73c8330cbd
RBreak remembers the CI location
...
It is now possible to specify return destination directly.
This allows callinfo to distinguish between calls to the same proc object.
At the same time, the `Kernel#catch` method is adjusted.
By removing the previously required double lambda object, the REnv object is no longer created as well.
2023-12-07 21:33:14 +09:00
ca9e8d91e4
Needs cipop() before crossing the C boundary
...
fix #6108
2023-12-05 22:03:38 +09:00
87453fb70b
vm.c (CI_PROC_SET): consider alias proc
2023-12-03 22:38:31 +09:00
e8da16cdf5
vm.c (exec_irep): need to support aliases
2023-12-03 22:38:30 +09:00
ccdd81eb69
Merge pull request #6101 from dearblue/unwinding
...
Reorganization of unwinding at `break` / `return`.
2023-11-22 08:11:15 +09:00
f2dc76e3b5
Change alias handling to avoid using mid in Env; ref 6098
...
Since `Env` objects are shared by Blocks/Procs in the same context,
making multiple aliases in one context screws up with alias names.
New implementation uses alias bodies (Procs) to refer new names.
As a side effect, `__callee__` stop working correctly for aliases.
To fix this `__callee__` problem, we need to keep alias method names in
`callinfo`, which consumes more memory. We are wondering that is worth
the compatibility.
2023-11-21 09:37:17 +09:00
9c1c9ba8e5
vm.c: rename local variable m to p
...
`m` usually stands for a method but this one was `struct RProc*`.
2023-11-19 23:21:05 +09:00
91b4da3a2b
class.c: keep alias flag in proc object; fix #6098
2023-11-16 17:18:28 +09:00
2e881d3297
Reorganization of unwinding at break / return.
...
The unwinding process that has been done at each point in time is now done in a merged block after looking for the last ci to `break` / `return`.
2023-11-12 20:39:33 +09:00
fe6f9faba4
vm.c: reduce number of check_method_noarg() calls
2023-10-22 22:04:58 +09:00
e7828bac4d
Fixed unwinding block that could point to invalid PC
...
Switching back from a fiber with an exception could point to an invalid PC.
The investigation was inspired by #6060 .
2023-10-10 21:29:38 +09:00
83d12f8d52
vm.c: GCC does not provide __has_feature()
2023-10-03 01:05:26 +09:00
b9536ea35f
vm.c: use smaller MRB_CALL_LEVEL_MAX with address sanitizer; fix #6068
...
Since ASAN uses lower memory bound. The other ideas to address #6068
might follow, e.g. removing `args=` from Enumerator.
2023-10-02 21:15:54 +09:00
d298da8e72
vm.c (eval_under): check if irep is NULL; fix #6065
2023-09-28 11:37:07 +09:00
ec3baece4f
vm.c (catch_handler_find): check if irep is NULL; fix #6052
2023-09-19 14:22:49 +09:00
75e8a72649
vm.c (CI_PROC_SET): check whether p->body.irep is not NULL; ref #6052
2023-09-19 14:22:49 +09:00
4810706ccc
vm.c: refactor redundant if conditions
2023-09-16 14:49:58 +09:00
acf5b74daa
vm.c (ary_new_from_regs): avoid mrb_ary_push()
...
Use mrb_ary_new_from_values(), which is faster instead.
2023-09-05 23:29:23 +09:00
9914243d68
vm.c: use direct array access if possible
2023-09-05 23:29:23 +09:00
0578e518e6
src/vm.c: adjust mrb_gc_protect() for kdict argument
...
It was called repeatedly in some cases.
2023-08-25 23:12:08 +09:00
d6d434eb3f
Inline loop variables everywhere
2023-07-13 08:05:00 +09:00
93648fc954
gc.c (mrb_env_unshare): fix a write barrier bug
2023-06-02 11:00:13 +09:00
9369679e36
vm.c: reduce local variables
2023-06-01 09:14:39 +09:00
2ad3f0e34b
vm.c (mrb_yield_with_class): keep mid of the callee
...
Otherwise `super` from within blocks may call wrong methods.
2023-05-26 11:17:51 +09:00
9040baa270
vm.c (mrb_yield_with_class): avoid redundant initialization of mid
2023-05-26 11:14:58 +09:00
8f745b7728
vm.c (OP_ENTER): fix issues by unreferenced kdict and blk
...
There's short a period where kdict and blk are not pointed from the stack.
2023-05-23 14:54:36 +09:00
2f7f797473
remove extra spaces before ;
2023-05-22 12:05:35 +09:00
eea72ec84a
fix spaces in the type cast expressions (cosmetic changes)
2023-05-18 23:29:16 +09:00
5ead26be3d
vm.c (ci_bidx): inline the function if possible
2023-05-16 22:41:32 +09:00
27a14f0a9a
vm.c (OP_CMP): fixed wrong numeric comparison for integers; fix #5975
...
MRB_TT_INTEGER value may not be fixnum (immediate integer), so we should
not use mrb_fixnum() here.
2023-04-07 10:46:01 +09:00
ce86040a0f
Merge pull request #5973 from dearblue/envadjust
...
Small improvements to `envadjust()`
2023-04-06 10:48:49 +09:00
9e4b3af553
Merge pull request #5972 from dearblue/simple-mrb_top_run
...
Simplify `mrb_top_run()`
2023-04-06 10:46:56 +09:00
17abe548a7
Small improvements to envadjust()
...
- Calculate the difference between `oldbase` and `newbase` only once outside the loop.
- To make sure they are within range, the comparison is done only once instead of twice.
2023-04-05 22:29:27 +09:00
b90f59d36e
Simplify mrb_top_run()
2023-04-05 22:11:36 +09:00
b46a852acb
Remove unnecessary branches in L_RAISE block
...
The previous `while` loop ends if the catch handler pointer `ch` is not `NULL`.
Supplement to #5843 .
2023-04-05 21:52:21 +09:00
f6c9191cd0
Replace MRB_NO_DIRECT_THREADING with MRB_USE_VM_SWITCH_DISPATCH
...
The correct technique now is "Token Threading".
However, the new name was chosen because it was felt that if a different technique was added in the future, there would be no need to replace it.
2023-03-18 17:39:41 +09:00
2675477f07
Safely retrieve self from proc in mrb_yield()
...
Passing a proc generated by the following method to `mrb_yield()` or `mrb_yield_argv()` no longer causes `SIGSEGV`:
- C functions made into proc objects with `mrb_proc_new_cfunc()`.
- A proc object which is the top level of a program generated by `mrb_load_string_cxt()` etc. with `mrbc_context::no_exec` enabled.
See also: #5932
2023-03-17 21:20:48 +09:00
924d605712
vm.c: remove redundant assignment (already done in cipush)
2023-03-13 08:05:55 +09:00
90e53cad39
Merge pull request #5945 from dearblue/env-internal
...
Avoid exposure for `REnv` objects
2023-03-11 16:18:28 +09:00
f1ffedc3fa
vm.c: kdict may be converted to a normal argument; fix #5944
2023-03-10 08:52:04 +09:00
ddaf3bdc0f
Added note to CINFO_* constants
2023-03-09 22:13:52 +09:00
38d5ed44e1
Avoid exposure for REnv objects
...
The `REnv` object is difficult to deal with, and it would be ideal if the user did not have to manipulate it directly.
In some previous situations, it was necessary to call `mrb_env_unshare()`, a non-API function, after `mrb_load_string()` or similar.
With this patch, it is no longer necessary for users to use `mrb_env_unshare()` directly, as it is now handled internally simply by using the `mrb_vm_ci_env_clear()` function.
Also, `mrb_vm_ci_env_set()` is demoted from the `MRB_API` function for the same reason.
ref. commit 1ab3da6f08
2023-03-05 17:45:47 +09:00
1ab3da6f08
vm.c (ci_env_set): inline ci_env_set()
...
Along with making preparing mrb_vm_ci_env_clar() as a replacement of
mrb_vm_ci_env_set(mrb, NULL).
2023-02-23 17:16:10 +09:00
d8d1fe27f6
vm.c (stack_extend): inline mrb_stack_extend() in vm.c
2023-02-22 12:05:46 +09:00
43922a5bf9
vm.c (mrb_gc_arena_shrink): use local variable
2023-02-21 21:59:42 +09:00
aecbdf72bf
vm.c (mrb_env_unshare): remove unnecessary mrb_write_barrier() call
...
Since this function does not modify references from env, we don't need
to put write barrier here.
2023-02-21 09:45:17 +09:00
76157c5f78
add inline specifier to frequently called functions
...
- entry_deleted_p(): hash.c
- ensure_block(): vm.c
2023-02-19 19:06:37 +09:00
4384a87516
class.c, vm.c: use loop variable declaration in for statement
2023-02-13 14:16:53 +09:00
58aefd6010
class.c, vm.c: use NULL instead of 0 for pointers
2023-02-13 14:16:30 +09:00
079244c36a
src/class.c: add new function mrb_vm_find_method
...
The function skips `cp` dereference, and improve performance of method
calls slightly.
2023-01-23 00:17:32 +09:00
dearblue