed). with c function boundary checks
preventing suspension in c functions, proc should always be valid on resume.
Co-authored-by: Claude <noreply@anthropic.com>
removes duplicate proc field and adds state-based union for result/timeslice,
achieving 16 bytes total savings per task (12.5% reduction):
optimizations:
- removed proc field (stored in c.ci->proc, already marked by gc): 8 bytes
- unified result/timeslice into state union (mutually exclusive): ~4 bytes
- combined with previous commit savings (priority_preemption, started, etc)
total reduction: 128 -> 112 bytes per task
impact:
- 10 tasks: 160 bytes saved
- 50 tasks: 800 bytes saved
- 100 tasks: 1.6 KB saved
all 1770 tests pass with zero functionality changes.
Co-authored-by: Claude <noreply@anthropic.com>
reduces per-task memory usage by 8 bytes (6.2%) through:
- removing priority_preemption field (always equals priority)
- removing started flag (inferred from context status)
- unifying wakeup_tick/join/mutex into single union
old size: 128 bytes
new size: 120 bytes
all tests pass with no functionality changes.
Co-authored-by: Claude <noreply@anthropic.com>
when task.pass is called from within a C function (such as Module.new's
block evaluation), attempting to yield would cause a segfault because C
functions lack valid bytecode program counters (see #6642).
this commit adds C function boundary detection to task.pass, raising a
runtime error when cci > 0 (indicating execution is inside a C function).
this matches fiber's behavior and provides a clear error message instead of
a cryptic segfault.
unlike the previous commit which allowed sleep to fall back to blocking
sleep, task.pass raises an exception because its sole purpose is cooperative
yielding - there is no sensible blocking fallback behavior.
Co-authored-by: Claude <noreply@anthropic.com>
when sleep() was called from within a C function (such as module.new's block
evaluation), the task scheduler would segfault while attempting to resume the
task. this occurred because C functions don't execute bytecode and thus their
callinfo has no valid program counter (pc). when the task tried to resume
execution, mrb_vm_exec() received a null pc, causing a segmentation fault.
the fix adds two safeguards in task.c:
1. C function boundary detection: before suspending a task for sleep, check
if we're inside a C function by examining the cci (c call info) field.
if cci > 0, fall back to blocking sleep via HAL instead of attempting
cooperative context switch. this preserves sleep functionality without
raising exceptions, though it blocks other tasks during the sleep period.
2. proc fallback in execute_task(): use the task's stored proc if the
current callinfo's proc is null, ensuring mrb_vm_exec() always receives
a valid proc pointer.
this approach prioritizes functionality over strict cooperative multitasking
semantics - tasks can still sleep inside C functions, but the sleep becomes
blocking. the alternative would be raising an exception like fiber does, but
that would break existing code unexpectedly.
Co-authored-by: Claude <noreply@anthropic.com>
rename all HAL functions from mrb_<feature>_hal_<name>() to
mrb_hal_<feature>_<name>() for better grouping and clarity. this makes all
HAL functions immediately identifiable with the mrb_hal_* prefix.
affected gems:
- mruby-task: mrb_task_hal_* -> mrb_hal_task_*
- mruby-io: mrb_io_hal_* -> mrb_hal_io_*
- mruby-socket: mrb_socket_hal_* -> mrb_hal_socket_*
- mruby-dir: mrb_dir_hal_* -> mrb_hal_dir_*
Co-authored-by: Claude <noreply@anthropic.com>
changed from angle brackets to quotes for gem-local HAL headers
(task.h, io_hal.h, socket_hal.h), and removed relative path prefix
from task.h include. this follows the mrbgem build system convention
where gem/include/ is automatically added to the include path.
Co-authored-by: Claude <noreply@anthropic.com>
task.c used clock_gettime() directly, breaking portability. added
mrb_task_hal_sleep_us() to hal interface.
Co-authored-by: Claude <noreply@anthropic.com>
follows mrb_{gem_name}_{operation} naming convention consistently
with other hal functions like mrb_task_hal_init. the plural form was
semantically correct but inconsistent with gem naming patterns.
Co-authored-by: Claude <noreply@anthropic.com>
removes mrb_tasks_run and mrb_task_mark_all from task_hal.h as these
are core scheduler functions, not HAL interface functions. only
mrb_tick remains as it must be called by HAL timer callbacks.
Co-authored-by: Claude <noreply@anthropic.com>
separates platform-specific timer and interrupt code into hal-posix-task
and hal-win-task gems. mruby-task now uses HAL interface defined in
task_hal.h, making it easier to port to new platforms.
hal-posix-task: uses sigalrm/setitimer for timer, sigprocmask for irq
protection, and SA_RESTART flag to prevent EINTR on system calls.
hal-win-task: uses multimedia timer API and critical_section for irq
protection.
both HALs support multiple mrb_state instances with single shared timer.
auto-detection loads appropriate HAL based on platform.
Co-authored-by: Claude <noreply@anthropic.com>
mruby-task uses mrb_context and mrb_fiber_state enum, but these are
part of core mruby, not the mruby-fiber gem. the dependency was not
needed.
Co-authored-by: Claude <noreply@anthropic.com>
windows multimedia timer api requires linking with winmm.lib. added
conditional linker library using spec.for_windows? to match mruby
build system conventions.
Co-authored-by: Claude <noreply@anthropic.com>
extended posix platform detection to include macos via __APPLE__ and
__MACH__ defines. implemented full windows hal using multimedia timer
(timeSetEvent) and CRITICAL_SECTION for thread synchronization. added
task_count_update stub for unsupported platforms with clear warnings.
Co-authored-by: Claude <noreply@anthropic.com>
remove MRB_USE_TASK_SCHEDULER ifdef guards from task.h and task.c
since the macro is always defined when compiling this gem
Co-authored-by: Claude <noreply@anthropic.com>
- remove redundant MRB_TASK_CREATED/STOPPED macros from task.h since
they are now properly defined in mrb_fiber_state enum in mruby.h
- declare kw_names array separately to avoid taking address of
temporary array in c++ compilation
Co-authored-by: Claude <noreply@anthropic.com>
clarify that task.new name parameter must be string, document
task#name returns "(noname)" for unnamed tasks, and provide full
structure of task.stat return value.
Co-authored-by: Claude <noreply@anthropic.com>
add tests for sleep/usleep validation, task creation, status/inspect
methods, control methods, task.stat, priority handling, and name
handling.
Co-authored-by: Claude <noreply@anthropic.com>
fix uninitialized kwargs array causing crashes, add type validation for
name (must be String) and priority (must be Integer) parameters, return
"(noname)" for unnamed tasks.
Co-authored-by: Claude <noreply@anthropic.com>
replaced stub with full implementation that returns a hash containing
scheduler statistics:
- tick: current tick counter
- wakeup_tick: next scheduled wakeup time
- dormant/ready/waiting/suspended: per-queue statistics
each queue stat includes:
- count: number of tasks in queue
- tasks: array of task objects in that queue
implements helper function mrb_stat_sub() to walk queues and collect
task information. uses irq disable/enable to ensure consistent snapshot.
returns hash directly as requested, not wrapped in stat object.
Co-authored-by: Claude <noreply@anthropic.com>
added inspect method that returns formatted string showing:
- task pointer address
- task name (string/symbol), or "(unnamed)" for nil/other types
- task status (RUNNING, READY, WAITING, SUSPENDED, DORMANT, UNKNOWN)
format matches original implementation: #<Task:0x12345678 name:STATUS>
avoids mrb_funcall during inspection to prevent vm state issues.
handles string and symbol names directly, treats other types as unnamed.
Co-authored-by: Claude <noreply@anthropic.com>
replaced stub implementation with proper status reporting that returns
symbols representing task state:
- :RUNNING for executing tasks
- :READY for tasks ready to execute
- :WAITING for tasks waiting (sleeping, blocked, etc.)
- :SUSPENDED for manually suspended tasks
- :DORMANT for terminated tasks
- :UNKNOWN for invalid states
implementation matches original mruby-task design using ternary operators
and MRB_SYM() macros for efficient symbol lookup.
Co-authored-by: Claude <noreply@anthropic.com>
improved sleep_us_impl() in several ways:
1. dynamic sleep intervals: now sleeps for actual remaining time instead
of fixed 1ms polling, reducing unnecessary wakeups and improving
efficiency for longer sleeps
2. error handling: added checks for clock_gettime() failures with fallback
to usleep(), and input validation to handle negative values
3. overflow prevention: use named constant USEC_PER_MSEC instead of
literal 1000 for microsecond-to-nanosecond conversion, and validate
input before conversion
4. wraparound handling: fixed tick comparison at line 580 to use signed
arithmetic like other tick comparisons in the codebase
5. code clarity: added time conversion constants (NSEC_PER_MSEC,
NSEC_PER_SEC, USEC_PER_MSEC) to replace magic numbers
all tests pass.
Co-authored-by: Claude <noreply@anthropic.com>
when sleep is called from root context (not within a task), it was
instantly advancing the simulated tick counter instead of actually
delaying. this caused task_pass.rb example to run tasks 0-5 instantly
without proper delays between iterations.
fixed by using clock_gettime() to track elapsed real time and sleeping
in 1ms intervals. also clear switching_ flag when returning from root
context sleep to prevent unwanted context switches.
removed find_earliest_wakeup_tick() function and time-advancing logic
from task_run_one_iteration() as real delays are now handled by sleep
itself.
Co-authored-by: Claude <noreply@anthropic.com>
Renamed constants to use more descriptive underscores:
- MRB_TASKSTATUS_* -> MRB_TASK_STATUS_*
- MRB_TASKREASON_* -> MRB_TASK_REASON_*
This improves code readability by making the constant names clearer.
Co-authored-by: Claude <noreply@anthropic.com>
Eliminated approximately 160 lines of duplicated code (~10% of file) by
extracting common patterns into reusable helpers. This improves
maintainability by consolidating task execution logic, validation
patterns, and state transitions into single locations.
Co-authored-by: Claude <noreply@anthropic.com>
Implement main task wrapper following Fiber's pattern, where root context
is represented by a special task object. This matches PicoRuby behavior
where Task.current always returns a task object, even from root context.
The main task is lazy-allocated on first Task.current call from root,
stored in mrb->task.main_task, and has name "main", status RUNNING,
priority 0. It wraps the root context without allocating a separate
execution context.
Co-authored-by: Claude <noreply@anthropic.com>
Enable Task.pass to work from root context by implementing mini-scheduler
iteration. When called from root context, Task.pass now runs one task
iteration, allowing cooperative multitasking without Task.run. This matches
PicoRuby behavior.
Co-authored-by: Claude <noreply@anthropic.com>
Fix task termination crash caused by fiber_terminate freeing task
context resources. When a task completes, the VM would call
fiber_terminate which frees cibase/stbase, then next resume attempt
crashes dereferencing NULL pointers.
Solution unifies task and fiber lifecycle management:
- Set vmexec flag before calling mrb_vm_exec to prevent fiber_terminate
from being called during normal task completion
- Save proc/pc to local variables to avoid CI_PROC_SET macro corruption
- Add termination check in mrb_task_free to prevent double-free
Tasks now follow the same execution pattern as Fiber, leveraging
VM's built-in context management.
Co-authored-by: Claude <noreply@anthropic.com>
Replace manual mrb_immediate_p check with mrb_gc_mark_value macro
which already includes the immediate check internally.
Co-authored-by: Claude <noreply@anthropic.com>
Implements dual-mechanism GC protection and optimizes task lookup
using pointer arithmetic based on PicoRuby reference implementation.
GC Protection:
- Add mrb_gc_register/unregister to protect Task objects
- Implement mrb_task_mark_all() to mark task contexts during GC
- Store proc reference in mrb_task to prevent premature collection
- Integrate marking into gc.c root_scan_phase
Performance Optimizations:
- Add MRB2TASK macro for O(1) context-to-task conversion
- Optimize Task.current: O(n) queue search -> O(1) pointer arithmetic
- Optimize Task.pass: simplify to root context check
- Optimize Task.join: use MRB2TASK for current task lookup
Bug Fixes:
- Fix MRB_TASK_CREATED/STOPPED to use MRB_FIBER_TERMINATED
- Add safety check to prevent execution of terminated tasks
- Initialize callinfo PC to bytecode start in task_init_context
Co-authored-by: Claude <noreply@anthropic.com>
dynamically enable/disable timer interrupts based on scheduler state.
timer disabled when only one runnable task exists.
timer enabled when multiple tasks need preemption or sleeping tasks need wakeup.
use counter arrays to track ready/waiting tasks per vm.
separate platform-specific timer control from generic decision logic.
update counters at all task state transitions.
eliminates 250 interrupts/second in single-task workloads.
improves cpu efficiency and power consumption.
simplifies porting to new platforms.
Co-authored-by: Claude <noreply@anthropic.com>
change sleep, usleep, sleep_ms from private methods to module functions
to match cruby behavior where sleep can be called as both bare sleep and
kernel.sleep.
Co-authored-by: Claude <noreply@anthropic.com>
replace global_mrb with vm_list to support up to 8 concurrent mrb_state
instances. sigalrm handler now ticks all registered VMs. first VM
initializes timer, last VM stops timer. proper cleanup in hal_final.
Co-authored-by: Claude <noreply@anthropic.com>
Rename sleep_ms_impl to sleep_us_impl as the base implementation,
providing true microsecond precision for usleep. sleep_ms_impl now
simply calls sleep_us_impl with converted values.
This ensures usleep provides proper microsecond granularity instead of
losing precision by converting to milliseconds.
Co-authored-by: Claude <noreply@anthropic.com>
Add usleep method that provides task-aware sleep behavior with
microsecond precision. This overrides mruby-sleep's usleep when both
gems are loaded.
The implementation converts microseconds to milliseconds and uses the
same sleep_ms_impl as sleep_ms, providing cooperative sleep within
tasks and signal-safe blocking sleep otherwise.
Co-authored-by: Claude <noreply@anthropic.com>
The sleep implementation now properly handles signal interruptions from
the sigalrm timer by using nanosleep with retry loop instead of usleep.
This commit also makes sleep override mruby-sleep's implementation when
both gems are loaded, providing task-aware sleep behavior.
Changes:
- replace usleep with nanosleep for signal-safe blocking sleep
- add retry loop to handle eintr interruptions
- use mrb_define_private_method_id for both sleep and sleep_ms
- add time.h and presym.h headers
Co-authored-by: Claude <noreply@anthropic.com>
This commit fixes several critical issues in the task scheduler:
1. Context switching now properly saves and restores ci/cci pointers
and sets prev links, following the fiber implementation pattern.
This prevents crashes when tasks complete.
2. Sleep implementation now falls back to blocking sleep (usleep/Sleep)
when not in task context, fixing standalone sleep calls.
3. Removed unused functions q_find_task and task_free to eliminate
compiler warnings.
4. Added platform-specific headers for sleep functions on Unix/Windows.
5. Enabled HAL initialization which was previously commented out.
6. Added SA_RESTART flag to SIGALRM handler to prevent timer from
interrupting IO syscalls, fixing mrbtest IO.popen failures.
Co-authored-by: Claude <noreply@anthropic.com>
this patch fixes several critical issues in the task scheduler:
1. vm integration for computed goto dispatch mode:
- added task switching check in NEXT macro for computed goto
- previous implementation only worked with switch dispatch mode
- now Task.pass properly yields control to other tasks
2. task lifecycle tracking:
- added 'started' flag to mrb_task structure
- fixed first-run detection to avoid popping callinfo multiple times
- vm overwrites context status during execution, making it unreliable
3. removed mrblib/task.rb:
- empty Ruby method stubs were overriding C implementations
- all task methods now properly implemented in C
4. cleaned up task scheduler loop:
- proper task completion detection using switching flag
- round-robin scheduling for tasks at same priority
- clean scheduler exit when all tasks complete
tasks now cooperatively yield with Task.pass and complete cleanly.
Co-authored-by: Claude <noreply@anthropic.com>
add hardware abstraction layer with posix implementation:
- setitimer: generates periodic sigalrm for tick-based scheduling
- signal handler: calls mrb_tick on each timer interrupt
- sigprocmask: enables/disables interrupts by blocking sigalrm
- usleep: idle cpu implementation for posix platforms
the hal is initialized during gem init and starts the periodic
timer automatically. non-posix platforms get stub implementations.
tick period is configurable via MRB_TICK_UNIT (default 4ms).
Co-authored-by: Claude <noreply@anthropic.com>
Yukihiro "Matz" Matsumoto