Previously, a user of the debug inspector API was supposed to use
`rb_debug_inspector_backtrace_locations` to get an array of
`Thread::Backtrace::Location`, and then used its index to get more
information from `rb_debug_inspector _frame_binding_get(index)`, etc.
However, `rb_debug_inspector_backtrace_locations` returned an array of
backtraces excluding rescue/ensure frames. On the other hand,
`rb_debug_inspector_frame_binding_get(index)` interprets index with
rescue/ensure frames. This led to inconsistency of the index, and it was
very difficult to correctly use the debug inspector API.
This is a minimal fix for the issue by making
`rb_debug_inspector_backtrace_locations` return a raw backtrace
including rescue/ensure frames.
The FrameState on the SendWithoutBlock represents the entry state,
but for instructions that push to the stack in the middle of the
instruction before actually doing the send like opt_aref_with,
the FrameState is incorrect. We need to write to the stack using the
arguments for the instruction.
`Ractor#take` was deprecated but some libraries can use it as
an alias for `Ractor#value` (i.e., to wait for a Ractor's
temrination and retrieve its result).
Therefore `Ractor#take` is simply an alias for `Ractor#value`.
This method will remain available until the end of August 2025,
unless there is further discussion.
May not be supported on some 32bit architectures.
```
/usr/lib/gcc-cross/m68k-linux-gnu/14/../../../../m68k-linux-gnu/bin/ld: ../../libruby-static.a(vm.o): in function `rbimpl_atomic_u64_set_relaxed':
/home/ubuntu/build/ruby/master/m68k-linux/../src/ruby_atomic.h:60:(.text+0x2468): undefined reference to `__atomic_store_8'
/usr/lib/gcc-cross/m68k-linux-gnu/14/../../../../m68k-linux-gnu/bin/ld: ../../libruby-static.a(vm.o): in function `rbimpl_atomic_u64_load_relaxed':
/home/ubuntu/build/ruby/master/m68k-linux/../src/ruby_atomic.h:43:(.text+0x2950):
undefined reference to `__atomic_load_8'
```
Freezing an object changes its `shape_id` This is necessary
so that `setivar` routines can use the `shape_id` as a cache key
and save on checking the frozen status every time.
However for `getivar` routines, this causes needless cache misses.
By clearing that bit we increase hit rate in codepaths that see
both frozen and mutable objects.
Instead `shape_id_t` higher bits contain flags, and the first one
tells whether the shape is frozen.
This has multiple benefits:
- Can check if a shape is frozen with a single bit check instead of
dereferencing a pointer.
- Guarantees it is always possible to transition to frozen.
- This allow reclaiming `FL_FREEZE` (not done yet).
The downside is you have to be careful to preserve these flags
when transitioning.
Followup: https://github.com/ruby/ruby/pull/13341 / [Feature #21353]
Even thought `shape_id_t` has been make 32bits, we were still limited
to use only the lower 16 bits because they had to fit alongside `attr_index_t`
inside a `uintptr_t` in inline caches.
By enlarging inline caches we can unlock the full 32bits on all
platforms, allowing to use these extra bits for tagging.
Some GC implementations want to always know when an object is written to,
even if the written value is a special constant. Checking special constants
in rb_obj_written was a micro-optimization that made assumptions about
the GC implementation.
[Bug #21394]
There are two ways to make RubyVM::InstructionSequence.new raise which
would cause the options->scopes to leak memory:
1. Passing in any (non T_FILE) object where the to_str raises.
2. Passing in a T_FILE object where String#initialize_dup raises. This is
because rb_io_path dups the string.
Example 1:
10.times do
100_000.times do
RubyVM::InstructionSequence.new(nil)
rescue TypeError
end
puts `ps -o rss= -p #{$$}`
end
Before:
13392
17104
20256
23920
27264
30432
33584
36752
40032
43232
After:
9392
11072
11648
11648
11648
11712
11712
11712
11744
11744
Example 2:
require "tempfile"
MyError = Class.new(StandardError)
String.prepend(Module.new do
def initialize_dup(_)
if $raise_on_dup
raise MyError
else
super
end
end
end)
Tempfile.create do |f|
10.times do
100_000.times do
$raise_on_dup = true
RubyVM::InstructionSequence.new(f)
rescue MyError
else
raise "MyError was not raised during RubyVM::InstructionSequence.new"
end
puts `ps -o rss= -p #{$$}`
ensure
$raise_on_dup = false
end
end
Before:
14080
18512
22000
25184
28320
31600
34736
37904
41088
44256
After:
12016
12464
12880
12880
12880
12912
12912
12912
12912
12912
Let them run against the version resolved by the `test_gems.rb` gemfile.
This should fix ruby-core CI job that was broken by the release of rake
13.3.0.
