admin/mruby-mruby
1
0
Fork 0
mirror of https://github.com/mruby/mruby synced 2026-06-08 16:11:16 +00:00
Code Issues Packages Projects Releases Wiki Activity
Files
29f84030ca3ef4e106e4c6b5adee49d3b764c829
mruby-mruby/src/vm.c
T
Yukihiro "Matz" Matsumoto 3d8ccee7a8 vm.c: optimize cipush/cipop for common cases 
CI_PROC_SET: split NULL/non-NULL proc paths so the compiler can
eliminate the CFUNC/ALIAS checks when proc is a compile-time NULL
(8 of 11 cipush call sites).

cipop: add fast path for the common case where no env and no blk
are set. skips ci_env_set, orphan check, and env_unshare entirely.
most simple method calls (no blocks, no closures) take this path.

Co-authored-by: Claude <noreply@anthropic.com>
2026-04-23 19:25:32 +09:00

3819 lines
108 KiB
C
Raw Blame History

/*
** vm.c - virtual machine for mruby
**
** See Copyright Notice in mruby.h
*/
#include <mruby.h>
#include <mruby/array.h>
#include <mruby/class.h>
#include <mruby/hash.h>
#include <mruby/irep.h>
#include <mruby/numeric.h>
#include <mruby/proc.h>
#include <mruby/range.h>
#include <mruby/string.h>
#include <mruby/variable.h>
#include <mruby/error.h>
#include <mruby/opcode.h>
#include "value_array.h"
#include <mruby/throw.h>
#include <mruby/dump.h>
#include <mruby/internal.h>
#ifdef MRB_NO_STDIO
#if defined(__cplusplus)
extern "C" {
#endif
void abort(void);
#if defined(__cplusplus)
} /* extern "C" */
#endif
#endif
#define STACK_INIT_SIZE 128
#define CALLINFO_INIT_SIZE 32
/* Define amount of linear stack growth. */
#ifndef MRB_STACK_GROWTH
#define MRB_STACK_GROWTH 128
#endif
/* Maximum recursive depth. Should be set lower on memory constrained systems. */
#ifdef __clang__
#if __has_feature(address_sanitizer) && !defined(__SANITIZE_ADDRESS__)
#define __SANITIZE_ADDRESS__
#endif
#endif
#ifndef MRB_CALL_LEVEL_MAX
#if defined(__SANITIZE_ADDRESS__)
#define MRB_CALL_LEVEL_MAX 128
#else
#define MRB_CALL_LEVEL_MAX 512
#endif
#endif
/* Maximum stack depth. Should be set lower on memory constrained systems.
The value below allows about 60000 recursive calls in the simplest case. */
#ifndef MRB_STACK_MAX
#define MRB_STACK_MAX (0x40000 - MRB_STACK_GROWTH)
#endif
#ifdef VM_DEBUG
# define DEBUG(x) (x)
#else
# define DEBUG(x)
#endif
#ifndef MRB_GC_FIXED_ARENA
static void
mrb_gc_arena_shrink(mrb_state *mrb, int idx)
{
mrb_gc *gc = &mrb->gc;
int capa = gc->arena_capa;
gc->arena_idx = idx;
if (idx < capa / 4) {
capa >>= 2;
if (capa < MRB_GC_ARENA_SIZE) {
capa = MRB_GC_ARENA_SIZE;
}
if (capa != gc->arena_capa) {
gc->arena = (struct RBasic**)mrb_realloc(mrb, gc->arena, sizeof(struct RBasic*)*capa);
gc->arena_capa = capa;
}
}
}
#else
#define mrb_gc_arena_shrink(mrb, idx) mrb_gc_arena_restore(mrb, idx)
#endif
#define CALL_MAXARGS 15
#define CALL_VARARGS (CALL_MAXARGS<<4 | CALL_MAXARGS)
static inline void
stack_clear(mrb_value *from, size_t count)
{
while (count-- > 0) {
SET_NIL_VALUE(*from);
from++;
}
}
static inline void
stack_copy(mrb_value *dst, const mrb_value *src, size_t size)
{
if (!src) return;
memcpy(dst, src, sizeof(mrb_value)*size);
}
static void
stack_init(mrb_state *mrb)
{
struct mrb_context *c = mrb->c;
/* mrb_assert(mrb->stack == NULL); */
c->stbase = (mrb_value*)mrb_malloc(mrb, STACK_INIT_SIZE * sizeof(mrb_value));
c->stend = c->stbase + STACK_INIT_SIZE;
stack_clear(c->stbase, STACK_INIT_SIZE);
/* mrb_assert(ci == NULL); */
static const mrb_callinfo ci_zero = { 0 };
c->cibase = (mrb_callinfo*)mrb_malloc(mrb, CALLINFO_INIT_SIZE * sizeof(mrb_callinfo));
c->ciend = c->cibase + CALLINFO_INIT_SIZE;
c->cibase[0] = ci_zero;
c->ci = c->cibase;
c->ci->u.target_class = mrb->object_class;
c->ci->stack = c->stbase;
c->ci->vis = 1; /* private (2-bit packed) */
}
static inline void
envadjust(mrb_state *mrb, mrb_value *oldbase, mrb_value *newbase)
{
mrb_callinfo *ci = mrb->c->cibase;
/*
* Byte-level calculation to avoid truncation when allocator alignment is
* smaller than sizeof(mrb_value).
* eg: MRB_NO_BOXING + MRB_INT64 with MRB_32BIT => sizeof(mrb_value)=16
* And when memory allocator's alignment is 8 bytes
* Pointer subtraction on mrb_value* would truncate (8/16 -> 0).
* So, we use char* for pointer calculation to get the correct offset in bytes,
* then apply that offset to mrb_value* pointers.
*/
ptrdiff_t off = (char *)newbase - (char *)oldbase;
if (off == 0) return;
while (ci <= mrb->c->ci) {
struct REnv *e = mrb_vm_ci_env(ci);
mrb_value *new_stack = (mrb_value *)((char *)ci->stack + off);
if (e) {
mrb_assert(e->cxt == mrb->c && MRB_ENV_ONSTACK_P(e));
mrb_assert(e->stack == ci->stack);
e->stack = new_stack;
}
ci->stack = new_stack;
ci++;
}
}
/** def rec; $deep =+ 1; if $deep > 1000; return 0; end; rec; end **/
static void
stack_extend_alloc(mrb_state *mrb, mrb_int room)
{
mrb_value *oldbase = mrb->c->stbase;
size_t oldsize = mrb->c->stend - mrb->c->stbase;
size_t size = oldsize;
size_t off = mrb->c->ci->stack ? mrb->c->stend - mrb->c->ci->stack : 0;
if (off > size) size = off;
#ifdef MRB_STACK_EXTEND_DOUBLING
if ((size_t)room <= size)
size *= 2;
else
size += room;
#else
/* Use 1.5x stack growth.
It is slightly slower than doubling the stack space,
but it saves memory on small devices. */
{
size_t newsize = size + (size >> 1); /* 1.5x growth */
if (newsize < size + MRB_STACK_GROWTH)
newsize = size + MRB_STACK_GROWTH;
if (newsize < size + (size_t)room)
newsize = size + room;
size = newsize;
}
#endif
mrb_value *newstack = (mrb_value*)mrb_realloc(mrb, mrb->c->stbase, sizeof(mrb_value) * size);
stack_clear(&(newstack[oldsize]), size - oldsize);
envadjust(mrb, oldbase, newstack);
mrb->c->stbase = newstack;
mrb->c->stend = mrb->c->stbase + size;
/* Raise an exception if the new stack size will be too large,
to prevent infinite recursion. However, do this only after resizing the stack, so mrb_raise has stack space to work with. */
if (size > MRB_STACK_MAX) {
mrb_exc_raise(mrb, mrb_obj_value(mrb->stack_err));
}
}
static inline void
stack_extend(mrb_state *mrb, mrb_int room)
{
if (mrb_unlikely(!mrb->c->ci->stack || mrb->c->ci->stack + room >= mrb->c->stend)) {
stack_extend_alloc(mrb, room);
}
}
/**
* @brief Extends the VM stack.
*
* This function extends the virtual machine stack to accommodate more values.
* If the current stack size is insufficient, it reallocates the stack
* with a larger size.
*
* @param mrb The mruby state.
* @param room The additional number of mrb_value slots required.
*/
MRB_API void
mrb_stack_extend(mrb_state *mrb, mrb_int room)
{
stack_extend(mrb, room);
}
static void
stack_extend_adjust(mrb_state *mrb, mrb_int room, const mrb_value **argp)
{
const struct mrb_context *c = mrb->c;
ptrdiff_t voff = *argp - c->stbase;
if (voff < 0 || voff >= c->stend - c->stbase) {
stack_extend(mrb, room);
}
else {
stack_extend(mrb, room);
*argp = c->stbase + voff;
}
}
static inline struct REnv*
uvenv(mrb_state *mrb, mrb_int up)
{
const struct RProc *proc = mrb->c->ci->proc;
while (up--) {
proc = proc->upper;
if (!proc) return NULL;
}
struct REnv *e = MRB_PROC_ENV(proc);
if (e) return e; /* proc has enclosed env */
return NULL;
}
static inline const struct RProc*
top_proc(mrb_state *mrb, const struct RProc *proc, const struct REnv **envp)
{
while (proc->upper) {
if (MRB_PROC_SCOPE_P(proc) || MRB_PROC_STRICT_P(proc))
return proc;
*envp = proc->e.env;
proc = proc->upper;
}
return proc;
}
#define CI_PROC_SET(ci, p) do {\
ci->proc = p;\
if (p) {\
mrb_assert(!MRB_PROC_ALIAS_P(p));\
ci->pc = (!MRB_PROC_CFUNC_P(p) && p->body.irep) ? p->body.irep->iseq : NULL;\
}\
else {\
ci->pc = NULL;\
}\
} while (0)
void
mrb_vm_ci_proc_set(mrb_callinfo *ci, const struct RProc *p)
{
CI_PROC_SET(ci, p);
}
#define MRB_PROC_RESOLVE_ALIAS(ci, p) do {\
if (MRB_PROC_ALIAS_P(p)) {\
(ci)->mid = (p)->body.mid;\
(p) = (p)->upper;\
}\
} while (0)
#define CI_TARGET_CLASS(ci) (((ci)->u.env && (ci)->u.env->tt == MRB_TT_ENV)? (ci)->u.env->c : (ci)->u.target_class)
struct RClass*
mrb_vm_ci_target_class(const mrb_callinfo *ci)
{
return CI_TARGET_CLASS(ci);
}
void
mrb_vm_ci_target_class_set(mrb_callinfo *ci, struct RClass *tc)
{
struct REnv *e = ci->u.env;
if (e && e->tt == MRB_TT_ENV) {
e->c = tc;
}
else {
ci->u.target_class = tc;
}
}
#define CI_ENV(ci) (((ci)->u.env && (ci)->u.env->tt == MRB_TT_ENV)? (ci)->u.env : NULL)
struct REnv*
mrb_vm_ci_env(const mrb_callinfo *ci)
{
return CI_ENV(ci);
}
static inline void
ci_env_set(mrb_callinfo *ci, struct REnv *e)
{
if (ci->u.env) {
if (ci->u.env->tt == MRB_TT_ENV) {
if (e) {
e->c = ci->u.env->c;
ci->u.env = e;
}
else {
ci->u.target_class = ci->u.env->c;
}
}
else if (e) {
e->c = ci->u.target_class;
ci->u.env = e;
}
}
else {
ci->u.env = e;
}
}
void
mrb_vm_ci_env_set(mrb_callinfo *ci, struct REnv *e)
{
ci_env_set(ci, e);
}
MRB_API void
mrb_vm_ci_env_clear(mrb_state *mrb, mrb_callinfo *ci)
{
struct REnv *e = ci->u.env;
if (e && e->tt == MRB_TT_ENV) {
ci->u.target_class = e->c;
mrb_env_unshare(mrb, e, FALSE);
}
}
#define CINFO_NONE 0 // called method from mruby VM (without C functions)
#define CINFO_SKIP 1 // ignited mruby VM from C
#define CINFO_DIRECT 2 // called method from C
#define CINFO_RESUMED 3 // resumed by `Fiber.yield` (probably the main call is `mrb_fiber_resume()`)
#define BLK_PTR(b) ((mrb_proc_p(b)) ? mrb_proc_ptr(b) : NULL)
static inline mrb_callinfo*
cipush(mrb_state *mrb, mrb_int push_stacks, uint8_t cci, struct RClass *target_class,
const struct RProc *proc, struct RProc *blk, mrb_sym mid, uint16_t argc)
{
struct mrb_context *c = mrb->c;
mrb_callinfo *ci = c->ci + 1;
if (ci < c->ciend) {
c->ci = ci;
}
else {
ptrdiff_t size = ci - c->cibase;
if (size >= MRB_CALL_LEVEL_MAX) {
mrb_exc_raise(mrb, mrb_obj_value(mrb->stack_err));
}
c->cibase = (mrb_callinfo*)mrb_realloc(mrb, c->cibase, sizeof(mrb_callinfo)*size*2);
c->ci = ci = c->cibase + size;
c->ciend = c->cibase + size * 2;
}
ci->mid = mid;
CI_PROC_SET(ci, proc);
ci->blk = blk;
ci->stack = ci[-1].stack + push_stacks;
ci->n = argc & 0xf;
ci->nk = (argc>>4) & 0xf;
ci->cci = cci;
ci->vis = MRB_METHOD_PUBLIC_FL;
ci->u.target_class = target_class;
return ci;
}
static void
fiber_terminate(mrb_state *mrb, struct mrb_context *c, mrb_callinfo *ci)
{
mrb_assert(c != mrb->root_c);
struct REnv *env = CI_ENV(ci);
mrb_assert(env == NULL || MRB_ENV_LEN(env) <= c->stend - ci->stack);
c->status = MRB_FIBER_TERMINATED;
mrb_free(mrb, c->cibase);
c->cibase = c->ciend = c->ci = NULL;
mrb_value *stack = c->stbase;
c->stbase = c->stend = NULL;
if (!env) {
mrb_free(mrb, stack);
}
else {
size_t len = (size_t)MRB_ENV_LEN(env);
if (len == 0) {
env->stack = NULL;
MRB_ENV_CLOSE(env);
mrb_free(mrb, stack);
}
else {
mrb_assert(stack == env->stack);
mrb_write_barrier(mrb, (struct RBasic*)env);
// don't call MRB_ENV_CLOSE() before mrb_realloc().
// the reason is that env->stack may be freed by mrb_realloc() if MRB_DEBUG + MRB_GC_STRESS are enabled.
// realloc() on a freed heap will cause double-free.
stack = (mrb_value*)mrb_realloc(mrb, stack, len * sizeof(mrb_value));
if (mrb_object_dead_p(mrb, (struct RBasic*)env)) {
mrb_free(mrb, stack);
}
else {
env->stack = stack;
MRB_ENV_CLOSE(env);
}
}
}
/* fiber termination should automatic yield or transfer to root */
mrb->c = c->prev;
if (!mrb->c) mrb->c = mrb->root_c;
else c->prev = NULL;
mrb->c->status = MRB_FIBER_RUNNING;
}
mrb_bool
mrb_env_unshare(mrb_state *mrb, struct REnv *e, mrb_bool noraise)
{
mrb_assert(e != NULL);
mrb_assert(MRB_ENV_ONSTACK_P(e));
size_t len = (size_t)MRB_ENV_LEN(e);
if (len == 0) {
e->stack = NULL;
MRB_ENV_CLOSE(e);
return TRUE;
}
size_t live = mrb->gc.live;
mrb_value *p = (mrb_value*)mrb_malloc_simple(mrb, sizeof(mrb_value)*len);
if (live != mrb->gc.live && mrb_object_dead_p(mrb, (struct RBasic*)e)) {
// The e object is now subject to GC inside mrb_malloc_simple().
// Moreover, if NULL is returned due to mrb_malloc_simple() failure, simply ignore it.
mrb_free(mrb, p);
return TRUE;
}
else if (p) {
stack_copy(p, e->stack, len);
e->stack = p;
MRB_ENV_CLOSE(e);
mrb_write_barrier(mrb, (struct RBasic*)e);
return TRUE;
}
else {
e->stack = NULL;
MRB_ENV_CLOSE(e);
MRB_ENV_SET_LEN(e, 0);
MRB_ENV_SET_BIDX(e, 0);
if (!noraise) {
mrb_exc_raise(mrb, mrb_obj_value(mrb->nomem_err));
}
return FALSE;
}
}
static inline mrb_callinfo*
cipop(mrb_state *mrb)
{
struct mrb_context *c = mrb->c;
mrb_callinfo *ci = c->ci;
/* Fast path: no env and no blk (most common for simple method calls) */
if (mrb_likely((!ci->u.env || ci->u.env->tt != MRB_TT_ENV) && !ci->blk)) {
c->ci--;
return c->ci;
}
struct REnv *env = CI_ENV(ci);
ci_env_set(ci, NULL); // make possible to free env by GC if not needed
struct RProc *b = ci->blk;
if (b && !MRB_PROC_STRICT_P(b) && MRB_PROC_ENV(b) == CI_ENV(&ci[-1])) {
b->flags |= MRB_PROC_ORPHAN;
}
if (env && !mrb_env_unshare(mrb, env, TRUE)) {
c->ci--; // exceptions are handled at the method caller; see #3087
mrb_exc_raise(mrb, mrb_obj_value(mrb->nomem_err));
}
c->ci--;
return c->ci;
}
/**
* @brief Protects a C function call from mruby exceptions.
*
* This function executes a C function (`body`) within a protected environment.
* If an mruby exception occurs during the execution of `body`, this function
* catches the exception, sets the `error` flag, and returns the exception object.
* Otherwise, it returns the result of the `body` function and `error` remains FALSE.
*
* This is crucial for calling mruby-related C functions from within C code
* that needs to handle potential mruby exceptions gracefully.
*
* @param mrb The mruby state.
* @param body A pointer to the C function to be executed.
* The function should have the signature: `mrb_value func(mrb_state *mrb, void *userdata)`
* @param userdata A pointer to arbitrary data that will be passed to the `body` function.
* @param error A pointer to an mrb_bool that will be set to TRUE if an exception
* occurred, and FALSE otherwise. Can be NULL if not needed.
* @return The value returned by the `body` function if no exception occurred,
* or the exception object if an exception occurred.
*/
MRB_API mrb_value
mrb_protect_error(mrb_state *mrb, mrb_protect_error_func *body, void *userdata, mrb_bool *error)
{
struct mrb_jmpbuf *prev_jmp = mrb->jmp;
struct mrb_jmpbuf c_jmp;
mrb_value result;
int ai = mrb_gc_arena_save(mrb);
const struct mrb_context *c = mrb->c;
ptrdiff_t ci_index = c->ci - c->cibase;
if (error) { *error = FALSE; }
MRB_TRY(&c_jmp) {
mrb->jmp = &c_jmp;
result = body(mrb, userdata);
mrb->jmp = prev_jmp;
}
MRB_CATCH(&c_jmp) {
mrb->jmp = prev_jmp;
result = mrb_obj_value(mrb->exc);
mrb->exc = NULL;
if (error) { *error = TRUE; }
if (mrb->c == c) {
while (c->ci - c->cibase > ci_index) {
cipop(mrb);
}
}
else {
// It was probably switched by mrb_fiber_resume().
// Simply destroy all successive CINFO_DIRECTs once the fiber has been switched.
c = mrb->c;
while (c->ci > c->cibase && c->ci->cci == CINFO_DIRECT) {
cipop(mrb);
}
}
}
MRB_END_EXC(&c_jmp);
mrb_gc_arena_restore(mrb, ai);
mrb_gc_protect(mrb, result);
return result;
}
void mrb_exc_set(mrb_state *mrb, mrb_value exc);
static mrb_value mrb_run(mrb_state *mrb, const struct RProc* proc, mrb_value self);
#ifndef MRB_FUNCALL_ARGC_MAX
#define MRB_FUNCALL_ARGC_MAX 16
#endif
/**
* @brief Calls a method on an object.
*
* This function invokes a method identified by its name on the `self` object,
* passing the given arguments.
*
* @param mrb The mruby state.
* @param self The receiver object of the method call.
* @param name The name of the method to call (C string).
* @param argc The number of arguments to pass to the method.
* @param ... The variable arguments to pass to the method.
* Each argument must be of type `mrb_value`.
* @return The result of the method call.
* @raise E_ARGUMENT_ERROR if `argc` is greater than `MRB_FUNCALL_ARGC_MAX`.
*/
MRB_API mrb_value
mrb_funcall(mrb_state *mrb, mrb_value self, const char *name, mrb_int argc, ...)
{
mrb_value argv[MRB_FUNCALL_ARGC_MAX];
mrb_sym mid = mrb_intern_cstr(mrb, name);
if (argc > MRB_FUNCALL_ARGC_MAX) {
mrb_raise(mrb, E_ARGUMENT_ERROR, "Too long arguments. (limit=" MRB_STRINGIZE(MRB_FUNCALL_ARGC_MAX) ")");
}
va_list ap;
va_start(ap, argc);
for (mrb_int i = 0; i < argc; i++) {
argv[i] = va_arg(ap, mrb_value);
}
va_end(ap);
return mrb_funcall_argv(mrb, self, mid, argc, argv);
}
/**
* @brief Calls a method on an object using a method ID.
*
* This function invokes a method identified by its symbol ID (`mid`) on
* the `self` object, passing the given arguments. Using a method ID
* can be more efficient than using a string name if the method is called
* frequently, as it avoids repeated string-to-symbol lookups.
*
* @param mrb The mruby state.
* @param self The receiver object of the method call.
* @param mid The symbol ID of the method to call.
* @param argc The number of arguments to pass to the method.
* @param ... The variable arguments to pass to the method.
* Each argument must be of type `mrb_value`.
* @return The result of the method call.
* @raise E_ARGUMENT_ERROR if `argc` is greater than `MRB_FUNCALL_ARGC_MAX`.
*/
MRB_API mrb_value
mrb_funcall_id(mrb_state *mrb, mrb_value self, mrb_sym mid, mrb_int argc, ...)
{
mrb_value argv[MRB_FUNCALL_ARGC_MAX];
if (argc > MRB_FUNCALL_ARGC_MAX) {
mrb_raise(mrb, E_ARGUMENT_ERROR, "Too long arguments. (limit=" MRB_STRINGIZE(MRB_FUNCALL_ARGC_MAX) ")");
}
va_list ap;
va_start(ap, argc);
for (mrb_int i = 0; i < argc; i++) {
argv[i] = va_arg(ap, mrb_value);
}
va_end(ap);
return mrb_funcall_argv(mrb, self, mid, argc, argv);
}
static mrb_int
mrb_ci_kidx(const mrb_callinfo *ci)
{
if (ci->nk == 0) return -1;
return (ci->n == CALL_MAXARGS) ? 2 : ci->n + 1;
}
static inline mrb_int
mrb_bidx(uint8_t n, uint8_t k)
{
if (n == 15) n = 1;
if (k == 15) n += 1;
else n += k*2;
return n + 1; /* self + args + kargs */
}
static inline mrb_int
ci_bidx(mrb_callinfo *ci)
{
return mrb_bidx(ci->n, ci->nk);
}
mrb_int
mrb_ci_bidx(mrb_callinfo *ci)
{
return ci_bidx(ci);
}
mrb_int
mrb_ci_nregs(mrb_callinfo *ci)
{
if (!ci) return 4;
mrb_int nregs = ci_bidx(ci) + 1; /* self + args + kargs + blk */
const struct RProc *p = ci->proc;
if (p && !MRB_PROC_CFUNC_P(p) && p->body.irep && p->body.irep->nregs > nregs) {
return p->body.irep->nregs;
}
return nregs;
}
mrb_value mrb_obj_missing(mrb_state *mrb, mrb_value mod);
static mrb_method_t
prepare_missing(mrb_state *mrb, mrb_callinfo *ci, mrb_value recv, mrb_sym mid, mrb_value blk, mrb_bool super)
{
mrb_sym missing = MRB_SYM(method_missing);
mrb_value *argv = &ci->stack[1];
mrb_value args;
mrb_method_t m;
/* pack positional arguments */
if (ci->n == 15) args = argv[0];
else args = mrb_ary_new_from_values(mrb, ci->n, argv);
if (mrb_func_basic_p(mrb, recv, missing, mrb_obj_missing)) {
method_missing:
if (super) mrb_no_method_error(mrb, mid, args, "no superclass method '%n' for %T", mid, recv);
else mrb_method_missing(mrb, mid, recv, args);
/* not reached */
}
if (mid != missing) {
ci->u.target_class = mrb_class(mrb, recv);
}
m = mrb_vm_find_method(mrb, ci->u.target_class, &ci->u.target_class, missing);
if (MRB_METHOD_UNDEF_P(m)) goto method_missing; /* just in case */
stack_extend(mrb, 4);
argv = &ci->stack[1]; /* maybe reallocated */
if (ci->nk == 0) {
argv[1] = blk;
}
else {
mrb_assert(ci->nk == 15);
if (ci->n != CALL_MAXARGS) {
argv[1] = argv[ci->n]; /* keyword arguments */
}
argv[2] = blk;
}
argv[0] = args; /* must be replaced after saving argv[0] as it may be a keyword argument */
ci->n = CALL_MAXARGS;
/* ci->nk is already set to zero or CALL_MAXARGS */
mrb_ary_unshift(mrb, args, mrb_symbol_value(mid));
ci->mid = missing;
return m;
}
static void
funcall_args_capture(mrb_state *mrb, int stoff, mrb_int argc, const mrb_value *argv, mrb_value block, mrb_callinfo *ci)
{
if (argc < 0 || argc > INT32_MAX) {
mrb_raisef(mrb, E_ARGUMENT_ERROR, "negative or too big argc for funcall (%i)", argc);
}
ci->nk = 0; /* funcall does not support keyword arguments */
if (argc < CALL_MAXARGS) {
mrb_int extends = stoff + argc + 2 /* self + block */;
stack_extend_adjust(mrb, extends, &argv);
mrb_value *args = mrb->c->ci->stack + stoff + 1 /* self */;
stack_copy(args, argv, argc);
args[argc] = block;
ci->n = (uint8_t)argc;
}
else {
int extends = stoff + 3 /* self + splat + block */;
stack_extend_adjust(mrb, extends, &argv);
mrb_value *args = mrb->c->ci->stack + stoff + 1 /* self */;
args[0] = mrb_ary_new_from_values(mrb, argc, argv);
args[1] = block;
ci->n = CALL_MAXARGS;
}
}
static inline mrb_value
ensure_block(mrb_state *mrb, mrb_value blk)
{
if (!mrb_nil_p(blk) && !mrb_proc_p(blk)) {
blk = mrb_type_convert(mrb, blk, MRB_TT_PROC, MRB_SYM(to_proc));
/* The stack might have been reallocated during mrb_type_convert(), see #3622 */
}
return blk;
}
/**
* @brief Calls a method on an object with a block.
*
* This function invokes a method identified by its symbol ID (`mid`) on
* the `self` object, passing the given arguments (`argv`) and a block (`blk`).
*
* @param mrb The mruby state.
* @param self The receiver object of the method call.
* @param mid The symbol ID of the method to call.
* @param argc The number of arguments in `argv`.
* @param argv A pointer to an array of `mrb_value` arguments.
* @param blk The block to pass to the method. If no block is to be passed,
* use `mrb_nil_value()`. If `blk` is not nil and not a proc,
* it will be converted to a proc using `to_proc`.
* @return The result of the method call.
* @raise E_ARGUMENT_ERROR if `argc` is negative or too large.
* @raise E_STACK_ERROR if the call level exceeds `MRB_CALL_LEVEL_MAX`.
*/
MRB_API mrb_value
mrb_funcall_with_block(mrb_state *mrb, mrb_value self, mrb_sym mid, mrb_int argc, const mrb_value *argv, mrb_value blk)
{
mrb_value val;
int ai = mrb_gc_arena_save(mrb);
if (!mrb->jmp) {
struct mrb_jmpbuf c_jmp;
ptrdiff_t nth_ci = mrb->c->ci - mrb->c->cibase;
MRB_TRY(&c_jmp) {
mrb->jmp = &c_jmp;
/* recursive call */
val = mrb_funcall_with_block(mrb, self, mid, argc, argv, blk);
mrb->jmp = NULL;
}
MRB_CATCH(&c_jmp) { /* error */
while (nth_ci < (mrb->c->ci - mrb->c->cibase)) {
cipop(mrb);
}
mrb->jmp = 0;
val = mrb_obj_value(mrb->exc);
}
MRB_END_EXC(&c_jmp);
mrb->jmp = NULL;
}
else {
mrb_method_t m;
mrb_callinfo *ci = mrb->c->ci;
mrb_int n = mrb_ci_nregs(ci);
if (!mrb->c->stbase) {
stack_init(mrb);
}
if (ci - mrb->c->cibase > MRB_CALL_LEVEL_MAX) {
mrb_exc_raise(mrb, mrb_obj_value(mrb->stack_err));
}
blk = ensure_block(mrb, blk);
ci = cipush(mrb, n, CINFO_DIRECT, NULL, NULL, BLK_PTR(blk), 0, 0);
funcall_args_capture(mrb, 0, argc, argv, blk, ci);
ci->u.target_class = mrb_class(mrb, self);
m = mrb_vm_find_method(mrb, ci->u.target_class, &ci->u.target_class, mid);
if (MRB_METHOD_UNDEF_P(m)) {
m = prepare_missing(mrb, ci, self, mid, mrb_nil_value(), FALSE);
}
else {
ci->mid = mid;
}
ci->proc = MRB_METHOD_PROC_P(m) ? MRB_METHOD_PROC(m) : NULL;
if (MRB_METHOD_CFUNC_P(m)) {
mrb->exc = NULL;
ci->stack[0] = self;
val = MRB_METHOD_CFUNC(m)(mrb, self);
cipop(mrb);
if (mrb->exc != NULL) {
mrb_exc_raise(mrb, mrb_obj_value(mrb->exc));
}
}
else {
/* handle alias */
MRB_PROC_RESOLVE_ALIAS(ci, ci->proc);
ci->cci = CINFO_SKIP;
val = mrb_run(mrb, ci->proc, self);
}
}
mrb_gc_arena_restore(mrb, ai);
mrb_gc_protect(mrb, val);
return val;
}
/**
* @brief Calls a method on an object with an array of arguments.
*
* This function is similar to `mrb_funcall_with_block` but takes arguments
* as a C array (`argv`) and does not take an explicit block argument.
* If a block is needed, `mrb_funcall_with_block` should be used.
* This function is essentially a convenience wrapper around
* `mrb_funcall_with_block` with `mrb_nil_value()` for the block.
*
* @param mrb The mruby state.
* @param self The receiver object of the method call.
* @param mid The symbol ID of the method to call.
* @param argc The number of arguments in `argv`.
* @param argv A pointer to an array of `mrb_value` arguments.
* @return The result of the method call.
* @see mrb_funcall_with_block
*/
MRB_API mrb_value
mrb_funcall_argv(mrb_state *mrb, mrb_value self, mrb_sym mid, mrb_int argc, const mrb_value *argv)
{
return mrb_funcall_with_block(mrb, self, mid, argc, argv, mrb_nil_value());
}
static void
check_argument_count(mrb_state *mrb, const mrb_callinfo *ci, mrb_aspec aspec)
{
mrb_int argc = ci->n;
if (mrb_unlikely(argc == CALL_MAXARGS)) {
argc = RARRAY_LEN(ci->stack[1]);
}
/* keyword hash counts as positional if method doesn't accept keywords */
if (ci->nk > 0 && MRB_ASPEC_KEY(aspec) == 0 && !MRB_ASPEC_KDICT(aspec)) {
mrb_value kdict = ci->stack[mrb_ci_kidx(ci)];
if (mrb_hash_p(kdict) && !mrb_hash_empty_p(mrb, kdict)) {
argc++;
}
}
int min = MRB_ASPEC_REQ(aspec) + MRB_ASPEC_POST(aspec);
int max = MRB_ASPEC_REST(aspec) ? -1 : min + MRB_ASPEC_OPT(aspec);
if (mrb_unlikely(argc < min || (max >= 0 && argc > max))) {
mrb_argnum_error(mrb, argc, min, max);
}
}
static mrb_value
exec_irep(mrb_state *mrb, mrb_value self, const struct RProc *p)
{
mrb_callinfo *ci = mrb->c->ci;
ci->stack[0] = self;
/* handle alias */
MRB_PROC_RESOLVE_ALIAS(ci, p);
CI_PROC_SET(ci, p);
if (MRB_PROC_CFUNC_P(p)) {
uint32_t caspec_bits = p->flags & MRB_PROC_CASPEC_MASK;
if (caspec_bits != 0) {
check_argument_count(mrb, ci, mrb_proc_decompress_caspec(caspec_bits));
}
else if (MRB_PROC_NOARG_P(p) && (ci->n > 0 || ci->nk > 0)) {
check_argument_count(mrb, ci, 0);
}
return MRB_PROC_CFUNC(p)(mrb, self);
}
mrb_int nregs = p->body.irep->nregs;
mrb_int keep = ci_bidx(ci)+1;
if (nregs < keep) {
stack_extend(mrb, keep);
}
else {
stack_extend(mrb, nregs);
stack_clear(ci->stack+keep, nregs-keep);
}
cipush(mrb, 0, 0, NULL, NULL, NULL, 0, 0);
return self;
}
mrb_value
mrb_exec_irep(mrb_state *mrb, mrb_value self, const struct RProc *p)
{
mrb_callinfo *ci = mrb->c->ci;
if (ci->cci == CINFO_NONE) {
return exec_irep(mrb, self, p);
}
else {
mrb_value ret;
if (MRB_PROC_CFUNC_P(p)) {
if (MRB_PROC_NOARG_P(p) && (ci->n > 0 || ci->nk > 0)) {
check_argument_count(mrb, ci, 0);
}
ci = cipush(mrb, 0, CINFO_DIRECT, CI_TARGET_CLASS(ci), p, NULL, ci->mid, ci->n|(ci->nk<<4));
mrb->exc = NULL;
ret = MRB_PROC_CFUNC(p)(mrb, self);
cipop(mrb);
}
else {
mrb_int keep = ci_bidx(ci) + 1; /* receiver + block */
ci = cipush(mrb, 0, CINFO_SKIP, CI_TARGET_CLASS(ci), p, NULL, ci->mid, ci->n|(ci->nk<<4));
ret = mrb_vm_run(mrb, p, self, keep);
}
if (mrb->exc && mrb->jmp) {
mrb_exc_raise(mrb, mrb_obj_value(mrb->exc));
}
return ret;
}
}
mrb_value
mrb_object_exec(mrb_state *mrb, mrb_value self, struct RClass *target_class)
{
mrb_callinfo *ci = mrb->c->ci;
mrb_int bidx = ci_bidx(ci);
mrb_value blk = ci->stack[bidx];
if (mrb_nil_p(blk)) {
mrb_raise(mrb, E_ARGUMENT_ERROR, "no block given");
}
mrb_assert(mrb_proc_p(blk));
mrb_gc_protect(mrb, blk);
ci->stack[bidx] = mrb_nil_value();
mrb_vm_ci_target_class_set(ci, target_class);
return mrb_exec_irep(mrb, self, mrb_proc_ptr(blk));
}
static mrb_noreturn void
vis_error(mrb_state *mrb, mrb_sym mid, mrb_value args, mrb_value recv, mrb_bool priv)
{
mrb_no_method_error(mrb, mid, args, "%s method '%n' called for %T", (priv ? "private" : "protected"), mid, recv);
}
static mrb_value
send_method(mrb_state *mrb, mrb_value self, mrb_bool pub)
{
mrb_callinfo *ci = mrb->c->ci;
int n = ci->n;
mrb_sym name;
if (ci->cci > CINFO_NONE) {
funcall:;
const mrb_value *argv;
mrb_int argc;
mrb_value block;
mrb_get_args(mrb, "n*&", &name, &argv, &argc, &block);
return mrb_funcall_with_block(mrb, self, name, argc, argv, block);
}
mrb_method_t m;
mrb_value *regs = mrb->c->ci->stack+1;
if (n == 0) {
argnum_error:
mrb_argnum_error(mrb, 0, 1, -1);
}
else if (n == 15) {
if (RARRAY_LEN(regs[0]) == 0) goto argnum_error;
name = mrb_obj_to_sym(mrb, RARRAY_PTR(regs[0])[0]);
}
else {
name = mrb_obj_to_sym(mrb, regs[0]);
}
struct RClass *c = mrb_class(mrb, self);
m = mrb_vm_find_method(mrb, c, &c, name);
if (MRB_METHOD_UNDEF_P(m)) { /* call method_missing */
goto funcall;
}
if (pub) {
mrb_bool priv = TRUE;
if (m.flags & MRB_METHOD_PRIVATE_FL) {
vis_err:;
if (n == 15) {
n = (int)(RARRAY_LEN(regs[0]) - 1);
regs = RARRAY_PTR(regs[0]);
}
vis_error(mrb, name, mrb_ary_new_from_values(mrb, n, regs+1), self, priv);
}
else if ((m.flags & MRB_METHOD_PROTECTED_FL) && mrb_obj_is_kind_of(mrb, self, ci->u.target_class)) {
priv = FALSE;
goto vis_err;
}
}
ci->mid = name;
ci->u.target_class = c;
/* remove first symbol from arguments */
if (n == 15) { /* variable length arguments */
regs[0] = mrb_ary_subseq(mrb, regs[0], 1, RARRAY_LEN(regs[0]) - 1);
}
else { /* n > 0 */
for (int i=0; i<n; i++) {
regs[i] = regs[i+1];
}
regs[n] = regs[n+1]; /* copy kdict or block */
if (ci->nk > 0) {
regs[n+1] = regs[n+2]; /* copy block */
}
ci->n--;
}
if (MRB_METHOD_FUNC_P(m)) {
check_argument_count(mrb, ci, MRB_MT_ASPEC(m.flags));
return MRB_METHOD_FUNC(m)(mrb, self);
}
const struct RProc *p = MRB_METHOD_PROC(m);
MRB_PROC_RESOLVE_ALIAS(ci, p);
CI_PROC_SET(ci, p);
if (MRB_PROC_CFUNC_P(p)) {
uint32_t caspec_bits = p->flags & MRB_PROC_CASPEC_MASK;
if (caspec_bits != 0) {
check_argument_count(mrb, ci, mrb_proc_decompress_caspec(caspec_bits));
}
else if (MRB_PROC_NOARG_P(p) && (ci->n > 0 || ci->nk > 0)) {
check_argument_count(mrb, ci, 0);
}
return MRB_PROC_CFUNC(p)(mrb, self);
}
return exec_irep(mrb, self, p);
}
/* 15.3.1.3.4 */
/* 15.3.1.3.44 */
/*
* call-seq:
* obj.send(symbol [, args...]) -> obj
* obj.__send__(symbol [, args...]) -> obj
*
* Invokes the method identified by _symbol_, passing it any
* arguments specified. You can use `__send__` if the name
* `send` clashes with an existing method in _obj_.
*
* class Klass
* def hello(*args)
* "Hello " + args.join(' ')
* end
* end
* k = Klass.new
* k.send :hello, "gentle", "readers" #=> "Hello gentle readers"
*/
mrb_value
mrb_f_send(mrb_state *mrb, mrb_value self)
{
return send_method(mrb, self, FALSE);
}
/*
* call-seq:
* obj.public_send(symbol [, args...]) -> obj
*
* Invokes the method identified by symbol, passing it any
* arguments specified. Unlike send, public_send calls public methods only.
* When the method is identified by a string, the string is converted to a
* symbol.
*
* 1.public_send(:puts, "hello") # causes NoMethodError
*/
mrb_value
mrb_f_public_send(mrb_state *mrb, mrb_value self)
{
return send_method(mrb, self, TRUE);
}
static void
check_block(mrb_state *mrb, mrb_value blk)
{
if (mrb_nil_p(blk)) {
mrb_raise(mrb, E_ARGUMENT_ERROR, "no block given");
}
if (!mrb_proc_p(blk)) {
mrb_raise(mrb, E_TYPE_ERROR, "not a block");
}
}
static mrb_value
eval_under(mrb_state *mrb, mrb_value self, mrb_value blk, struct RClass *c)
{
check_block(mrb, blk);
mrb_callinfo *ci = mrb->c->ci;
if (ci->cci == CINFO_DIRECT) {
return mrb_yield_with_class(mrb, blk, 1, &self, self, c);
}
ci->u.target_class = c;
const struct RProc *p = mrb_proc_ptr(blk);
/* just in case irep is NULL; #6065 */
if (p->body.irep == NULL) return mrb_nil_value();
CI_PROC_SET(ci, p);
ci->n = 1;
ci->nk = 0;
ci->mid = ci[-1].mid;
MRB_CI_SET_VISIBILITY_BREAK(ci);
if (MRB_PROC_CFUNC_P(p)) {
stack_extend(mrb, 4);
mrb->c->ci->stack[0] = self;
mrb->c->ci->stack[1] = self;
mrb->c->ci->stack[2] = mrb_nil_value();
return MRB_PROC_CFUNC(p)(mrb, self);
}
int nregs = p->body.irep->nregs;
if (nregs < 4) nregs = 4;
stack_extend(mrb, nregs);
mrb->c->ci->stack[0] = self;
mrb->c->ci->stack[1] = self;
stack_clear(mrb->c->ci->stack+2, nregs-2);
cipush(mrb, 0, 0, NULL, NULL, NULL, 0, 0);
return self;
}
/* 15.2.2.4.35 */
/*
* call-seq:
* mod.class_eval {| | block } -> obj
* mod.module_eval {| | block } -> obj
*
* Evaluates block in the context of _mod_. This can
* be used to add methods to a class. `module_eval` returns
* the result of evaluating its argument.
*/
mrb_value
mrb_mod_module_eval(mrb_state *mrb, mrb_value mod)
{
mrb_value a, b;
if (mrb_get_args(mrb, "|S&", &a, &b) == 1) {
mrb_raise(mrb, E_NOTIMP_ERROR, "module_eval/class_eval with string not implemented");
}
return eval_under(mrb, mod, b, mrb_class_ptr(mod));
}
/* 15.3.1.3.18 */
/*
* call-seq:
* obj.instance_eval {| | block } -> obj
*
* Evaluates the given block,within the context of the receiver (_obj_).
* In order to set the context, the variable `self` is set to _obj_ while
* the code is executing, giving the code access to _obj_'s
* instance variables. In the version of `instance_eval`
* that takes a `String`, the optional second and third
* parameters supply a filename and starting line number that are used
* when reporting compilation errors.
*
* class KlassWithSecret
* def initialize
* @secret = 99
* end
* end
* k = KlassWithSecret.new
* k.instance_eval { @secret } #=> 99
*/
mrb_value
mrb_obj_instance_eval(mrb_state *mrb, mrb_value self)
{
mrb_value a, b;
if (mrb_get_args(mrb, "|S&", &a, &b) == 1) {
mrb_raise(mrb, E_NOTIMP_ERROR, "instance_eval with string not implemented");
}
return eval_under(mrb, self, b, mrb_singleton_class_ptr(mrb, self));
}
static mrb_value
yield_with_attr(mrb_state *mrb, mrb_value b, mrb_int argc, const mrb_value *argv, mrb_value self, struct RClass *c,
mrb_bool vis_break)
{
check_block(mrb, b);
mrb_callinfo *ci = mrb->c->ci;
mrb_int n = mrb_ci_nregs(ci);
const struct RProc *p = mrb_proc_ptr(b);
mrb_sym mid;
if (MRB_PROC_ENV_P(p)) {
mid = p->e.env->mid;
}
else {
mid = ci->mid;
}
ci = cipush(mrb, n, CINFO_DIRECT, NULL, NULL, NULL, mid, 0);
funcall_args_capture(mrb, 0, argc, argv, mrb_nil_value(), ci);
ci->u.target_class = c;
ci->proc = p;
if (vis_break) {
MRB_CI_SET_VISIBILITY_BREAK(ci);
}
mrb_value val;
if (MRB_PROC_CFUNC_P(p)) {
mrb->exc = NULL;
ci->stack[0] = self;
val = MRB_PROC_CFUNC(p)(mrb, self);
cipop(mrb);
if (mrb->exc && mrb->jmp) {
mrb_exc_raise(mrb, mrb_obj_value(mrb->exc));
}
}
else {
ci->cci = CINFO_SKIP;
val = mrb_run(mrb, p, self);
}
return val;
}
/**
* @brief Yields to a block with a specific `self` object and class context.
*
* This function executes a given block (`b`) with the provided arguments (`argv`).
* The `self` object within the block will be `self`, and the class context
* will be `c`. This allows for more control over the execution environment of
* the block. The `vis_break` flag is set to TRUE, meaning visibility checks
* (public/private/protected) are enforced.
*
* @param mrb The mruby state.
* @param b The block (proc) to yield to.
* @param argc The number of arguments in `argv`.
* @param argv A pointer to an array of `mrb_value` arguments to pass to the block.
* @param self The object that will be `self` inside the block.
* @param c The class context for the block execution.
* @return The result of the block execution.
* @raise E_TYPE_ERROR if `b` is not a proc or nil.
* @see mrb_yield_argv
* @see mrb_yield
*/
MRB_API mrb_value
mrb_yield_with_class(mrb_state *mrb, mrb_value b, mrb_int argc, const mrb_value *argv, mrb_value self, struct RClass *c)
{
return yield_with_attr(mrb, b, argc, argv, self, c, TRUE);
}
/**
* @brief Yields to a block with an array of arguments.
*
* This function executes a given block (`b`) with the provided arguments (`argv`).
* The `self` object and class context for the block execution are determined
* from the block itself (its captured environment).
* Visibility checks (public/private/protected) are not strictly enforced
* in the same way as `mrb_yield_with_class` (vis_break is FALSE).
*
* @param mrb The mruby state.
* @param b The block (proc) to yield to.
* @param argc The number of arguments in `argv`.
* @param argv A pointer to an array of `mrb_value` arguments to pass to the block.
* @return The result of the block execution.
* @raise E_TYPE_ERROR if `b` is not a proc or nil.
* @see mrb_yield_with_class
* @see mrb_yield
*/
MRB_API mrb_value
mrb_yield_argv(mrb_state *mrb, mrb_value b, mrb_int argc, const mrb_value *argv)
{
const struct RProc *p = mrb_proc_ptr(b);
struct RClass *tc;
mrb_value self = mrb_proc_get_self(mrb, p, &tc);
return yield_with_attr(mrb, b, argc, argv, self, tc, FALSE);
}
/**
* @brief Yields to a block with a single argument.
*
* This function executes a given block (`b`) with a single argument (`arg`).
* It's a convenience function for the common case of yielding with one argument.
* The `self` object and class context for the block execution are determined
* from the block itself.
* Visibility checks are not strictly enforced (vis_break is FALSE).
*
* @param mrb The mruby state.
* @param b The block (proc) to yield to.
* @param arg The single `mrb_value` argument to pass to the block.
* @return The result of the block execution.
* @raise E_TYPE_ERROR if `b` is not a proc or nil.
* @see mrb_yield_with_class
* @see mrb_yield_argv
*/
MRB_API mrb_value
mrb_yield(mrb_state *mrb, mrb_value b, mrb_value arg)
{
const struct RProc *p = mrb_proc_ptr(b);
struct RClass *tc;
mrb_value self = mrb_proc_get_self(mrb, p, &tc);
return yield_with_attr(mrb, b, 1, &arg, self, tc, FALSE);
}
mrb_value
mrb_yield_cont(mrb_state *mrb, mrb_value b, mrb_value self, mrb_int argc, const mrb_value *argv)
{
check_block(mrb, b);
const struct RProc *p = mrb_proc_ptr(b);
mrb_callinfo *ci = mrb->c->ci;
stack_extend_adjust(mrb, 4, &argv);
mrb->c->ci->stack[1] = mrb_ary_new_from_values(mrb, argc, argv);
mrb->c->ci->stack[2] = mrb_nil_value();
mrb->c->ci->stack[3] = mrb_nil_value();
ci->n = 15;
ci->nk = 0;
return exec_irep(mrb, self, p);
}
#define RBREAK_TAG_FOREACH(f) \
f(RBREAK_TAG_BREAK, 0) \
f(RBREAK_TAG_JUMP, 1) \
f(RBREAK_TAG_STOP, 2)
#define RBREAK_TAG_DEFINE(tag, i) tag = i,
enum {
RBREAK_TAG_FOREACH(RBREAK_TAG_DEFINE)
};
#undef RBREAK_TAG_DEFINE
#define RBREAK_TAG_BIT 3
#define RBREAK_TAG_BIT_OFF 8
#define RBREAK_TAG_MASK (~(~UINT32_C(0) << RBREAK_TAG_BIT))
static inline uint32_t
mrb_break_tag_get(struct RBreak *brk)
{
return (brk->flags >> RBREAK_TAG_BIT_OFF) & RBREAK_TAG_MASK;
}
static inline void
mrb_break_tag_set(struct RBreak *brk, uint32_t tag)
{
brk->flags &= ~(RBREAK_TAG_MASK << RBREAK_TAG_BIT_OFF);
brk->flags |= (tag & RBREAK_TAG_MASK) << RBREAK_TAG_BIT_OFF;
}
static struct RBreak*
break_new(mrb_state *mrb, uint32_t tag, const mrb_callinfo *return_ci, mrb_value val)
{
mrb_assert((size_t)(return_ci - mrb->c->cibase) <= (size_t)(mrb->c->ci - mrb->c->cibase));
struct RBreak *brk = MRB_OBJ_ALLOC(mrb, MRB_TT_BREAK, NULL);
brk->ci_break_index = return_ci - mrb->c->cibase;
mrb_break_value_set(brk, val);
mrb_break_tag_set(brk, tag);
return brk;
}
#define MRB_CATCH_FILTER_RESCUE (UINT32_C(1) << MRB_CATCH_RESCUE)
#define MRB_CATCH_FILTER_ENSURE (UINT32_C(1) << MRB_CATCH_ENSURE)
#define MRB_CATCH_FILTER_ALL (MRB_CATCH_FILTER_RESCUE | MRB_CATCH_FILTER_ENSURE)
static const struct mrb_irep_catch_handler *
catch_handler_find(const mrb_irep *irep, const mrb_code *pc, uint32_t filter)
{
/* The comparison operators use `>` and `<=` because pc already points to the next instruction */
#define catch_cover_p(pc, beg, end) ((pc) > (ptrdiff_t)(beg) && (pc) <= (ptrdiff_t)(end))
mrb_assert(irep && irep->clen > 0);
ptrdiff_t xpc = pc - irep->iseq;
/* If it retry at the top level, pc will be 0, so check with -1 as the start position */
mrb_assert(catch_cover_p(xpc, -1, irep->ilen));
if (!catch_cover_p(xpc, -1, irep->ilen)) return NULL;
/* Currently uses a simple linear search to avoid processing complexity. */
size_t cnt = irep->clen;
const struct mrb_irep_catch_handler *e = mrb_irep_catch_handler_table(irep) + cnt - 1;
for (; cnt > 0; cnt--, e--) {
if (((UINT32_C(1) << e->type) & filter) &&
catch_cover_p(xpc, mrb_irep_catch_handler_unpack(e->begin), mrb_irep_catch_handler_unpack(e->end))) {
return e;
}
}
#undef catch_cover_p
return NULL;
}
#define RAISE_EXC(mrb, exc) do { \
mrb_value exc_value = (exc); \
mrb_exc_set(mrb, exc_value); \
goto L_RAISE; \
} while (0)
#define RAISE_LIT(mrb, c, str) RAISE_EXC(mrb, mrb_exc_new_lit(mrb, c, str))
#define RAISE_FORMAT(mrb, c, fmt, ...) RAISE_EXC(mrb, mrb_exc_new_str(mrb, c, mrb_format(mrb, fmt, __VA_ARGS__)))
/* return codes for extracted opcode handlers */
#define VM_NEXT 0 /* continue to next instruction */
#define VM_RAISE 1 /* exception: goto L_RAISE */
#define VM_SEND_SYM 2 /* fallback send: goto L_SEND_SYM */
#define VM_SENDB_SYM 3 /* fallback sendb: goto L_SENDB_SYM */
#define VM_RETURN_NIL 4 /* nil irep: return nil via L_OP_RETURN */
#if defined(__GNUC__) || defined(__clang__)
#define MRB_FLATTEN __attribute__((flatten))
#else
#define MRB_FLATTEN
#endif
static void
argnum_error(mrb_state *mrb, mrb_int num)
{
mrb_int argc = mrb->c->ci->n;
if (argc == 15) {
mrb_value args = mrb->c->ci->stack[1];
if (mrb_array_p(args)) {
argc = RARRAY_LEN(args);
}
}
if (argc == 0 && mrb->c->ci->nk != 0 && !mrb_hash_empty_p(mrb, mrb->c->ci->stack[1])) {
argc++;
}
mrb_value str = mrb_format(mrb, "wrong number of arguments (given %i, expected %i)", argc, num);
mrb_value exc = mrb_exc_new_str(mrb, E_ARGUMENT_ERROR, str);
mrb_exc_set(mrb, exc);
}
static mrb_bool
break_tag_p(struct RBreak *brk, uint32_t tag)
{
return (brk != NULL && brk->tt == MRB_TT_BREAK) ? TRUE : FALSE;
}
static void
prepare_tagged_break(mrb_state *mrb, uint32_t tag, const mrb_callinfo *return_ci, mrb_value val)
{
if (break_tag_p((struct RBreak*)mrb->exc, tag)) {
mrb_break_tag_set((struct RBreak*)mrb->exc, tag);
}
else {
mrb->exc = (struct RObject*)break_new(mrb, tag, return_ci, val);
}
}
#define THROW_TAGGED_BREAK(mrb, tag, return_ci, val) \
do { \
prepare_tagged_break(mrb, tag, return_ci, val); \
goto L_CATCH_TAGGED_BREAK; \
} while (0)
#define UNWIND_ENSURE(mrb, ci, pc, tag, return_ci, val) \
do { \
const struct RProc *proc = (ci)->proc; \
if (proc && !MRB_PROC_CFUNC_P(proc) && (irep = proc->body.irep) && irep->clen > 0 && \
(ch = catch_handler_find(irep, pc, MRB_CATCH_FILTER_ENSURE))) { \
THROW_TAGGED_BREAK(mrb, tag, return_ci, val); \
} \
} while (0)
/*
* CHECKPOINT_RESTORE(tag) {
* This part is executed when jumping by the same "tag" of RBreak (it is not executed the first time).
* Write the code required (initialization of variables, etc.) for the subsequent processing.
* }
* CHECKPOINT_MAIN(tag) {
* This part is always executed.
* }
* CHECKPOINT_END(tag);
*
* ...
*
* // Jump to CHECKPOINT_RESTORE with the same "tag".
* goto CHECKPOINT_LABEL_MAKE(tag);
*/
#define CHECKPOINT_LABEL_MAKE(tag) L_CHECKPOINT_ ## tag
#define CHECKPOINT_RESTORE(tag) \
do { \
if (FALSE) { \
CHECKPOINT_LABEL_MAKE(tag): \
do {
#define CHECKPOINT_MAIN(tag) \
} while (0); \
} \
do {
#define CHECKPOINT_END(tag) \
} while (0); \
} while (0)
#ifdef MRB_USE_DEBUG_HOOK
#define CODE_FETCH_HOOK(mrb, irep, pc, regs) if ((mrb)->code_fetch_hook) (mrb)->code_fetch_hook((mrb), (irep), (pc), (regs));
#else
#define CODE_FETCH_HOOK(mrb, irep, pc, regs)
#endif
#ifdef MRB_BYTECODE_DECODE_OPTION
#define BYTECODE_DECODER(x) ((mrb)->bytecode_decoder)?(mrb)->bytecode_decoder((mrb), (x)):(x)
#else
#define BYTECODE_DECODER(x) (x)
#endif
#ifndef MRB_USE_VM_SWITCH_DISPATCH
#if !defined __GNUC__ && !defined __clang__ && !defined __INTEL_COMPILER
#define MRB_USE_VM_SWITCH_DISPATCH
#endif
#endif /* ifndef MRB_USE_VM_SWITCH_DISPATCH */
#ifdef MRB_USE_VM_SWITCH_DISPATCH
#define INIT_DISPATCH for (;;) { CALL_CODE_HOOKS(); switch (insn) {
#define CASE(insn,ops) case insn: DECODE_OPERANDS(ops); L_ ## insn ## _BODY:
#define NEXT goto L_END_DISPATCH
#define JUMP NEXT
#define END_DISPATCH L_END_DISPATCH: RETURN_IF_TASK_STOPPED(mrb);}}
#else
#define INIT_DISPATCH JUMP; return mrb_nil_value();
#define CASE(insn,ops) L_ ## insn: DECODE_OPERANDS(ops); L_ ## insn ## _BODY:
#define NEXT RETURN_IF_TASK_STOPPED(mrb); CALL_CODE_HOOKS(); goto *optable[insn]
#define JUMP NEXT
#define END_DISPATCH RETURN_IF_TASK_STOPPED(mrb)
#endif
#define DECODE_OPERANDS(ops) do { const mrb_code *pc = ci->pc+1; FETCH_ ## ops (); ci->pc = pc; } while (0)
#define CALL_CODE_HOOKS() do { insn = BYTECODE_DECODER(*ci->pc); CODE_FETCH_HOOK(mrb, irep, ci->pc, regs); } while (0)
#ifdef MRB_USE_TASK_SCHEDULER
#define RETURN_IF_TASK_STOPPED(mrb) do { \
if ((mrb)->task.switching || (mrb)->c->status == MRB_TASK_STOPPED) \
return mrb_nil_value(); \
} while (0)
#define TASK_STOP(mrb) do { \
if (mrb->c->status != MRB_TASK_STOPPED) \
mrb->c->status = MRB_TASK_STOPPED; \
} while (0)
#else
#define RETURN_IF_TASK_STOPPED(mrb)
#define TASK_STOP(mrb)
#endif
/**
* @brief Executes a mruby bytecode sequence (iseq) within the VM.
*
* This function is a core part of the mruby execution process. It sets up
* the VM environment for executing the bytecode instructions associated with
* the given proc (Ruby procedure/method).
*
* It initializes the stack if necessary, extends it to accommodate the
* required number of registers for the proc, and then calls `mrb_vm_exec`
* to actually execute the bytecode.
*
* @param mrb The mruby state.
* @param proc The RProc object containing the bytecode (iseq) to execute.
* This proc represents a Ruby method or block.
* @param self The `self` object for the context of this execution.
* @param stack_keep The number of values to preserve on the stack from the
* previous context. This is used for managing nested calls
* and ensuring that arguments or local variables from the
* caller are accessible if needed, or that the stack is
* correctly cleared.
* @return The result of the bytecode execution (typically the value of the
* last evaluated expression).
* @see mrb_vm_exec
* @see mrb_top_run
*/
MRB_API mrb_value
mrb_vm_run(mrb_state *mrb, const struct RProc *proc, mrb_value self, mrb_int stack_keep)
{
const mrb_irep *irep = proc->body.irep;
struct mrb_context *c = mrb->c;
#ifdef MRB_DEBUG
ptrdiff_t cioff = c->ci - c->cibase;
#endif
mrb_int nregs = irep->nregs;
if (!c->stbase) {
stack_init(mrb);
}
if (stack_keep > nregs)
nregs = stack_keep;
else {
struct REnv *e = CI_ENV(mrb->c->ci);
if (e && (stack_keep == 0 || irep->nlocals < MRB_ENV_LEN(e))) {
ci_env_set(mrb->c->ci, NULL);
mrb_env_unshare(mrb, e, FALSE);
}
}
stack_extend(mrb, nregs);
stack_clear(c->ci->stack + stack_keep, nregs - stack_keep);
c->ci->stack[0] = self;
mrb_value result = mrb_vm_exec(mrb, proc, irep->iseq);
mrb_assert(mrb->c == c); /* do not switch fibers via mrb_vm_run(), unlike mrb_vm_exec() */
mrb_assert(c->ci == c->cibase || (c->ci - c->cibase) == cioff - 1);
return result;
}
static struct RClass*
check_target_class(mrb_state *mrb)
{
struct RClass *target = CI_TARGET_CLASS(mrb->c->ci);
if (!target) {
mrb_raise(mrb, E_TYPE_ERROR, "no class/module to add method");
}
return target;
}
#define regs (ci->stack)
static mrb_value
hash_new_from_regs(mrb_state *mrb, mrb_int argc, mrb_int idx)
{
mrb_value hash = mrb_hash_new_capa(mrb, argc);
mrb_callinfo *ci = mrb->c->ci;
while (argc--) {
mrb_hash_set(mrb, hash, regs[idx+0], regs[idx+1]);
ci = mrb->c->ci;
idx += 2;
}
return hash;
}
#define ary_new_from_regs(mrb, argc, idx) mrb_ary_new_from_values(mrb, (argc), &regs[idx]);
/* type pair for arithmetic/comparison dispatch */
#define TYPES2(a,b) ((((uint16_t)(a))<<8)|(((uint16_t)(b))&0xff))
/*
* Extracted opcode handlers.
* These are static functions force-inlined back into mrb_vm_exec via
* __attribute__((flatten)). The source stays clean while the compiled
* output is identical to having the code inline.
*
* Return VM_NEXT to continue, VM_RAISE when an exception has been set.
* VM_SEND_SYM/VM_SENDB_SYM for method fallback (mid set via out-param).
*/
static int
vm_op_blkpush(mrb_state *mrb, uint32_t a, uint16_t b)
{
mrb_callinfo *ci = mrb->c->ci;
int m1 = (b>>11)&0x3f;
int r = (b>>10)&0x1;
int m2 = (b>>5)&0x1f;
int kd = (b>>4)&0x1;
int lv = (b>>0)&0xf;
int offset = m1+r+m2+kd;
mrb_value *stack;
if (lv == 0) stack = regs + 1;
else {
struct REnv *e = uvenv(mrb, lv-1);
if (!e || (!MRB_ENV_ONSTACK_P(e) && e->mid == 0) ||
MRB_ENV_LEN(e) <= offset+1) {
mrb_exc_set(mrb, mrb_exc_new_lit(mrb, E_LOCALJUMP_ERROR, "unexpected yield"));
return VM_RAISE;
}
stack = e->stack + 1;
}
if (mrb_nil_p(stack[offset])) {
mrb_exc_set(mrb, mrb_exc_new_lit(mrb, E_LOCALJUMP_ERROR, "unexpected yield"));
return VM_RAISE;
}
regs[a] = stack[offset];
return VM_NEXT;
}
static int
vm_op_argary(mrb_state *mrb, uint32_t a, uint16_t b)
{
mrb_callinfo *ci = mrb->c->ci;
mrb_int m1 = (b>>11)&0x3f;
mrb_int r = (b>>10)&0x1;
mrb_int m2 = (b>>5)&0x1f;
mrb_int kd = (b>>4)&0x1;
mrb_int lv = (b>>0)&0xf;
mrb_value *stack;
if (ci->mid == 0 || CI_TARGET_CLASS(ci) == NULL) {
L_NOSUPER:
mrb_exc_set(mrb, mrb_exc_new_lit(mrb, E_NOMETHOD_ERROR, "super called outside of method"));
return VM_RAISE;
}
if (lv == 0) stack = regs + 1;
else {
struct REnv *e = uvenv(mrb, lv-1);
if (!e) goto L_NOSUPER;
if (MRB_ENV_LEN(e) <= m1+r+m2+1)
goto L_NOSUPER;
stack = e->stack + 1;
}
if (r == 0) {
regs[a] = mrb_ary_new_from_values(mrb, m1+m2, stack);
}
else {
mrb_value *pp = NULL;
struct RArray *rest;
mrb_int len = 0;
if (mrb_array_p(stack[m1])) {
struct RArray *ary = mrb_ary_ptr(stack[m1]);
pp = ARY_PTR(ary);
len = ARY_LEN(ary);
}
regs[a] = mrb_ary_new_capa(mrb, m1+len+m2);
rest = mrb_ary_ptr(regs[a]);
if (m1 > 0) {
stack_copy(ARY_PTR(rest), stack, m1);
}
if (len > 0) {
stack_copy(ARY_PTR(rest)+m1, pp, len);
}
if (m2 > 0) {
stack_copy(ARY_PTR(rest)+m1+len, stack+m1+1, m2);
}
ARY_SET_LEN(rest, m1+len+m2);
}
if (kd) {
regs[a+1] = stack[m1+r+m2];
regs[a+2] = stack[m1+r+m2+1];
}
else {
regs[a+1] = stack[m1+r+m2];
}
return VM_NEXT;
}
static int
vm_op_enter(mrb_state *mrb, uint32_t a)
{
mrb_callinfo *ci = mrb->c->ci;
const mrb_irep *irep = ci->proc->body.irep;
mrb_int argc = ci->n;
mrb_value *argv = regs+1;
mrb_int m1 = MRB_ASPEC_REQ(a);
/* no other args */
if ((a & ~0x7c0001) == 0 && argc < 15 && MRB_PROC_STRICT_P(ci->proc)) {
if (mrb_unlikely(argc+(ci->nk==15) != m1)) { /* count kdict too */
argnum_error(mrb, m1);
return VM_RAISE;
}
/* clear local (but non-argument) variables */
mrb_int pos = m1+2; /* self+m1+blk */
if (irep->nlocals-pos > 0) {
stack_clear(&regs[pos], irep->nlocals-pos);
}
return VM_NEXT;
}
mrb_int o = MRB_ASPEC_OPT(a);
mrb_int r = MRB_ASPEC_REST(a);
mrb_int m2 = MRB_ASPEC_POST(a);
mrb_int kd = (MRB_ASPEC_KEY(a) > 0 || MRB_ASPEC_KDICT(a))? 1 : 0;
/* unused
int b = MRB_ASPEC_BLOCK(a);
*/
mrb_int const len = m1 + o + r + m2;
mrb_value * const argv0 = argv;
mrb_value blk = regs[ci_bidx(ci)];
/* &nil: reject block */
if (MRB_ASPEC_NOBLOCK(a) && !mrb_nil_p(blk)) {
mrb_exc_set(mrb, mrb_exc_new_lit(mrb, E_ARGUMENT_ERROR, "no block accepted"));
return VM_RAISE;
}
mrb_value kdict = mrb_nil_value();
/* keyword arguments */
if (ci->nk == 15) {
kdict = regs[mrb_ci_kidx(ci)];
}
if (!kd) {
if (!mrb_nil_p(kdict) && mrb_hash_p(kdict) && mrb_hash_size(mrb, kdict) > 0) {
if (argc < 14) {
ci->n++;
argc++; /* include kdict in normal arguments */
}
else if (argc == 14) {
/* pack arguments and kdict */
regs[1] = mrb_ary_new_from_values(mrb, argc+1, &regs[1]);
argc = ci->n = 15;
}
else {/* argc == 15 */
/* push kdict to packed arguments */
mrb_ary_push(mrb, regs[1], kdict);
}
}
kdict = mrb_nil_value();
ci->nk = 0;
}
else if (!mrb_nil_p(kdict)) {
mrb_gc_protect(mrb, kdict);
}
/* arguments is passed with Array */
if (argc == 15) {
struct RArray *ary = mrb_ary_ptr(regs[1]);
argv = ARY_PTR(ary);
argc = (int)ARY_LEN(ary);
mrb_gc_protect(mrb, regs[1]);
}
/* strict argument check */
if (ci->proc && MRB_PROC_STRICT_P(ci->proc)) {
if (mrb_unlikely(argc < m1 + m2 || (r == 0 && argc > len))) {
argnum_error(mrb, m1+m2);
return VM_RAISE;
}
}
/* extract first argument array to arguments */
else if (len > 1 && argc == 1 && mrb_array_p(argv[0])) {
mrb_gc_protect(mrb, argv[0]);
argc = (int)RARRAY_LEN(argv[0]);
argv = RARRAY_PTR(argv[0]);
}
/* rest arguments */
mrb_value rest;
if (argc < len) {
mrb_int mlen = m2;
if (argc < m1+m2) {
mlen = m1 < argc ? argc - m1 : 0;
}
/* copy mandatory and optional arguments */
if (argv0 != argv && argv) {
value_move(&regs[1], argv, argc-mlen); /* m1 + o */
}
if (argc < m1) {
stack_clear(&regs[argc+1], m1-argc);
}
/* copy post mandatory arguments */
if (mlen) {
value_move(&regs[len-m2+1], &argv[argc-mlen], mlen);
}
if (mlen < m2) {
stack_clear(&regs[len-m2+mlen+1], m2-mlen);
}
/* initialize rest arguments with empty Array */
if (r) {
rest = mrb_ary_new_capa(mrb, 0);
regs[m1+o+1] = rest;
}
/* skip initializer of passed arguments */
if (o > 0 && argc > m1+m2)
ci->pc += (argc - m1 - m2)*3;
}
else {
mrb_int rnum = 0;
if (argv0 != argv) {
mrb_gc_protect(mrb, blk);
value_move(&regs[1], argv, m1+o);
}
if (r) {
rnum = argc-m1-o-m2;
rest = mrb_ary_new_from_values(mrb, rnum, argv+m1+o);
regs[m1+o+1] = rest;
}
if (m2 > 0 && argc-m2 > m1) {
value_move(&regs[m1+o+r+1], &argv[m1+o+rnum], m2);
}
ci->pc += o*3;
}
/* need to be update blk first to protect blk from GC */
mrb_int const kw_pos = len + kd; /* where kwhash should be */
mrb_int const blk_pos = kw_pos + 1; /* where block should be */
regs[blk_pos] = blk; /* move block */
if (kd) {
if (mrb_nil_p(kdict)) {
kdict = mrb_hash_new_capa(mrb, 0);
}
regs[kw_pos] = kdict; /* set kwhash */
ci->nk = 15;
}
/* format arguments for generated code */
ci->n = (uint8_t)len;
/* clear local (but non-argument) variables */
if (irep->nlocals-blk_pos-1 > 0) {
stack_clear(&regs[blk_pos+1], irep->nlocals-blk_pos-1);
}
return VM_NEXT;
}
static int
vm_op_getidx(mrb_state *mrb, uint32_t a, mrb_sym *midp)
{
mrb_callinfo *ci = mrb->c->ci;
mrb_value va = regs[a], vb = regs[a+1];
enum mrb_vtype tt = mrb_type(va);
/* Array case is most common - check first with branch hint */
if (mrb_likely(tt == MRB_TT_ARRAY)) {
struct RArray *ary = mrb_ary_ptr(va);
/* optimize only for Array class; subclasses/singleton may override [] */
if (mrb_unlikely(ary->c != mrb->array_class)) goto getidx_fallback;
if (mrb_likely(mrb_integer_p(vb))) {
mrb_int idx = mrb_integer(vb);
mrb_int len;
mrb_value *ptr;
/* Single ARY_EMBED_P check instead of two */
#ifndef MRB_ARY_NO_EMBED
if (ARY_EMBED_P(ary)) {
len = ARY_EMBED_LEN(ary);
ptr = ary->as.ary;
}
else
#endif
{
len = ary->as.heap.len;
ptr = ary->as.heap.ptr;
}
/* Unsigned comparison: handles negative idx as large positive */
if (mrb_likely((mrb_uint)idx < (mrb_uint)len)) {
regs[a] = ptr[idx];
}
else {
regs[a] = mrb_ary_entry(va, idx);
}
return VM_NEXT;
}
goto getidx_fallback;
}
else if (tt == MRB_TT_HASH) {
/* optimize only for Hash class; subclasses/singleton may override [] */
if (mrb_obj_ptr(va)->c != mrb->hash_class) goto getidx_fallback;
va = mrb_hash_get(mrb, va, vb);
ci = mrb->c->ci;
regs[a] = va;
return VM_NEXT;
}
else if (tt == MRB_TT_STRING) {
/* optimize only for String class; subclasses/singleton may override [] */
if (mrb_obj_ptr(va)->c != mrb->string_class) goto getidx_fallback;
switch (mrb_type(vb)) {
case MRB_TT_INTEGER:
case MRB_TT_STRING:
case MRB_TT_RANGE:
va = mrb_str_aref(mrb, va, vb, mrb_undef_value());
regs[a] = va;
return VM_NEXT;
default:
break;
}
}
getidx_fallback:
*midp = MRB_OPSYM(aref);
return VM_SEND_SYM;
}
static int
vm_op_getidx0(mrb_state *mrb, uint32_t a, uint16_t b, mrb_sym *midp)
{
mrb_callinfo *ci = mrb->c->ci;
mrb_value recv = regs[b];
enum mrb_vtype tt = mrb_type(recv);
if (mrb_likely(tt == MRB_TT_ARRAY)) {
struct RArray *ary = mrb_ary_ptr(recv);
if (mrb_unlikely(ary->c != mrb->array_class)) goto getidx0_fallback;
#ifndef MRB_ARY_NO_EMBED
if (ARY_EMBED_P(ary)) {
regs[a] = ARY_EMBED_LEN(ary) > 0 ? ary->as.ary[0] : mrb_nil_value();
}
else
#endif
{
regs[a] = ary->as.heap.len > 0 ? ary->as.heap.ptr[0] : mrb_nil_value();
}
return VM_NEXT;
}
else if (tt == MRB_TT_HASH) {
if (mrb_obj_ptr(recv)->c != mrb->hash_class) goto getidx0_fallback;
regs[a] = mrb_hash_get(mrb, recv, mrb_fixnum_value(0));
return VM_NEXT;
}
getidx0_fallback:
regs[a] = recv;
SET_FIXNUM_VALUE(regs[a+1], 0);
*midp = MRB_OPSYM(aref);
return VM_SEND_SYM;
}
static int
vm_op_setidx(mrb_state *mrb, uint32_t a, mrb_sym *midp)
{
mrb_callinfo *ci = mrb->c->ci;
mrb_value va = regs[a], vb = regs[a+1], vc = regs[a+2];
switch (mrb_type(va)) {
case MRB_TT_ARRAY:
/* optimize only for Array class; subclasses/singleton may override []= */
if (mrb_obj_ptr(va)->c != mrb->array_class) goto setidx_fallback;
if (!mrb_integer_p(vb)) goto setidx_fallback;
mrb_ary_set(mrb, va, mrb_integer(vb), vc);
ci = mrb->c->ci;
regs[a] = vc;
return VM_NEXT;
case MRB_TT_HASH:
/* optimize only for Hash class; subclasses/singleton may override []= */
if (mrb_obj_ptr(va)->c != mrb->hash_class) goto setidx_fallback;
mrb_hash_set(mrb, va, vb, vc);
ci = mrb->c->ci;
regs[a] = vc;
return VM_NEXT;
default:
setidx_fallback:
SET_NIL_VALUE(regs[a+3]);
*midp = MRB_OPSYM(aset);
return VM_SENDB_SYM;
}
}
static int
vm_op_div(mrb_state *mrb, uint32_t a, mrb_sym *midp)
{
mrb_callinfo *ci = mrb->c->ci;
#ifndef MRB_NO_FLOAT
mrb_float x, y, f;
#endif
/* need to check if op is overridden */
switch (TYPES2(mrb_type(regs[a]),mrb_type(regs[a+1]))) {
case TYPES2(MRB_TT_INTEGER,MRB_TT_INTEGER):
{
mrb_int x = mrb_integer(regs[a]);
mrb_int y = mrb_integer(regs[a+1]);
regs[a] = mrb_div_int_value(mrb, x, y);
}
return VM_NEXT;
#ifndef MRB_NO_FLOAT
case TYPES2(MRB_TT_INTEGER,MRB_TT_FLOAT):
x = (mrb_float)mrb_integer(regs[a]);
y = mrb_float(regs[a+1]);
break;
case TYPES2(MRB_TT_FLOAT,MRB_TT_INTEGER):
x = mrb_float(regs[a]);
y = (mrb_float)mrb_integer(regs[a+1]);
break;
case TYPES2(MRB_TT_FLOAT,MRB_TT_FLOAT):
x = mrb_float(regs[a]);
y = mrb_float(regs[a+1]);
break;
#endif
default:
*midp = MRB_OPSYM(div);
return VM_SEND_SYM;
}
#ifndef MRB_NO_FLOAT
f = mrb_div_float(x, y);
SET_FLOAT_VALUE(mrb, regs[a], f);
#endif
return VM_NEXT;
}
static mrb_sym
vm_define_method(mrb_state *mrb, struct RClass *tc, const mrb_irep *irep, uint16_t b, uint16_t c)
{
struct RProc *p = mrb_proc_new(mrb, irep->reps[c]);
mrb_sym mid = irep->syms[b];
mrb_method_t m;
p->flags |= MRB_PROC_SCOPE | MRB_PROC_STRICT;
MRB_METHOD_FROM_PROC(m, p);
MRB_METHOD_SET_VISIBILITY(m, MRB_METHOD_VDEFAULT_FL);
mrb_define_method_raw(mrb, tc, mid, m);
mrb_method_added(mrb, tc, mid);
return mid;
}
/* Common proc dispatch for OP_CALL and OP_BLKCALL.
Returns VM_NEXT, VM_RAISE, or VM_RETURN_NIL. */
static int
vm_call_proc(mrb_state *mrb, const struct RProc *p, mrb_int nargs,
const mrb_irep **irepp, int ai)
{
mrb_callinfo *ci = mrb->c->ci;
mrb_value recv = ci->stack[0];
/* handle alias */
MRB_PROC_RESOLVE_ALIAS(ci, p);
if (MRB_PROC_ENV_P(p)) {
ci->mid = MRB_PROC_ENV(p)->mid;
}
ci->u.target_class = MRB_PROC_TARGET_CLASS(p);
CI_PROC_SET(ci, p);
if (MRB_PROC_CFUNC_P(p)) {
recv = MRB_PROC_CFUNC(p)(mrb, recv);
mrb_gc_arena_shrink(mrb, ai);
if (mrb_unlikely(mrb->exc)) return VM_RAISE;
ci = cipop(mrb);
ci[1].stack[0] = recv;
*irepp = ci->proc->body.irep;
}
else {
const mrb_irep *irep = p->body.irep;
if (!irep) {
ci->stack[0] = mrb_nil_value();
return VM_RETURN_NIL;
}
if (nargs < irep->nregs) {
stack_extend(mrb, irep->nregs);
stack_clear(ci->stack+nargs, irep->nregs-nargs);
}
if (MRB_PROC_ENV_P(p)) {
ci->stack[0] = MRB_PROC_ENV(p)->stack[0];
}
ci->pc = irep->iseq;
*irepp = irep;
}
return VM_NEXT;
}
/**
* @brief Executes a sequence of mruby bytecode instructions.
*
* This is the main bytecode interpreter loop. It takes a starting proc
* (`begin_proc`) and a pointer to the initial instruction (`iseq`) within
* that proc's instruction sequence. It then enters a loop, fetching and
* dispatching bytecode operations until an OP_STOP instruction is encountered,
* an exception occurs, or a C function call returns.
*
* This function handles the low-level details of instruction decoding,
* stack manipulation, exception handling (try/catch blocks within mruby code),
* and calling C functions or other mruby methods.
*
* @param mrb The mruby state.
* @param begin_proc The initial RProc whose bytecode is to be executed.
* While the name suggests it's the "beginning" proc,
* execution might involve other procs called from this one.
* @param iseq A pointer to the first bytecode instruction to execute within
* `begin_proc`'s instruction sequence.
* @return The result of the execution. This could be the return value of
* the executed Ruby code, an exception object if an unhandled
* exception occurred, or the result of a fiber switch.
* @note This function is highly complex and central to mruby's operation.
* It uses a jump table (`optable`) for efficient instruction dispatch
* when not using switch-based dispatch. It also manages the callinfo
* stack (`ci`) for tracking method/block calls.
*/
MRB_FLATTEN MRB_API mrb_value
mrb_vm_exec(mrb_state *mrb, const struct RProc *begin_proc, const mrb_code *iseq)
{
/* mrb_assert(MRB_PROC_CFUNC_P(begin_proc)) */
const mrb_irep *irep = begin_proc->body.irep;
mrb_code insn;
int ai = mrb_gc_arena_save(mrb);
struct mrb_jmpbuf *prev_jmp = mrb->jmp;
struct mrb_jmpbuf c_jmp;
uint32_t a;
uint16_t b;
uint16_t c;
mrb_sym mid;
const struct mrb_irep_catch_handler *ch;
#ifndef MRB_USE_VM_SWITCH_DISPATCH
static const void * const optable[] = {
#define OPCODE(x,_) &&L_OP_ ## x,
#include <mruby/ops.h>
#undef OPCODE
};
#endif
mrb->exc = NULL;
mrb_callinfo *ci = mrb->c->ci;
CI_PROC_SET(ci, begin_proc);
ci->pc = iseq;
RETRY_TRY_BLOCK:
MRB_TRY(&c_jmp) {
if (mrb_unlikely(mrb->exc)) {
mrb_gc_arena_restore(mrb, ai);
if (mrb->exc->tt == MRB_TT_BREAK)
goto L_BREAK;
goto L_RAISE;
}
/* Intentionally store stack variable address for exception handling.
* This is safe because the pointer is cleared before function returns.
* Suppress GCC 12+ warning about dangling pointer. */
#if defined(__GNUC__) && !defined(__clang__)
#if __GNUC__ >= 12
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wdangling-pointer"
#endif
#endif
mrb->jmp = &c_jmp;
#if defined(__GNUC__) && !defined(__clang__)
#if __GNUC__ >= 12
#pragma GCC diagnostic pop
#endif
#endif
INIT_DISPATCH {
CASE(OP_NOP, Z) {
/* do nothing */
NEXT;
}
CASE(OP_MOVE, BB) {
regs[a] = regs[b];
NEXT;
}
CASE(OP_LOADL, BB) {
switch (irep->pool[b].tt) { /* number */
case IREP_TT_INT32:
regs[a] = mrb_int_value(mrb, (mrb_int)irep->pool[b].u.i32);
break;
case IREP_TT_INT64:
#if defined(MRB_INT64)
regs[a] = mrb_int_value(mrb, (mrb_int)irep->pool[b].u.i64);
break;
#else
#if defined(MRB_64BIT)
if (INT32_MIN <= irep->pool[b].u.i64 && irep->pool[b].u.i64 <= INT32_MAX) {
regs[a] = mrb_int_value(mrb, (mrb_int)irep->pool[b].u.i64);
break;
}
#endif
goto L_INT_OVERFLOW;
#endif
case IREP_TT_BIGINT:
#ifdef MRB_USE_BIGINT
{
const char *s = irep->pool[b].u.str;
regs[a] = mrb_bint_new_str(mrb, s+2, (uint8_t)s[0], (int8_t)s[1]);
}
break;
#else
goto L_INT_OVERFLOW;
#endif
#ifndef MRB_NO_FLOAT
case IREP_TT_FLOAT:
regs[a] = mrb_float_value(mrb, irep->pool[b].u.f);
break;
#endif
default:
/* should not happen (tt:string) */
regs[a] = mrb_nil_value();
break;
}
NEXT;
}
CASE(OP_LOADI8, BB) {
SET_FIXNUM_VALUE(regs[a], b);
NEXT;
}
CASE(OP_LOADINEG, BB) {
SET_FIXNUM_VALUE(regs[a], -b);
NEXT;
}
CASE(OP_LOADI__1,B) goto L_LOADI;
CASE(OP_LOADI_0,B) goto L_LOADI;
CASE(OP_LOADI_1,B) goto L_LOADI;
CASE(OP_LOADI_2,B) goto L_LOADI;
CASE(OP_LOADI_3,B) goto L_LOADI;
CASE(OP_LOADI_4,B) goto L_LOADI;
CASE(OP_LOADI_5,B) goto L_LOADI;
CASE(OP_LOADI_6,B) goto L_LOADI;
CASE(OP_LOADI_7, B) {
L_LOADI:
SET_FIXNUM_VALUE(regs[a], (mrb_int)insn - (mrb_int)OP_LOADI_0);
NEXT;
}
CASE(OP_LOADI16, BS) {
SET_FIXNUM_VALUE(regs[a], (mrb_int)(int16_t)b);
NEXT;
}
CASE(OP_LOADI32, BSS) {
SET_INT_VALUE(mrb, regs[a], (int32_t)(((uint32_t)b<<16)+c));
NEXT;
}
CASE(OP_LOADSYM, BB) {
SET_SYM_VALUE(regs[a], irep->syms[b]);
NEXT;
}
CASE(OP_LOADNIL, B) {
SET_NIL_VALUE(regs[a]);
NEXT;
}
CASE(OP_LOADSELF, B) {
regs[a] = regs[0];
NEXT;
}
CASE(OP_LOADTRUE, B) {
SET_TRUE_VALUE(regs[a]);
NEXT;
}
CASE(OP_LOADFALSE, B) {
SET_FALSE_VALUE(regs[a]);
NEXT;
}
CASE(OP_GETGV, BB) {
mrb_value val = mrb_gv_get(mrb, irep->syms[b]);
ci = mrb->c->ci;
regs[a] = val;
NEXT;
}
CASE(OP_SETGV, BB) {
mrb_gv_set(mrb, irep->syms[b], regs[a]);
ci = mrb->c->ci;
NEXT;
}
CASE(OP_GETSV, BB) {
mrb_value val = mrb_vm_special_get(mrb, irep->syms[b]);
ci = mrb->c->ci;
regs[a] = val;
NEXT;
}
CASE(OP_SETSV, BB) {
mrb_vm_special_set(mrb, irep->syms[b], regs[a]);
ci = mrb->c->ci;
NEXT;
}
CASE(OP_GETIV, BB) {
regs[a] = mrb_iv_get(mrb, regs[0], irep->syms[b]);
ci = mrb->c->ci;
NEXT;
}
CASE(OP_SETIV, BB) {
mrb_iv_set(mrb, regs[0], irep->syms[b], regs[a]);
ci = mrb->c->ci;
NEXT;
}
CASE(OP_GETCV, BB) {
mrb_value val;
val = mrb_vm_cv_get(mrb, irep->syms[b]);
ci = mrb->c->ci;
regs[a] = val;
NEXT;
}
CASE(OP_SETCV, BB) {
mrb_vm_cv_set(mrb, irep->syms[b], regs[a]);
ci = mrb->c->ci;
NEXT;
}
CASE(OP_GETIDX, B) {
int r = vm_op_getidx(mrb, a, &mid);
ci = mrb->c->ci;
if (r == VM_SEND_SYM) goto L_SEND_SYM;
NEXT;
}
CASE(OP_GETIDX0, BB) {
int r = vm_op_getidx0(mrb, a, b, &mid);
ci = mrb->c->ci;
if (r == VM_SEND_SYM) goto L_SEND_SYM;
NEXT;
}
CASE(OP_SETIDX, B) {
int r = vm_op_setidx(mrb, a, &mid);
ci = mrb->c->ci;
if (r == VM_SENDB_SYM) { c = 2; goto L_SENDB_SYM; }
NEXT;
}
CASE(OP_GETCONST, BB) {
#ifndef MRB_NO_CONST_CACHE
mrb_sym sym = irep->syms[b];
uint32_t h = mrb_int_hash_func(mrb, ((intptr_t)irep) ^ sym) & (MRB_CONST_CACHE_SIZE-1);
struct mrb_const_cache_entry *cc = &mrb->const_cache[h];
if (cc->irep == irep && cc->sym == sym) {
regs[a] = cc->value;
NEXT;
}
#endif
{
mrb_value v = mrb_vm_const_get(mrb, irep->syms[b]);
ci = mrb->c->ci;
regs[a] = v;
#ifndef MRB_NO_CONST_CACHE
cc->irep = irep;
cc->sym = sym;
cc->value = v;
#endif
}
NEXT;
}
CASE(OP_SETCONST, BB) {
ci = mrb->c->ci;
struct RClass *c = MRB_PROC_TARGET_CLASS(ci->proc);
if (!c) c = mrb->object_class;
mrb_const_set(mrb, mrb_obj_value(c), irep->syms[b], regs[a]);
NEXT;
}
CASE(OP_GETMCNST, BB) {
mrb_value v = mrb_const_get(mrb, regs[a], irep->syms[b]);
ci = mrb->c->ci;
regs[a] = v;
NEXT;
}
CASE(OP_SETMCNST, BB) {
mrb_const_set(mrb, regs[a+1], irep->syms[b], regs[a]);
ci = mrb->c->ci;
NEXT;
}
CASE(OP_GETUPVAR, BBB) {
struct REnv *e = uvenv(mrb, c);
if (e && b < MRB_ENV_LEN(e)) {
regs[a] = e->stack[b];
}
else {
regs[a] = mrb_nil_value();
}
NEXT;
}
CASE(OP_SETUPVAR, BBB) {
struct REnv *e = uvenv(mrb, c);
if (e) {
if (b < MRB_ENV_LEN(e)) {
e->stack[b] = regs[a];
mrb_write_barrier(mrb, (struct RBasic*)e);
}
}
NEXT;
}
CASE(OP_JMP, S) {
ci->pc += (int16_t)a;
JUMP;
}
CASE(OP_JMPIF, BS) {
if (mrb_test(regs[a])) {
ci->pc += (int16_t)b;
JUMP;
}
NEXT;
}
CASE(OP_JMPNOT, BS) {
if (!mrb_test(regs[a])) {
ci->pc += (int16_t)b;
JUMP;
}
NEXT;
}
CASE(OP_JMPNIL, BS) {
if (mrb_nil_p(regs[a])) {
ci->pc += (int16_t)b;
JUMP;
}
NEXT;
}
CASE(OP_JMPUW, S) {
a = (uint32_t)((ci->pc - irep->iseq) + (int16_t)a);
CHECKPOINT_RESTORE(RBREAK_TAG_JUMP) {
struct RBreak *brk = (struct RBreak*)mrb->exc;
mrb_value target = mrb_break_value_get(brk);
mrb_assert(mrb_integer_p(target));
a = (uint32_t)mrb_integer(target);
mrb_assert(a >= 0 && a < irep->ilen);
}
CHECKPOINT_MAIN(RBREAK_TAG_JUMP) {
if (irep->clen > 0 &&
(ch = catch_handler_find(irep, ci->pc, MRB_CATCH_FILTER_ENSURE))) {
/* avoiding a jump from a catch handler into the same handler */
if (a < mrb_irep_catch_handler_unpack(ch->begin) || a > mrb_irep_catch_handler_unpack(ch->end)) {
THROW_TAGGED_BREAK(mrb, RBREAK_TAG_JUMP, mrb->c->ci, mrb_fixnum_value(a));
}
}
}
CHECKPOINT_END(RBREAK_TAG_JUMP);
mrb->exc = NULL; /* clear break object */
ci->pc = irep->iseq + a;
JUMP;
}
CASE(OP_EXCEPT, B) {
mrb_value exc;
if (mrb->exc == NULL) {
exc = mrb_nil_value();
}
else {
switch (mrb->exc->tt) {
case MRB_TT_BREAK:
case MRB_TT_EXCEPTION:
exc = mrb_obj_value(mrb->exc);
break;
default:
mrb_assert(!"bad mrb_type");
exc = mrb_nil_value();
break;
}
mrb->exc = NULL;
}
regs[a] = exc;
NEXT;
}
CASE(OP_RESCUE, BB) {
mrb_value exc = regs[a]; /* exc on stack */
mrb_value e = regs[b];
struct RClass *ec;
switch (mrb_type(e)) {
case MRB_TT_CLASS:
case MRB_TT_MODULE:
break;
default:
RAISE_LIT(mrb, E_TYPE_ERROR, "class or module required for rescue clause");
}
ec = mrb_class_ptr(e);
regs[b] = mrb_bool_value(mrb_obj_is_kind_of(mrb, exc, ec));
NEXT;
}
CASE(OP_RAISEIF, B) {
mrb_value exc;
exc = regs[a];
if (mrb_likely(mrb_nil_p(exc))) {
mrb->exc = NULL;
}
else if (mrb_break_p(exc)) {
struct RBreak *brk;
mrb->exc = mrb_obj_ptr(exc);
L_BREAK:
brk = (struct RBreak*)mrb->exc;
switch (mrb_break_tag_get(brk)) {
#define DISPATCH_CHECKPOINTS(n, i) case n: goto CHECKPOINT_LABEL_MAKE(n);
RBREAK_TAG_FOREACH(DISPATCH_CHECKPOINTS)
#undef DISPATCH_CHECKPOINTS
default:
mrb_assert(!"wrong break tag");
}
}
else {
mrb_exc_set(mrb, exc);
L_RAISE:
ci = mrb->c->ci;
while (!ci->proc || MRB_PROC_CFUNC_P(ci->proc) || !(irep = ci->proc->body.irep) || irep->clen < 1 ||
(ch = catch_handler_find(irep, ci->pc, MRB_CATCH_FILTER_ALL)) == NULL) {
if (ci != mrb->c->cibase) {
ci = cipop(mrb);
if (ci[1].cci == CINFO_SKIP) {
mrb_assert(prev_jmp != NULL);
mrb->jmp = prev_jmp;
MRB_THROW(prev_jmp);
}
}
else if (mrb->c == mrb->root_c) {
ci->stack = mrb->c->stbase;
mrb->jmp = prev_jmp;
return mrb_obj_value(mrb->exc);
}
else {
struct mrb_context *c = mrb->c;
fiber_terminate(mrb, c, ci);
if (mrb_unlikely(!c->vmexec)) goto L_RAISE;
mrb->jmp = prev_jmp;
if (!prev_jmp) return mrb_obj_value(mrb->exc);
MRB_THROW(prev_jmp);
}
}
if (FALSE) {
L_CATCH_TAGGED_BREAK: /* from THROW_TAGGED_BREAK() or UNWIND_ENSURE() */
ci = mrb->c->ci;
}
irep = ci->proc->body.irep;
stack_extend(mrb, irep->nregs);
ci->pc = irep->iseq + mrb_irep_catch_handler_unpack(ch->target);
}
NEXT;
}
CASE(OP_MATCHERR, B) {
if (mrb_unlikely(!mrb_test(regs[a]))) {
RAISE_LIT(mrb, mrb_exc_get_id(mrb, MRB_ERROR_SYM(NoMatchingPatternError)), "pattern not matched");
}
NEXT;
}
CASE(OP_SSEND, BBB) {
regs[a] = regs[0];
}
goto L_SENDB;
CASE(OP_SSEND0, BB) {
regs[a] = regs[0];
c = 0;
}
goto L_SENDB;
CASE(OP_SSENDB, BBB) {
regs[a] = regs[0];
}
goto L_SENDB;
CASE(OP_SEND, BBB)
goto L_SENDB;
CASE(OP_SEND0, BB) {
c = 0;
}
goto L_SENDB;
L_SEND_SYM:
c = 1;
/* push nil after arguments */
SET_NIL_VALUE(regs[a+2]);
goto L_SENDB_SYM;
CASE(OP_SENDB, BBB)
L_SENDB:
mid = irep->syms[b];
L_SENDB_SYM:
{
mrb_method_t m;
mrb_value recv, blk;
mrb_int bidx, new_bidx;
if (mrb_likely(c < CALL_MAXARGS)) {
/* fast path limited to fixed length arguments of less than 15 */
bidx = a + c + 1 /* self */;
new_bidx = bidx;
}
else {
int n = c&0xf;
int nk = (c>>4)&0xf;
bidx = a + mrb_bidx(n,nk);
new_bidx = bidx;
if (nk == CALL_MAXARGS) {
mrb_ensure_hash_type(mrb, regs[a+(n==CALL_MAXARGS?1:n)+1]);
}
else if (nk > 0) { /* pack keyword arguments */
mrb_int kidx = a+(n==CALL_MAXARGS?1:n)+1;
mrb_value kdict = hash_new_from_regs(mrb, nk, kidx);
ci = mrb->c->ci;
regs[kidx] = kdict;
nk = CALL_MAXARGS;
c = n | (nk<<4);
new_bidx = a+mrb_bidx(n, nk);
}
}
mrb_assert(bidx < irep->nregs);
if (insn == OP_SEND || insn == OP_SEND0 || insn == OP_SSEND || insn == OP_SSEND0) {
/* clear block argument */
SET_NIL_VALUE(regs[new_bidx]);
SET_NIL_VALUE(blk);
}
else {
blk = ensure_block(mrb, regs[bidx]);
ci = mrb->c->ci;
regs[new_bidx] = blk;
}
ci = cipush(mrb, a, CINFO_DIRECT, NULL, NULL, BLK_PTR(blk), 0, c);
recv = regs[0];
ci->u.target_class = (insn == OP_SUPER) ? CI_TARGET_CLASS(ci - 1)->super : mrb_class(mrb, recv);
m = mrb_vm_find_method(mrb, ci->u.target_class, &ci->u.target_class, mid);
if (mrb_unlikely(MRB_METHOD_UNDEF_P(m))) {
m = prepare_missing(mrb, ci, recv, mid, blk, (insn == OP_SUPER));
}
else {
ci->mid = mid;
}
if (insn == OP_SEND || insn == OP_SEND0 || insn == OP_SENDB) {
mrb_bool priv = TRUE;
if (m.flags & MRB_METHOD_PRIVATE_FL) {
vis_err:;
mrb_value args = (ci->n == 15) ? regs[1] : mrb_ary_new_from_values(mrb, ci->n, regs+1);
vis_error(mrb, mid, args, recv, priv);
}
else if ((m.flags & MRB_METHOD_PROTECTED_FL) && mrb_obj_is_kind_of(mrb, recv, ci->u.target_class)) {
priv = FALSE;
goto vis_err;
}
}
ci->cci = CINFO_NONE;
if (MRB_METHOD_PROC_P(m)) {
const struct RProc *p = MRB_METHOD_PROC(m);
/* handle alias */
MRB_PROC_RESOLVE_ALIAS(ci, p);
CI_PROC_SET(ci, p);
if (!MRB_PROC_CFUNC_P(p)) {
/* setup environment for calling method */
irep = p->body.irep;
stack_extend(mrb, (irep->nregs < 4) ? 4 : irep->nregs);
ci->pc = irep->iseq;
JUMP;
}
else {
if (MRB_PROC_NOARG_P(p) && (ci->n > 0 || ci->nk > 0)) {
check_argument_count(mrb, ci, 0);
}
recv = MRB_PROC_CFUNC(p)(mrb, recv);
}
}
else {
check_argument_count(mrb, ci, MRB_MT_ASPEC(m.flags));
recv = MRB_METHOD_FUNC(m)(mrb, recv);
}
/* cfunc epilogue */
mrb_gc_arena_shrink(mrb, ai);
if (mrb_unlikely(mrb->exc)) goto L_RAISE;
ci = mrb->c->ci;
if (!ci->u.keep_context) { /* return from context modifying method (resume/yield) */
if (ci->cci == CINFO_RESUMED) {
mrb->jmp = prev_jmp;
return recv;
}
else {
mrb_assert(!MRB_PROC_CFUNC_P(ci[-1].proc));
irep = ci[-1].proc->body.irep;
}
}
mrb_assert(ci > mrb->c->cibase);
ci->stack[0] = recv;
/* pop stackpos */
ci = cipop(mrb);
JUMP;
}
CASE(OP_CALL, Z) {
const struct RProc *p = mrb_proc_ptr(ci->stack[0]);
int r = vm_call_proc(mrb, p, ci_bidx(ci)+1, &irep, ai);
ci = mrb->c->ci;
if (r == VM_RAISE) goto L_RAISE;
if (r == VM_RETURN_NIL) { a = 0; goto L_OP_RETURN_BODY; }
JUMP;
}
CASE(OP_BLKCALL, BB) {
/* Direct block call: R[a] = R[a].call(R[a+1],...,R[a+b]) */
if (mrb_unlikely(!mrb_proc_p(regs[a]))) {
mrb_raisef(mrb, E_TYPE_ERROR, "wrong type %T (expected Proc)", regs[a]);
}
const struct RProc *p = mrb_proc_ptr(regs[a]);
ci = cipush(mrb, a, CINFO_DIRECT, NULL, NULL, NULL, 0, b);
ci->cci = CINFO_NONE; /* mark as VM-to-VM call for proper break handling */
int r = vm_call_proc(mrb, p, b+1, &irep, ai);
ci = mrb->c->ci;
if (r == VM_RAISE) goto L_RAISE;
if (r == VM_RETURN_NIL) { a = 0; goto L_OP_RETURN_BODY; }
JUMP;
}
CASE(OP_SUPER, BB) {
mrb_value recv;
struct RClass* target_class = CI_TARGET_CLASS(ci);
mid = ci->mid;
if (mid == 0 || !target_class) {
RAISE_LIT(mrb, E_NOMETHOD_ERROR, "super called outside of method");
}
if ((target_class->flags & MRB_FL_CLASS_IS_PREPENDED) || target_class->tt == MRB_TT_MODULE) {
goto super_typeerror;
}
recv = regs[0];
if (!mrb_obj_is_kind_of(mrb, recv, target_class)) {
super_typeerror:
RAISE_LIT(mrb, E_TYPE_ERROR, "self has wrong type to call super in this context");
}
c = b; // arg info
regs[a] = recv;
goto L_SENDB_SYM;
}
CASE(OP_ARGARY, BS) {
if (vm_op_argary(mrb, a, b) == VM_RAISE) goto L_RAISE;
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_ENTER, W) {
if (vm_op_enter(mrb, a) == VM_RAISE) goto L_RAISE;
ci = mrb->c->ci;
irep = ci->proc->body.irep;
JUMP;
}
CASE(OP_KARG, BB) {
mrb_value k = mrb_symbol_value(irep->syms[b]);
mrb_int kidx = mrb_ci_kidx(ci);
mrb_value kdict, v;
if (kidx < 0 || !mrb_hash_p(kdict=regs[kidx]) || !mrb_hash_key_p(mrb, kdict, k)) {
RAISE_FORMAT(mrb, E_ARGUMENT_ERROR, "missing keyword: %v", k);
}
v = mrb_hash_delete_key(mrb, kdict, k);
ci = mrb->c->ci;
regs[a] = v;
NEXT;
}
CASE(OP_KEY_P, BB) {
mrb_value k = mrb_symbol_value(irep->syms[b]);
mrb_int kidx = mrb_ci_kidx(ci);
mrb_value kdict;
mrb_bool key_p = FALSE;
if (kidx >= 0 && mrb_hash_p(kdict=regs[kidx])) {
key_p = mrb_hash_key_p(mrb, kdict, k);
ci = mrb->c->ci;
}
regs[a] = mrb_bool_value(key_p);
NEXT;
}
CASE(OP_KEYEND, Z) {
mrb_int kidx = mrb_ci_kidx(ci);
mrb_value kdict;
if (kidx >= 0 && mrb_hash_p(kdict=regs[kidx]) && !mrb_hash_empty_p(mrb, kdict)) {
mrb_value key1 = mrb_hash_first_key(mrb, kdict);
RAISE_FORMAT(mrb, E_ARGUMENT_ERROR, "unknown keyword: %v", key1);
}
NEXT;
}
CASE(OP_BREAK, B) {
if (MRB_PROC_STRICT_P(ci->proc)) goto NORMAL_RETURN;
if (!MRB_PROC_ORPHAN_P(ci->proc) && MRB_PROC_ENV_P(ci->proc) && ci->proc->e.env->cxt == mrb->c) {
const struct RProc *dst = ci->proc->upper;
for (ptrdiff_t i = ci - mrb->c->cibase; i > 0; i--, ci--) {
if (ci[-1].proc == dst) {
goto L_UNWINDING;
}
}
}
RAISE_LIT(mrb, E_LOCALJUMP_ERROR, "break from proc-closure");
/* not reached */
}
CASE(OP_RETURN_BLK, B) {
if (!MRB_PROC_ENV_P(ci->proc) || MRB_PROC_STRICT_P(ci->proc)) {
goto NORMAL_RETURN;
}
const struct REnv *env = ci->u.env;
const struct RProc *dst = top_proc(mrb, ci->proc, &env);
if (!MRB_PROC_ENV_P(dst) || dst->e.env->cxt == mrb->c) {
/* check jump destination */
for (ptrdiff_t i = ci - mrb->c->cibase; i >= 0; i--, ci--) {
if (ci->u.env == env) {
goto L_UNWINDING;
}
}
}
/* no jump destination */
RAISE_LIT(mrb, E_LOCALJUMP_ERROR, "unexpected return");
/* not reached */
}
CASE(OP_RETSELF, Z) {
a = 0;
goto NORMAL_RETURN;
}
CASE(OP_RETNIL, Z) {
a = 0;
goto L_RETURN_NIL;
}
CASE(OP_RETTRUE, Z) {
a = 0;
goto L_RETURN_TRUE;
}
CASE(OP_RETFALSE, Z) {
a = 0;
goto L_RETURN_FALSE;
}
CASE(OP_RETURN, B) {
mrb_int acc;
mrb_value v;
mrb_callinfo *return_ci;
NORMAL_RETURN:
v = regs[a];
goto L_RETURN;
L_RETURN_NIL:
v = mrb_nil_value();
goto L_RETURN;
L_RETURN_TRUE:
v = mrb_true_value();
goto L_RETURN;
L_RETURN_FALSE:
v = mrb_false_value();
L_RETURN:
mrb_gc_protect(mrb, v);
return_ci = ci;
CHECKPOINT_RESTORE(RBREAK_TAG_BREAK) {
if (TRUE) {
struct RBreak *brk = (struct RBreak*)mrb->exc;
return_ci = &mrb->c->cibase[brk->ci_break_index];
v = mrb_break_value_get(brk);
}
else {
L_UNWINDING:
return_ci = ci;
ci = mrb->c->ci;
v = ci->stack[a];
}
mrb_gc_protect(mrb, v);
}
CHECKPOINT_MAIN(RBREAK_TAG_BREAK) {
for (;;) {
UNWIND_ENSURE(mrb, ci, ci->pc, RBREAK_TAG_BREAK, return_ci, v);
if (ci == return_ci) {
break;
}
ci = cipop(mrb);
if (ci[1].cci != CINFO_NONE) {
mrb_assert(prev_jmp != NULL);
mrb->exc = (struct RObject*)break_new(mrb, RBREAK_TAG_BREAK, return_ci, v);
mrb_gc_arena_restore(mrb, ai);
mrb->c->vmexec = FALSE;
mrb->jmp = prev_jmp;
MRB_THROW(prev_jmp);
}
}
}
CHECKPOINT_END(RBREAK_TAG_BREAK);
mrb->exc = NULL; /* clear break object */
if (ci == mrb->c->cibase) {
struct mrb_context *c = mrb->c;
if (c == mrb->root_c) {
/* toplevel return */
mrb_gc_arena_restore(mrb, ai);
mrb->jmp = prev_jmp;
return v;
}
#ifdef MRB_USE_TASK_SCHEDULER
if (mrb->c->status == MRB_TASK_CREATED) {
mrb_gc_arena_restore(mrb, ai);
mrb->jmp = prev_jmp;
TASK_STOP(mrb);
return v;
}
#endif
fiber_terminate(mrb, c, ci);
if (c->vmexec ||
(mrb->c == mrb->root_c && mrb->c->ci == mrb->c->cibase) /* case using Fiber#transfer in mrb_fiber_resume() */) {
mrb_gc_arena_restore(mrb, ai);
c->vmexec = FALSE;
mrb->jmp = prev_jmp;
return v;
}
ci = mrb->c->ci;
}
if (mrb->c->vmexec && !ci->u.keep_context) {
mrb_gc_arena_restore(mrb, ai);
mrb->c->vmexec = FALSE;
mrb->jmp = prev_jmp;
return v;
}
acc = ci->cci;
ci = cipop(mrb);
if (acc == CINFO_SKIP || acc == CINFO_DIRECT) {
mrb_gc_arena_restore(mrb, ai);
mrb->jmp = prev_jmp;
return v;
}
DEBUG(fprintf(stderr, "from :%s\n", mrb_sym_name(mrb, ci->mid)));
irep = ci->proc->body.irep;
ci[1].stack[0] = v;
mrb_gc_arena_restore(mrb, ai);
JUMP;
}
CASE(OP_BLKPUSH, BS) {
if (vm_op_blkpush(mrb, a, b) == VM_RAISE) goto L_RAISE;
NEXT;
}
#if !defined(MRB_USE_BIGINT) || defined(MRB_INT32)
L_INT_OVERFLOW:
RAISE_LIT(mrb, E_RANGE_ERROR, "integer overflow");
#endif
#define OP_MATH(op_name) do { \
/* need to check if op is overridden */ \
uint16_t tt = TYPES2(mrb_type(regs[a]),mrb_type(regs[a+1])); \
if (mrb_likely(tt == TYPES2(MRB_TT_INTEGER, MRB_TT_INTEGER))) { \
mrb_int x = mrb_integer(regs[a]), y = mrb_integer(regs[a+1]), z; \
if (mrb_int_##op_name##_overflow(x, y, &z)) { \
OP_MATH_OVERFLOW_INT(op_name,x,y); \
} \
else \
SET_INT_VALUE(mrb,regs[a], z); \
} \
else switch (tt) { \
OP_MATH_CASE_FLOAT(op_name, integer, float); \
OP_MATH_CASE_FLOAT(op_name, float, integer); \
OP_MATH_CASE_FLOAT(op_name, float, float); \
OP_MATH_CASE_STRING_##op_name(); \
default: \
mid = MRB_OPSYM(op_name); \
goto L_SEND_SYM; \
} \
} while(0); \
NEXT;
#define OP_MATH_CASE_INTEGER(op_name) \
case TYPES2(MRB_TT_INTEGER, MRB_TT_INTEGER): \
{ \
mrb_int x = mrb_integer(regs[a]), y = mrb_integer(regs[a+1]), z; \
if (mrb_int_##op_name##_overflow(x, y, &z)) { \
OP_MATH_OVERFLOW_INT(op_name,x,y); \
} \
else \
SET_INT_VALUE(mrb,regs[a], z); \
} \
break
#ifdef MRB_NO_FLOAT
#define OP_MATH_CASE_FLOAT(op_name, t1, t2) (void)0
#else
#define OP_MATH_CASE_FLOAT(op_name, t1, t2) \
case TYPES2(OP_MATH_TT_##t1, OP_MATH_TT_##t2): \
{ \
mrb_float z = mrb_##t1(regs[a]) OP_MATH_OP_##op_name mrb_##t2(regs[a+1]); \
SET_FLOAT_VALUE(mrb, regs[a], z); \
} \
break
#endif
#ifdef MRB_USE_BIGINT
#define OP_MATH_OVERFLOW_INT(op,x,y) regs[a] = mrb_bint_##op##_ii(mrb,x,y)
#else
#define OP_MATH_OVERFLOW_INT(op,x,y) goto L_INT_OVERFLOW
#endif
#define OP_MATH_CASE_STRING_add() \
case TYPES2(MRB_TT_STRING, MRB_TT_STRING): \
regs[a] = mrb_str_plus(mrb, regs[a], regs[a+1]); \
mrb_gc_arena_restore(mrb, ai); \
break
#define OP_MATH_CASE_STRING_sub() (void)0
#define OP_MATH_CASE_STRING_mul() (void)0
#define OP_MATH_OP_add +
#define OP_MATH_OP_sub -
#define OP_MATH_OP_mul *
#define OP_MATH_TT_integer MRB_TT_INTEGER
#define OP_MATH_TT_float MRB_TT_FLOAT
CASE(OP_ADD, B) {
OP_MATH(add);
}
CASE(OP_SUB, B) {
OP_MATH(sub);
}
CASE(OP_MUL, B) {
OP_MATH(mul);
}
CASE(OP_DIV, B) {
int r = vm_op_div(mrb, a, &mid);
ci = mrb->c->ci;
if (r == VM_SEND_SYM) goto L_SEND_SYM;
NEXT;
}
#define OP_MATHI(op_name) do { \
/* need to check if op is overridden */ \
if (mrb_likely(mrb_integer_p(regs[a]))) { \
mrb_int x = mrb_integer(regs[a]), y = (mrb_int)b, z; \
if (mrb_int_##op_name##_overflow(x, y, &z)) { \
OP_MATH_OVERFLOW_INT(op_name,x,y); \
} \
else \
SET_INT_VALUE(mrb,regs[a], z); \
} \
else switch (mrb_type(regs[a])) { \
OP_MATHI_CASE_FLOAT(op_name); \
default: \
SET_INT_VALUE(mrb,regs[a+1], b); \
mid = MRB_OPSYM(op_name); \
goto L_SEND_SYM; \
} \
} while(0); \
NEXT;
#define OP_MATHI_CASE_INTEGER(op_name) \
case MRB_TT_INTEGER: \
{ \
mrb_int x = mrb_integer(regs[a]), y = (mrb_int)b, z; \
if (mrb_int_##op_name##_overflow(x, y, &z)) { \
OP_MATH_OVERFLOW_INT(op_name,x,y); \
} \
else \
SET_INT_VALUE(mrb,regs[a], z); \
} \
break
#ifdef MRB_NO_FLOAT
#define OP_MATHI_CASE_FLOAT(op_name) (void)0
#else
#define OP_MATHI_CASE_FLOAT(op_name) \
case MRB_TT_FLOAT: \
{ \
mrb_float z = mrb_float(regs[a]) OP_MATH_OP_##op_name b; \
SET_FLOAT_VALUE(mrb, regs[a], z); \
} \
break
#endif
CASE(OP_ADDI, BB) {
OP_MATHI(add);
}
CASE(OP_SUBI, BB) {
OP_MATHI(sub);
}
#ifdef MRB_NO_FLOAT
#define OP_MATHILV_CASE_FLOAT(op_name) (void)0
#else
#define OP_MATHILV_CASE_FLOAT(op_name) \
case MRB_TT_FLOAT: \
{ \
mrb_float z = mrb_float(regs[a]) OP_MATH_OP_##op_name c; \
SET_FLOAT_VALUE(mrb, regs[a], z); \
} \
break
#endif
#define OP_MATHILV(op_name) \
/* a=local, b=working space, c=immediate */ \
switch (mrb_type(regs[a])) { \
case MRB_TT_INTEGER: \
{ \
mrb_int x = mrb_integer(regs[a]), y = (mrb_int)c, z; \
if (mrb_int_##op_name##_overflow(x, y, &z)) { \
OP_MATH_OVERFLOW_INT(op_name,x,y); \
} \
else { \
SET_INT_VALUE(mrb,regs[a], z); \
} \
} \
break; \
OP_MATHILV_CASE_FLOAT(op_name); \
default: \
SET_INT_VALUE(mrb,regs[a+1], c); \
mid = MRB_OPSYM(op_name); \
goto L_SEND_SYM; \
} \
NEXT
CASE(OP_ADDILV, BBB) {
OP_MATHILV(add);
}
CASE(OP_SUBILV, BBB) {
OP_MATHILV(sub);
}
#define OP_CMP_BODY(op,v1,v2) (v1(regs[a]) op v2(regs[a+1]))
#ifdef MRB_NO_FLOAT
#define OP_CMP(op,sym) do {\
int result;\
/* need to check if op is overridden */\
if (mrb_likely(TYPES2(mrb_type(regs[a]),mrb_type(regs[a+1])) == \
TYPES2(MRB_TT_INTEGER,MRB_TT_INTEGER))) {\
result = OP_CMP_BODY(op,mrb_fixnum,mrb_fixnum);\
}\
else {\
mid = MRB_OPSYM(sym);\
goto L_SEND_SYM;\
}\
if (result) {\
SET_TRUE_VALUE(regs[a]);\
}\
else {\
SET_FALSE_VALUE(regs[a]);\
}\
} while(0)
#else
#define OP_CMP(op, sym) do {\
int result;\
/* need to check if op is overridden */\
uint16_t tt = TYPES2(mrb_type(regs[a]),mrb_type(regs[a+1]));\
if (mrb_likely(tt == TYPES2(MRB_TT_INTEGER,MRB_TT_INTEGER))) {\
result = OP_CMP_BODY(op,mrb_integer,mrb_integer);\
}\
else switch (tt) {\
case TYPES2(MRB_TT_INTEGER,MRB_TT_FLOAT):\
result = OP_CMP_BODY(op,mrb_integer,mrb_float);\
break;\
case TYPES2(MRB_TT_FLOAT,MRB_TT_INTEGER):\
result = OP_CMP_BODY(op,mrb_float,mrb_integer);\
break;\
case TYPES2(MRB_TT_FLOAT,MRB_TT_FLOAT):\
result = OP_CMP_BODY(op,mrb_float,mrb_float);\
break;\
default:\
mid = MRB_OPSYM(sym);\
goto L_SEND_SYM;\
}\
if (result) {\
SET_TRUE_VALUE(regs[a]);\
}\
else {\
SET_FALSE_VALUE(regs[a]);\
}\
} while(0)
#endif
CASE(OP_EQ, B) {
if (mrb_obj_eq(mrb, regs[a], regs[a+1])) {
SET_TRUE_VALUE(regs[a]);
}
else if (mrb_symbol_p(regs[a])) {
SET_FALSE_VALUE(regs[a]);
}
else {
OP_CMP(==,eq);
}
NEXT;
}
CASE(OP_LT, B) {
OP_CMP(<,lt);
NEXT;
}
CASE(OP_LE, B) {
OP_CMP(<=,le);
NEXT;
}
CASE(OP_GT, B) {
OP_CMP(>,gt);
NEXT;
}
CASE(OP_GE, B) {
OP_CMP(>=,ge);
NEXT;
}
CASE(OP_ARRAY, BB) {
regs[a] = ary_new_from_regs(mrb, b, a);
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_ARRAY2, BBB) {
regs[a] = ary_new_from_regs(mrb, c, b);
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_ARYCAT, B) {
mrb_value splat = mrb_ary_splat(mrb, regs[a+1]);
ci = mrb->c->ci;
if (mrb_nil_p(regs[a])) {
regs[a] = splat;
}
else {
mrb_ensure_array_type(mrb, regs[a]);
mrb_ary_concat(mrb, regs[a], splat);
}
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_ARYPUSH, BB) {
mrb_ensure_array_type(mrb, regs[a]);
for (mrb_int i=0; i<b; i++) {
mrb_ary_push(mrb, regs[a], regs[a+i+1]);
}
NEXT;
}
CASE(OP_ARYSPLAT, B) {
mrb_value ary = mrb_ary_splat(mrb, regs[a]);
ci = mrb->c->ci;
regs[a] = ary;
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_AREF, BBB) {
mrb_value v = regs[b];
if (!mrb_array_p(v)) {
if (c == 0) {
regs[a] = v;
}
else {
SET_NIL_VALUE(regs[a]);
}
}
else {
v = mrb_ary_ref(mrb, v, c);
regs[a] = v;
}
NEXT;
}
CASE(OP_ASET, BBB) {
mrb_ensure_array_type(mrb, regs[b]);
mrb_ary_set(mrb, regs[b], c, regs[a]);
NEXT;
}
CASE(OP_APOST, BBB) {
mrb_value v = regs[a];
int pre = b;
int post = c;
if (!mrb_array_p(v)) {
v = ary_new_from_regs(mrb, 1, a);
}
struct RArray *ary = mrb_ary_ptr(v);
int len = (int)ARY_LEN(ary);
if (len > pre + post) {
v = mrb_ary_new_from_values(mrb, len - pre - post, ARY_PTR(ary)+pre);
regs[a++] = v;
while (post--) {
regs[a++] = ARY_PTR(ary)[len-post-1];
}
}
else {
v = mrb_ary_new_capa(mrb, 0);
regs[a++] = v;
int idx;
for (idx=0; idx+pre<len; idx++) {
regs[a+idx] = ARY_PTR(ary)[pre+idx];
}
while (idx < post) {
SET_NIL_VALUE(regs[a+idx]);
idx++;
}
}
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_INTERN, B) {
mrb_ensure_string_type(mrb, regs[a]);
mrb_sym sym = mrb_intern_str(mrb, regs[a]);
regs[a] = mrb_symbol_value(sym);
NEXT;
}
CASE(OP_SYMBOL, BB) {
size_t len;
mrb_sym sym;
mrb_assert((irep->pool[b].tt&IREP_TT_NFLAG)==0);
len = irep->pool[b].tt >> 2;
if (irep->pool[b].tt & IREP_TT_SFLAG) {
sym = mrb_intern_static(mrb, irep->pool[b].u.str, len);
}
else {
sym = mrb_intern(mrb, irep->pool[b].u.str, len);
}
regs[a] = mrb_symbol_value(sym);
NEXT;
}
CASE(OP_STRING, BB) {
mrb_int len;
mrb_assert((irep->pool[b].tt&IREP_TT_NFLAG)==0);
len = irep->pool[b].tt >> 2;
if (irep->pool[b].tt & IREP_TT_SFLAG) {
regs[a] = mrb_str_new_static(mrb, irep->pool[b].u.str, len);
}
else {
regs[a] = mrb_str_new(mrb, irep->pool[b].u.str, len);
}
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_STRCAT, B) {
mrb_ensure_string_type(mrb, regs[a]);
mrb_str_concat(mrb, regs[a], regs[a+1]);
ci = mrb->c->ci;
NEXT;
}
CASE(OP_HASH, BB) {
mrb_value hash = mrb_hash_new_capa(mrb, b);
int lim = a+b*2;
for (int i=a; i<lim; i+=2) {
mrb_hash_set(mrb, hash, regs[i], regs[i+1]);
ci = mrb->c->ci;
}
regs[a] = hash;
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_HASHADD, BB) {
mrb_value hash;
int lim = a+b*2+1;
hash = regs[a];
mrb_ensure_hash_type(mrb, hash);
for (int i=a+1; i<lim; i+=2) {
mrb_hash_set(mrb, hash, regs[i], regs[i+1]);
ci = mrb->c->ci;
}
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_HASHCAT, B) {
mrb_value hash = regs[a];
mrb_ensure_hash_type(mrb, hash);
mrb_hash_merge(mrb, hash, regs[a+1]);
ci = mrb->c->ci;
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_LAMBDA, BB)
c = OP_L_LAMBDA;
L_MAKE_LAMBDA:
{
struct RProc *p;
const mrb_irep *nirep = irep->reps[b];
if (c & OP_L_CAPTURE) {
p = mrb_closure_new(mrb, nirep);
}
else {
p = mrb_proc_new(mrb, nirep);
p->flags |= MRB_PROC_SCOPE;
}
if (c & OP_L_STRICT) p->flags |= MRB_PROC_STRICT;
regs[a] = mrb_obj_value(p);
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_BLOCK, BB) {
c = OP_L_BLOCK;
goto L_MAKE_LAMBDA;
}
CASE(OP_METHOD, BB) {
c = OP_L_METHOD;
goto L_MAKE_LAMBDA;
}
CASE(OP_RANGE_INC, B) {
mrb_value v = mrb_range_new(mrb, regs[a], regs[a+1], FALSE);
ci = mrb->c->ci;
regs[a] = v;
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_RANGE_EXC, B) {
mrb_value v = mrb_range_new(mrb, regs[a], regs[a+1], TRUE);
ci = mrb->c->ci;
regs[a] = v;
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_OCLASS, B) {
regs[a] = mrb_obj_value(mrb->object_class);
NEXT;
}
CASE(OP_CLASS, BB) {
struct RClass *c = 0, *baseclass;
mrb_sym id = irep->syms[b];
mrb_value base = regs[a];
mrb_value super = regs[a+1];
if (mrb_nil_p(base)) {
baseclass = MRB_PROC_TARGET_CLASS(ci->proc);
if (!baseclass) baseclass = mrb->object_class;
base = mrb_obj_value(baseclass);
}
c = mrb_vm_define_class(mrb, base, super, id);
ci = mrb->c->ci;
regs[a] = mrb_obj_value(c);
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_MODULE, BB) {
struct RClass *cls = 0, *baseclass;
mrb_sym id = irep->syms[b];
mrb_value base = regs[a];
if (mrb_nil_p(base)) {
baseclass = MRB_PROC_TARGET_CLASS(ci->proc);
if (!baseclass) baseclass = mrb->object_class;
base = mrb_obj_value(baseclass);
}
cls = mrb_vm_define_module(mrb, base, id);
ci = mrb->c->ci;
regs[a] = mrb_obj_value(cls);
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_EXEC, BB)
{
mrb_value recv = regs[a];
struct RClass *c = mrb_class_ptr(recv);
const mrb_irep *nirep = irep->reps[b];
/* prepare closure */
struct RProc *p = mrb_proc_new(mrb, nirep);
p->c = NULL;
mrb_field_write_barrier(mrb, (struct RBasic*)p, (struct RBasic*)ci->proc);
MRB_PROC_SET_TARGET_CLASS(p, c);
p->flags |= MRB_PROC_SCOPE;
/* prepare call stack */
ci = cipush(mrb, a, 0, c, p, NULL, 0, 0);
irep = p->body.irep;
stack_extend(mrb, irep->nregs);
stack_clear(regs+1, irep->nregs-1);
ci->pc = irep->iseq;
JUMP;
}
CASE(OP_DEF, BB) {
struct RClass *target = mrb_class_ptr(regs[a]);
const struct RProc *p = mrb_proc_ptr(regs[a+1]);
mrb_method_t m;
mrb_sym mid = irep->syms[b];
MRB_METHOD_FROM_PROC(m, p);
MRB_METHOD_SET_VISIBILITY(m, MRB_METHOD_VDEFAULT_FL);
mrb_define_method_raw(mrb, target, mid, m);
mrb_method_added(mrb, target, mid);
ci = mrb->c->ci;
mrb_gc_arena_restore(mrb, ai);
regs[a] = mrb_symbol_value(mid);
NEXT;
}
CASE(OP_TDEF, BBB) {
struct RClass *tc = check_target_class(mrb);
if (mrb_unlikely(!tc)) goto L_RAISE;
mid = vm_define_method(mrb, tc, irep, b, c);
ci = mrb->c->ci;
mrb_gc_arena_restore(mrb, ai);
regs[a] = mrb_symbol_value(mid);
NEXT;
}
CASE(OP_SDEF, BBB) {
struct RClass *tc = mrb_class_ptr(mrb_singleton_class(mrb, regs[a]));
mid = vm_define_method(mrb, tc, irep, b, c);
ci = mrb->c->ci;
mrb_gc_arena_restore(mrb, ai);
regs[a] = mrb_symbol_value(mid);
NEXT;
}
CASE(OP_SCLASS, B) {
regs[a] = mrb_singleton_class(mrb, regs[a]);
mrb_gc_arena_restore(mrb, ai);
NEXT;
}
CASE(OP_TCLASS, B) {
struct RClass *target = check_target_class(mrb);
if (mrb_unlikely(!target)) goto L_RAISE;
regs[a] = mrb_obj_value(target);
NEXT;
}
CASE(OP_ALIAS, BB) {
struct RClass *target = check_target_class(mrb);
if (mrb_unlikely(!target)) goto L_RAISE;
mrb_alias_method(mrb, target, irep->syms[a], irep->syms[b]);
mrb_method_added(mrb, target, irep->syms[a]);
ci = mrb->c->ci;
NEXT;
}
CASE(OP_UNDEF, B) {
struct RClass *target = check_target_class(mrb);
if (mrb_unlikely(!target)) goto L_RAISE;
mrb_undef_method_id(mrb, target, irep->syms[a]);
ci = mrb->c->ci;
NEXT;
}
CASE(OP_DEBUG, BBB) {
#ifdef MRB_USE_DEBUG_HOOK
if (mrb->debug_op_hook) mrb->debug_op_hook(mrb, irep, ci->pc, regs);
#else
#ifndef MRB_NO_STDIO
printf("OP_DEBUG %d %d %d\n", a, b, c);
#else
abort();
#endif
#endif
NEXT;
}
CASE(OP_ERR, B) {
size_t len = irep->pool[a].tt >> 2;
mrb_value exc;
mrb_assert((irep->pool[a].tt&IREP_TT_NFLAG)==0);
exc = mrb_exc_new(mrb, E_LOCALJUMP_ERROR, irep->pool[a].u.str, len);
RAISE_EXC(mrb, exc);
}
CASE(OP_EXT1, Z) {
const mrb_code *pc = ci->pc;
insn = READ_B();
switch (insn) {
#define OPCODE(insn,ops) case OP_ ## insn: FETCH_ ## ops ## _1(); ci->pc = pc; goto L_OP_ ## insn ## _BODY;
#include <mruby/ops.h>
#undef OPCODE
}
NEXT;
}
CASE(OP_EXT2, Z) {
const mrb_code *pc = ci->pc;
insn = READ_B();
switch (insn) {
#define OPCODE(insn,ops) case OP_ ## insn: FETCH_ ## ops ## _2(); ci->pc = pc; goto L_OP_ ## insn ## _BODY;
#include <mruby/ops.h>
#undef OPCODE
}
NEXT;
}
CASE(OP_EXT3, Z) {
const mrb_code *pc = ci->pc;
insn = READ_B();
switch (insn) {
#define OPCODE(insn,ops) case OP_ ## insn: FETCH_ ## ops ## _3(); ci->pc = pc; goto L_OP_ ## insn ## _BODY;
#include <mruby/ops.h>
#undef OPCODE
}
NEXT;
}
CASE(OP_STOP, Z) {
/* stop VM */
mrb_value v;
v = mrb->exc ? mrb_obj_value(mrb->exc) : mrb_nil_value();
CHECKPOINT_RESTORE(RBREAK_TAG_STOP) {
struct RBreak *brk = (struct RBreak*)mrb->exc;
v = mrb_break_value_get(brk);
}
CHECKPOINT_MAIN(RBREAK_TAG_STOP) {
UNWIND_ENSURE(mrb, ci, ci->pc, RBREAK_TAG_STOP, ci, v);
}
CHECKPOINT_END(RBREAK_TAG_STOP);
mrb->jmp = prev_jmp;
if (!mrb_nil_p(v)) {
mrb->exc = mrb_obj_ptr(v);
TASK_STOP(mrb);
return v;
}
mrb->exc = NULL;
TASK_STOP(mrb);
return regs[irep->nlocals];
}
}
END_DISPATCH;
#undef regs
}
MRB_CATCH(&c_jmp) {
mrb_assert(mrb->exc != NULL);
ci = mrb->c->ci;
while (ci > mrb->c->cibase && ci->cci == CINFO_DIRECT) {
ci = cipop(mrb);
}
goto RETRY_TRY_BLOCK;
}
MRB_END_EXC(&c_jmp);
}
static mrb_value
mrb_run(mrb_state *mrb, const struct RProc *proc, mrb_value self)
{
return mrb_vm_run(mrb, proc, self, ci_bidx(mrb->c->ci) + 1);
}
/**
* @brief Executes a mruby proc in the top-level environment.
*
* This function is used to execute a proc (like a script loaded from a file
* or a string) at the top level of the mruby environment. It's similar to
* `mrb_vm_run` but is specifically designed for top-level execution.
*
* It ensures that if there's an existing callinfo stack, the new execution
* is pushed on top with `CINFO_SKIP`, indicating it's a new, distinct
* execution context rather than a nested call from within the VM.
*
* @param mrb The mruby state.
* @param proc The RProc object (representing the script or code) to execute.
* @param self The `self` object for this top-level execution. Typically,
* this is the main `top_self` object in mruby.
* @param stack_keep The number of values to preserve on the stack. For
* top-level execution, this is often 0 or a small number
* to set up initial local variables if any.
* @return The result of the proc's execution.
* @see mrb_vm_run
*/
MRB_API mrb_value
mrb_top_run(mrb_state *mrb, const struct RProc *proc, mrb_value self, mrb_int stack_keep)
{
if (mrb->c->cibase && mrb->c->ci > mrb->c->cibase) {
cipush(mrb, 0, CINFO_SKIP, mrb->object_class, NULL, NULL, 0, 0);
}
return mrb_vm_run(mrb, proc, self, stack_keep);
}
#undef CASE
Reference in New Issue View Git Blame Copy Permalink