summaryrefslogtreecommitdiffhomepage
path: root/src
AgeCommit message (Collapse)Author
2017-12-04Remove temporary objects from GC arena in `mrb_vformat()'; #3863Yukihiro "Matz" Matsumoto
2017-12-04Check if `ci->proc` is not `NULL` and `MRB_PROC_CFUNC_P()`; fix #3867Yukihiro "Matz" Matsumoto
2017-12-04Pop exception objects from the bottom of GC arena; fix #3863Yukihiro "Matz" Matsumoto
2017-12-01The stack shift width should be determined by `p->upper`; fix #3864Yukihiro "Matz" Matsumoto
And check required register number from `ci->proc` as well. The fixes for #3859 and #3862 were incomplete.
2017-11-30Avoid calling `mrb_env_unshare()` when `env` is `NULL`.Yukihiro "Matz" Matsumoto
Small performance improvement by reducing function invocations.
2017-11-30Avoid double free() of env stack; fix #3860Yukihiro "Matz" Matsumoto
Should turn on `MRB_ENV_STACK_UNSHARED` flag only after env stack reallocation. `malloc()` may fail.
2017-11-30Wrong stack adjustment in `ecall()` (f35f975 #3859); fix #3862Yukihiro "Matz" Matsumoto
2017-11-30Stop infinite recursion in `Class#to_s`; fix #3861Yukihiro "Matz" Matsumoto
2017-11-29Clear `c->prev` on fiber termination.Yukihiro "Matz" Matsumoto
2017-11-29Need to free contexts when freeing fibers.Yukihiro "Matz" Matsumoto
Memory leak fixed; ref #3711
2017-11-29Set `MRB_FIBER_TERMINATED` flag on exception termination of a fiber.Yukihiro "Matz" Matsumoto
2017-11-29Add `most recent call last` message to the backtrace.Yukihiro "Matz" Matsumoto
2017-11-29Call all ensure clauses pushed at `OP_STOP`.Yukihiro "Matz" Matsumoto
2017-11-29No need to use `ecall_adjust()`.Yukihiro "Matz" Matsumoto
2017-11-29Fixed stack address adjustment in `ecall()`; fix #3859Yukihiro "Matz" Matsumoto
2017-11-29Need to evaluate ensure clauses at fiber termination.Yukihiro "Matz" Matsumoto
2017-11-29Remove unnecessary code; ref #3711Yukihiro "Matz" Matsumoto
2017-11-22Provide shortcut comparison methods for numbers for performance.Yukihiro "Matz" Matsumoto
2017-11-22Stack adjustment should be based on `p->upper`; fix #3857Yukihiro "Matz" Matsumoto
It used to be based on `ci->proc` but the callinfo position may be wrong when `ecall()` is called during stack rewinding from `OP_RETURN`.
2017-11-22Clear ensure stack after calling in `OP_EPOP`; #3855 3856Yukihiro "Matz" Matsumoto
2017-11-22No need to pop ensure stack if `eidx` is smaller than `epos`; #3855 #3856Yukihiro "Matz" Matsumoto
2017-11-22fixup! Remove fixed argument of `ecall()`; ref #3855 #3856Yukihiro "Matz" Matsumoto
2017-11-22Remove fixed argument of `ecall()`; ref #3855 #3856Yukihiro "Matz" Matsumoto
2017-11-22Should have marked all ensure stack entries; fix #3855 fix #3856Yukihiro "Matz" Matsumoto
2017-11-20Reduce the size of `struct RBreak` to reduce memory usage.Yukihiro "Matz" Matsumoto
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.
2017-11-20Add `MRB_METHOD_TABLE_INLINE` option.Yukihiro "Matz" Matsumoto
Now the method tables (in classes/modules and caches) keeps C function pointers without wrapping in `struct RProc` objects. For the sake of portability, `mrb_method_t` is represented by the struct and union, but if the most significant bit of the pointer is not used by the platform, `mrb_method_t` should be packed in `uintptr_t` to reduce memory usage. `MRB_METHOD_TABLE_INLINE` is turned on by default for linux.
2017-11-19Invoke `mrb_full_gc()` before `ObjectSpace.count_objects`.Yukihiro "Matz" Matsumoto
2017-11-19Leave `hash->ht` to be `NULL` if initial capacity is zero.Yukihiro "Matz" Matsumoto
2017-11-18Assign operands to local variables.Yukihiro "Matz" Matsumoto
2017-11-17Removed useless conditionUkrainskiy Sergey
2017-11-17Changed the assignment of a variable in variable.cUkrainskiy Sergey
2017-11-17Changed numbers to TRUE and FALSE for mrb_bool in variable.cUkrainskiy Sergey
2017-11-17Need to save stack space by number of registers from the current proc,Yukihiro "Matz" Matsumoto
Not from the execution ensure proc; fix #3849
2017-11-13The number of argument should be retrieved by `mrb_get_argc`; fix #3848Yukihiro "Matz" Matsumoto
You should not access `mrb->c->ci->argc` directly.
2017-11-11Fixed wrong stack extend size; fix #3847Yukihiro "Matz" Matsumoto
2017-11-10Need to clear stack region for local variables in eval; fix #3844Yukihiro "Matz" Matsumoto
2017-11-09Use proper target_class to define class/module; fix #3843Yukihiro "Matz" Matsumoto
The outer class of the class/module definition should be taken from `MRB_TARGET_CLASS(mrb->c->ci->proc)` not `mrb->c->ci->target_class` which is the target of constant lookups.
2017-11-08Should allocate memory region before updating `len`; fix #3842Yukihiro "Matz" Matsumoto
Otherwise half-baked string object will be allocated.
2017-11-04Avoid compiling `fmt_fp.c` if `MRB_WITHOUT_FLOAT` is set; #3827Yukihiro "Matz" Matsumoto
2017-11-04Merge branch 'mrb_without_float' of https://github.com/pandax381/mruby into ↵Yukihiro "Matz" Matsumoto
pandax381-mrb_without_float
2017-11-04Reimplement `block_given?`; ref #3841Yukihiro "Matz" Matsumoto
Make `block_given?` to search for the top of the scope first. The top of the scope means either: * the top method body * the enclosing class body * the top-level The special case is the method defined by `define_method` with a block as in #3841. In cases like this, the method body (given by a block) is not considered as the top of the scope. You need to use `&block` in the block parameter if you want to know if a block is given to the method. This commit also changes the behavior of `MRB_PROC_SCOPE` flag. Now it is only set if the `proc` is either a class body or a method body defined in Ruby. It is no longer set for a block that given to `define_method`.
2017-11-04The `bidx` saved in `env` may be useless; fix #3841Yukihiro "Matz" Matsumoto
When `block_given?` is called from a block given to `define_method` as a method body, the `bidx` may not be within `env` saved closure. In this case, it causes heap buffer overflow.
2017-11-04Merge branch 'master' of github.com:mruby/mrubyYAMAMOTO Masaya
2017-11-04Avoid pointer arithmetic in backtrace.c; #3816Yukihiro "Matz" Matsumoto
2017-11-03Always check division-by-zero to avoid undefined behavior; fix #3816Yukihiro "Matz" Matsumoto
Also removed the code to normalize NaN value for `MRB_NAN_BOXING`. Tha code was added to fix #1712 but no longer required after 249f05e7d.
2017-11-03Should not use `FSHARED` state for string from `irep` pools; fix #3829Yukihiro "Matz" Matsumoto
This strings in `irep` pools may be freed forcefully in `mrb_irep_free`. This commit probably fixes #3817 as well.
2017-11-02don't overwrite backtrace info.Tomoyuki Sahara
2017-10-31Fixed constant (and class variable) reference bug; fix #3839Yukihiro "Matz" Matsumoto
Unlike method definition, constant reference should start from `MRB_PROC_TARGET_CLASS(ci->proc)`, not `ci->target_class`. In addition, `MRB_PROC_TARGET_CLASS(ci->proc)` is always set.
2017-10-31Call stack may not reference the destination `proc`; fix #3838Yukihiro "Matz" Matsumoto
The destination `proc` may be an orphan.
2017-10-31Need to mark fibers referenced from `env`; fix #3837Yukihiro "Matz" Matsumoto
The issue #3837 is a regression of #3819. I was mistakenly removed a important code to mark fibers from `env`.
generated by cgit v1.2.3 (git 2.39.1) at 2026-04-17 10:40:00 +0000