Normal return of control from the `mrb_sys_fail()` function has unexpected results.
Example: double free, malloc with huge size used as error integer
The following is an example of an actual crash:
```console
% cat crash.rb
def SystemCallError._sys_fail(*a)
nil
end
File.readlink("/404")
% bin/mruby crash.rb
zsh: segmentation fault (core dumped) bin/mruby crash.rb
```
Internal functions can only be called from within the library.
Functions listed in `mruby/internal.h` can be called from:
* core (src/*.c)
* gems (mrbgems/**/*.c)
But not from the application linked with `libmruby`.
Since the possible values of the backtrace are limited to `nil`, `RData`, and `RAray`, they are now stored as object pointers.
This change saves memory by eliminating the need to use instance variables for common exceptions.
ref. #2485
There is no need to limit the type to `struct RString`.
Also, the `MRB_EXC_MESG_STRING_FLAG` flag can be eliminated by checking if `struct RException::mesg` is `NULL` or not.
ref. #2485
ref. #2485
The `SystemCallError#to_s` method also needed to be modified, but since it has no functional difference from the `Exception#to_s` method, it will be removed.
The purpose is two-fold:
1. to be able to specify a pointer directly when user data is used
When using `mrb_protect()`, it is necessary to allocate objects by `mrb_obj_cptr()` function when using user data.
Adding `mrb_protect_raw()` will make it simpler to reimplement `mrbgems/mruby-error`.
2. to correctly unwind callinfo when an exception is raised from a C function defined as a method (the main topic)
If a method call is made directly under `mrb_protect()` and a C function is called, control is returned from `mrb_protect()` if an exception occurs there.
In this case, callinfo is not restored, so it is out of sync.
Moreover, returning to mruby VM (`mrb_vm_exec()` function) in this state will indicate `ci->pc` of C function which is equal to `NULL`, and subsequent `JUMP` will cause `SIGSEGV`.
Following is an example that actually causes `SIGSEGV`:
- `crash.c`
```c
#include <mruby.h>
#include <mruby/compile.h>
#include <mruby/error.h>
static mrb_value
level1_body(mrb_state *mrb, mrb_value self)
{
return mrb_funcall(mrb, self, "level2", 0);
}
static mrb_value
level1(mrb_state *mrb, mrb_value self)
{
return mrb_protect(mrb, level1_body, self, NULL);
}
static mrb_value
level2(mrb_state *mrb, mrb_value self)
{
mrb_raise(mrb, E_RUNTIME_ERROR, "error!");
return mrb_nil_value();
}
int
main(int argc, char *argv[])
{
mrb_state *mrb = mrb_open();
mrb_define_method(mrb, mrb->object_class, "level1", level1, MRB_ARGS_NONE());
mrb_define_method(mrb, mrb->object_class, "level2", level2, MRB_ARGS_NONE());
mrb_p(mrb, mrb_load_string(mrb, "p level1"));
mrb_close(mrb);
return 0;
}
```
- compile & run
```console
% `bin/mruby-config --cc --cflags --ldflags` crash.c `bin/mruby-config --libs`
% ./a.out
zsh: segmentation fault (core dumped) ./a.out
```
After applying this patch, it will print exception object and exit normally.
The `mrb_protect()`, `mrb_ensure()` and `mrb_rescue_exceptions()` in `mrbgems/mruby-error` have been rewritten using `mrb_protect_raw()`.
- `MRB_WITHOUT_FLOAT` => `MRB_NO_FLOAT`
- `MRB_USE_FLOAT` => `MRB_USE_FLOAT32`
The former is to use `USE_XXX` naming convention. The latter is to make
sure `float` is 32bit float and not floating point number in general.
When a global jump occurs, look at the catch handler table to determine where to jump.
In that case, `pc` already shows the following instruction, but since the table shows `begin_offset ... end_offset`, the comparison is done with `begin_offset < pc && pc <= end_offset`.
If there is a corresponding handler, move `pc` to `handler.target_offset` and continue running the VM.
When a global jump across `ensure` is made by `return`, `break`, `next`, `redo` and `retry`, the extended `RBreak` object saves and restores the C-level execution position.
This extended `RBreak` can have tag information, which makes it a pseudo coroutine (the "tag" mimics CRuby).
The implementation of pseudo coroutines by `RBreak` is summarized by `CHECKPOINT_RESTORE ... CHECKPOINT_MAIN ... CHECKPOINT_END` and `throw_tagged_break` / `unwind_ensure` macros.
The restart of processing is branched by `RBREAK_TAG_FOREACH(DISPATCH_CHECKPOINTS)`.
- Not only `rescue` blocks but also `ensure` blocks are now sandwiched between `OP_EXCEPT` and `OP_RAISEIF`.
- Remove the function `ecall()`.
It is no longer necessary to re-enter the VM to perform an "ensure block".
This will resolves#1888.
- Added instruction `OP_JUW` (Jump while UnWind).
It jumps unconditionally like `OP_JMP`, but searches the catch handler table and executes the ensure block.
Since it searches the catch handler table, it is much heavier than `OP_JMP`.
Functions to add prototypes to headers:
* mrb_ary_splice()
* mrb_notimplement()
* mrb_vformat()
* mrb_cstr_to_dbl()
* mrb_cstr_to_inum()
Functions to be made `static` (`MRB_API` was not needed):
* mrb_mod_module_function()
* mrb_obj_hash()
* mrb_str_len_to_inum()
Functions to remove `MRB_API` from definitions (referenced from within `libmruby`):
* mrb_mod_cv_defined()
* mrb_mod_cv_get()
* mrb_f_send()
The old size of `struct RBreak` was 56 bytes (`MRB_NO_BOXING`)
and it's bigger than other object structures. That increase the
size of `RVALUE` thus increase the total amount of memory consumption.
Instead of preserving a backtrace in `mrb_state`, `mrb_exc_set`
keeps packed backtrace in an exception object. `#backtrace` unpacks
it to an array of strings.
Yukihiro "Matz" Matsumoto