| Age | Commit message (Collapse) | Author |
|---|
|
We don't want to increase number of files that should be compiled by C++
compiler when `enable_cxx_exception` is turned on.
|
|
|
|
We have introduced following new instructions.
* `OP_LAMBDA16`
* `OP_BLOCK16`
* `OP_METHOD16`
* `OP_EXEC16`
Each instruction uses 16 bits operand for `reps` index. Since new
instructions are added, `mruby/c` VM should be updated.
Due to new instructions, dump format compatibility is lost, we have
increased `RITE_BINARY_MAJOR_VER`.
In addition, we have decreased the size of `refcnt` in `mrb_irep` from
`uint32_t` to `uint16_t`, which is reasonably big enough.
|
|
In addition, stop eager allocation of `mt` table.
|
|
Introduced `MRB_PRESYM_INIT_SYMBOLS()`
|
|
Remove unnecessary `ci0` variables; ref #5272
|
|
The `init_SYMBOLS()` function implicitly defined in `MRB_PRESYM_DEFINE_VAR_AND_INITER()` requires some familiarity when trying to find it from the caller.
By introducing `MRB_PRESYM_INIT_SYMBOLS()`, it is possible to find directly from the identifier.
|
|
|
|
The following macros will be removed:
- `ENSURE_STACK_INIT_SIZE`
- `RESCUE_STACK_INIT_SIZE`
- `MRB_ECALL_DEPTH_MAX`
|
|
Raise SystemStackError if mruby VM stack expansion fails
|
|
|
|
This was because it caused `SIGSEGV` when `mruby -v` displayed an unnamed variable.
```console
% bin/mruby -ve 'call { |(a, b)| }'
...SNIP...
irep 0x8007d0050 nregs=3 nlocals=1 pools=0 syms=1 reps=1 iseq=12
file: -e
1 000 OP_LOADSELF R1
1 002 OP_BLOCK R2 I(0:0x8007d00a0)
1 005 OP_SENDB R1 :call 0
1 009 OP_RETURN R1
1 011 OP_STOP
irep 0x8007d00a0 nregs=6 nlocals=5 pools=0 syms=0 reps=0 iseq=29
local variable names:
zsh: segmentation fault (core dumped) bin/mruby -ve 'call { |(a, b)| }'
```
|
|
|
|
Because a structure that is an element of `presym_table` has padding, split
it into individual arrays for name and length.
#### Result (64-bit CPU with full-core gembox)
| | mruby | libmruby.a |
|--------|------------|------------|
| Before | 1,087,444B | 1,476,872B |
| After | 1,079,340B | 1,469,784B |
|
|
Because it may not be expected result.
example: https://wandbox.org/permlink/F5Mp7IEJ1VY3CFLp
|
|
This reverts commit dc51d89ac22acc60b9bfeed87115863565b74085.
|
|
|
|
|
|
|
|
|
|
Instead of including `mruby/presym.h` everywhere, we provided the
fallback `mruby/presym.inc` under `include/mruby` directory, and specify
`-I<build-dir>/include` before `-I<top-dir>/include` in `presym.rake`.
So even when someone drops `-I<build-dir>/include` in compiler options,
it just compiles without failure.
|
|
https://github.com/shuujii/mruby into shuujii-avoid-including-presym.inc-in-existing-header-files
|
|
`calc_crc_16_ccitt()` consumes a lot of clock cycles in programs like
`mrbtest` which loads a lot of dumped binary. Error detection for flaky
channels should be done in the higher level.
Note: `mruby/c` should be updated to support this change.
|
|
The Hash implementation assumed that there were always empty buckets, but
sometimes there were only active or deleted buckets (no empty buckets).
Therefore, fix it so that this situation does not occur.
### Example
```ruby
# example.rb
class A
attr_reader :v
def initialize(v) @v = v end
def ==(o) @v == o.v end
def hash; @v end
def to_s; "#{self.class}[#{@v}]" end
alias eql? ==
alias inspect to_s
end
keys = (0..31).map{A.new(_1)}
h = {}
(0..16).each{h[keys[_1]] = _1}
(17..31).each do
k = keys[_1]
h[k] = _1
h.delete(k)
end
p h.keys
```
#### Before this patch:
```console
$ bin/mruby example.rb
[A[0], A[1], A[2], A[3], A[4], A[5], A[6], A[7], A[8], A[9], A[10], A[11], A[12], A[13], A[14], A[15], A[16], A[30], A[31]]
```
#### After this patch:
```console
$ bin/mruby example.rb
[A[0], A[1], A[2], A[3], A[4], A[5], A[6], A[7], A[8], A[9], A[10], A[11], A[12], A[13], A[14], A[15], A[16]]
```
|
|
|
|
Thinking `Ratinal` and `Complex` in mind.
|
|
When I `#call` the "proc" object created by the `mrb_proc_new_cfunc()` function from Ruby space, the return value did not go into the correct stack position.
This can destroy the calling variable.
This issue is now caused by #5272. sorry.
|
|
dearblue-reorganize-ci
|
|
|
|
|
|
To make inline symbols packed in 30 bits.
|
|
|
|
Addressed an issue where existing programs linking `libmruby.a` could only
be built by adding `<build-dir>/include` to compiler's include path.
|
|
Previously I used the `RData` object to avoid a memory leak in `mrb_irep` if `src/load.c` failed.
ref: https://github.com/mruby/mruby/pull/4250
commit: f1523d24042ca3416dc5b9be7b3fc220ddaed896
Considering that the `RProc` object will be created in the subsequent process, it is preferable to create the `RProc` object from the beginning.
Along with this, the inside of `read_irep()` is replaced with the processing centered on the `RProc` object.
The global function that returns the `mrb_irep` pointer is still provided for compatibility.
|
|
This enhances self-containment.
- Changed the `mrb_callinfo::pc` field to point to itself.
Previously it indicated the return destination of the previous call level.
`mrb_callinfo::pc` will now hold the address to its own `proc->body.irep->iseq`.
- Removed `mrb_callinfo::err` field.
This is because `mrb_callinfo::pc - 1` is semantically the same as the previous `err`.
- The `pc0` and `pc_save` variables in `mrb_vm_exec()` are no longer needed and have been deleted.
- It removes the argument because `cipush()` doesn't need to save the previous `pc`.
|
|
This enhances self-containment.
Previously `mrb_context::stack` had the current call level stack, but now it owns it.
The `mrb_context::stack` field, which is no longer needed, will be removed.
|
|
If there is `env`, `env->c` means `target_class`.
|
|
The first byte of UTF-8 character should not be `80..c1`.
|
|
|
|
|
|
Non-ASCII characters are allowed to local variable names, so they are not
always named presym.
|
|
https://github.com/shuujii/mruby into shuujii-improve-source-scanning-for-presym
|
|
But you still cannot cross C function boundary.
|
|
Avoid 64-bit operations in `src/hash.c`; close #5201
|
|
Maybe it's a typo.
|
|
The idea of using `size_t` in `ea_next_capa_for` is by @dearblue.
|
|
|
|
|
|
Negative integer `>-65535` had wrong value, e,g, `p(-40550)` printed
`4294926746` since Nov. 2020, sigh.
|
|
|
