In rb_gc_impl_before_fork, it locks the VM and barriers all the Ractors
before calling mmtk_before_fork. However, since rb_mmtk_block_for_gc is
a barrier point, one or more Ractors could be paused there. However,
mmtk_before_fork is not compatible with that because it assumes that the
MMTk workers are idle, but the workers are not idle because they are
busy working on a GC.
This commit essentially implements a trylock. It will optimistically
lock but will release the lock if it detects that any other Ractors are
waiting in rb_mmtk_block_for_gc.
For example, the following script demonstrates the issue:
puts "Hello #{Process.pid}"
100.times do |i|
puts "i = #{i}"
Ractor.new(i) do |j|
puts "Ractor #{j} hello"
1000.times do |i|
s = "#{j}-#{i}"
end
Ractor.receive
puts "Ractor #{j} goodbye"
end
pid = fork { }
puts "Child pid is #{pid}"
_, status = Process.waitpid2 pid
puts status.success?
end
puts "Goodbye"
We can see the MMTk worker thread is waiting to start the GC:
#4 0x00007ffff66538b1 in rb_mmtk_stop_the_world () at gc/mmtk/mmtk.c:101
#5 0x00007ffff6d04caf in mmtk_ruby::collection::{impl#0}::stop_all_mutators<mmtk::scheduler::gc_work::{impl#14}::do_work::{closure_env#0}<mmtk::plan::immix::gc_work::ImmixGCWorkContext<mmtk_ruby::Ruby, 0>>> (_tls=..., mutator_visitor=...) at src/collection.rs:23
However, the mutator thread is stuck in mmtk_before_fork trying to stop
that worker thread:
#4 0x00007ffff6c0b621 in std::sys:🧵:unix::Thread::join () at library/std/src/sys/thread/unix.rs:134
#5 0x00007ffff6658b6e in std:🧵:JoinInner<()>::join<()> (self=...)
#6 0x00007ffff6658d4c in std:🧵:JoinHandle<()>::join<()> (self=...)
#7 0x00007ffff665795e in mmtk_ruby::binding::RubyBinding::join_all_gc_threads (self=0x7ffff72462d0 <mmtk_ruby::BINDING+8>) at src/binding.rs:115
#8 0x00007ffff66561a8 in mmtk_ruby::api::mmtk_before_fork () at src/api.rs:309
#9 0x00007ffff66556ff in rb_gc_impl_before_fork (objspace_ptr=0x555555d17980) at gc/mmtk/mmtk.c:1054
#10 0x00005555556bbc3e in rb_gc_before_fork () at gc.c:5429
https://github.com/ruby/mmtk/commit/1a629504a7
We need the VM barrier in rb_gc_impl_before_fork to stop the other Ractors
because otherwise they could be allocating objects in the fast path which
could be calling mmtk_add_obj_free_candidate. Since mmtk_add_obj_free_candidate
acquires a lock on obj_free_candidates in weak_proc.rs, this lock may not
be released in the child process after the Ractor dies.
For example, the following script demonstrates the issue:
puts "Hello #{Process.pid}"
100.times do |i|
puts "i = #{i}"
Ractor.new(i) do |j|
puts "Ractor #{j} hello"
1000.times do |i|
s = "#{j}-#{i}"
end
Ractor.receive
puts "Ractor #{j} goodbye"
end
pid = fork { }
puts "Child pid is #{pid}"
_, status = Process.waitpid2 pid
puts status.success?
end
puts "Goodbye"
In the child process, we can see that it is stuck trying to acquire the
lock on obj_free_candidates:
#5 0x00007192bfb53f10 in mmtk_ruby::weak_proc::WeakProcessor::get_all_obj_free_candidates (self=0x7192c0657498 <mmtk_ruby::BINDING+72>) at src/weak_proc.rs:52
#6 0x00007192bfa634c3 in mmtk_ruby::api::mmtk_get_all_obj_free_candidates () at src/api.rs:295
#7 0x00007192bfa61d50 in rb_gc_impl_shutdown_call_finalizer (objspace_ptr=0x578c17abfc50) at gc/mmtk/mmtk.c:1032
#8 0x0000578c1601e48e in rb_ec_finalize (ec=0x578c17ac06d0) at eval.c:166
#9 rb_ec_cleanup (ec=<optimized out>, ex=<optimized out>) at eval.c:257
#10 0x0000578c1601ebf6 in ruby_cleanup (ex=<optimized out>) at eval.c:180
#11 ruby_stop (ex=<optimized out>) at eval.c:292
#12 0x0000578c16127124 in rb_f_fork (obj=<optimized out>) at process.c:4291
#13 rb_f_fork (obj=<optimized out>) at process.c:4281
https://github.com/ruby/mmtk/commit/eb4b229858
- In 31d67ecc05
we enforced the new behaviour where running `bundle` no longer
installs gems but displays the help.
Users now have a way to configure their preferred default command using
the `BUNDLE_DEFAULT_CLI_COMMAND` flag.
With the preview of Ruby 4.0 now being released, some people will
start to see this new change.
The problem is that the previous behaviour had existed for like an
eternity and we didn't warn users about this change in advance.
I'd like to provide a deprecation/warning cycle because this is
confusing users already and this breaks various CI setup that now
needs to be changed immediately.
https://github.com/ruby/rubygems/commit/e415480ac5
This is useful, in case you're using Docker, and an upstream
Dockerfile sets BUNDLER_VERSION to something you don't want.
It's impossible to unset it... only override to be the empty
string.
https://github.com/ruby/rubygems/commit/ffa3eb9ac6
(https://github.com/ruby/rubygems/pull/8183)
* Add new gem templates
* Add `--ext=go` in `bundle gem`
* Add setup-go to .github/workflows/main.yml
* Embed go version in go.mod
* Use go in bundler CI
* Add example method to template
* Install Go in .circleci/config.yml
* Install Go in .gitlab-ci.yml
* Allow hard tabs in go template
* Run `rake update_manifest`
* Fix test
* Move go_gem to gemspec
Respect to 9b0ec80
* nits: ⛳
* includes valid module name in go.mod
* generate header file
* Run `go mod tidy` to create `go.sum`
* Check if `go.sum` is generated only when Go is installed
To avoid test failure in environments where Go is not installed
* Run CI
* Workaround for hung up
c.f. https://github.com/rubygems/rubygems/actions/runs/11639408044/job/32415545422
* Write man for --ext=go
* Re-generate man with `./bin/rake man:build`
* pinning 📌
* Update with `./bin/rake man:build`
* nits: Extract to method
* nits: Use `sys_exec` instead of `system`
* Clean go module cache after test
Workaround following error
```
1) bundle gem gem naming with underscore --ext parameter set with go includes go_gem extension in extconf.rb
Failure/Error: FileUtils.rm_r(dir)
Errno::EACCES:
Permission denied @ apply2files - /home/runner/work/rubygems/rubygems/bundler/tmp/2.2/home/go/pkg/mod/gopkg.in/yaml.v3@v3.0.1/decode_test.go
# ./spec/support/helpers.rb:37:in `block in reset!'
# ./spec/support/helpers.rb:21:in `each'
# ./spec/support/helpers.rb:21:in `reset!'
# ./spec/spec_helper.rb:130:in `block (2 levels) in <top (required)>'
# /home/runner/work/rubygems/rubygems/lib/rubygems.rb:303:in `load'
# /home/runner/work/rubygems/rubygems/lib/rubygems.rb:303:in `activate_and_load_bin_path'
```
Files installed with `go get` have permissions set to 444
ref. https://github.com/golang/go/issues/35615
```
$ ls -l /home/runner/work/rubygems/rubygems/bundler/tmp/2.2/home/go/pkg/mod/gopkg.in/yaml.v3@v3.0.1/decode_test.go
-r--r--r-- 1 runner runner 42320 Nov 15 06:38 /home/runner/work/rubygems/rubygems/bundler/tmp/2.2/home/go/pkg/mod/gopkg.in/yaml.v3@v3.0.1/decode_test.go
```
So they cannot be deleted by `FileUtils.rm_r`.
Therefore, this is necessary to execute `go clean -modcache` separately from `FileUtils.rm_r` to circumvent it.
* Remove needless changes
ref. https://github.com/ruby/rubygems/pull/8183#discussion_r2532902051
* ci: setup-go is needless
* Don't run go command in `bundle gem`
ref. https://github.com/ruby/rubygems/pull/8183#discussion_r2532765470
* Revert unrelated date changes
---------
https://github.com/ruby/rubygems/commit/260d7d60b3
Co-authored-by: Hiroshi SHIBATA <hsbt@ruby-lang.org>
`Gem::Installer.instance_methods(false).include?(:generate_plugins)` is
63x slower than `Gem::Installer.method_defined?(:generate_plugins)` in a
microbenchmark. The latter is a direct lookup, whereas the former will
create an array, which will be slower.
```ruby
require "benchmark/ips"
Benchmark.ips do |x|
x.report "instance_methods" do
Gem::Installer.instance_methods(false).include?(:generate_plugins)
end
x.report "method_defined" do
Gem::Installer.method_defined?(:generate_plugins)
end
x.compare!
end
```
```
$ ruby -I lib/ benchmark_methods.rb
ruby 3.4.4 (2025-05-14 revision https://github.com/ruby/rubygems/commit/a38531fd3f) +PRISM [arm64-darwin23]
Warming up --------------------------------------
instance_methods 58.449k i/100ms
method_defined 3.375M i/100ms
Calculating -------------------------------------
instance_methods 541.874k (± 5.7%) i/s (1.85 μs/i) - 2.747M in 5.087825s
method_defined 34.263M (± 1.1%) i/s (29.19 ns/i) - 172.135M in 5.024524s
Comparison:
method_defined: 34263189.1 i/s
instance_methods: 541874.3 i/s - 63.23x slower
```
This does not make much difference in an overall benchmark or profile,
but this is more idiomatic Ruby than the prior code.
https://github.com/ruby/rubygems/commit/49dec52cb2
The comptime receiver, which is a proc, is either shareable or from this
ractor so we don't need to assume single-ractor mode. We should never get
the "defined with an un-shareable Proc in a different ractor" error.
This implements Shopify#854:
- Splits boot-time and enable-time initialization,
tracks progress with `InitializationState` enum
- Introduces `RubyVM::ZJIT.enable` Ruby method for
enabling the JIT lazily, if not already enabled
- Introduces `--zjit-disable` flag, which can be
used alongside the other `--zjit-*` flags but
prevents enabling the JIT at boot time
- Adds ZJIT infra to support JIT hooks, but this
is not currently exercised (Shopify/ruby#667)
Left for future enhancements:
- Support kwargs for overriding the CLI flags in
`RubyVM::ZJIT.enable`
Closes Shopify#854
RHEL 9.7 ships OpenSSL 3.5.1 with ML-DSA support, but it is disabled
for TLS by default, according to the system configuration file:
/etc/crypto-policies/back-ends/opensslcnf.config
Specify SSLContext#sigalgs to override the default list.
https://github.com/ruby/openssl/commit/fac3a26748
