May 2023 - ruby-core - ml.ruby-lang.org

[ruby-core:113180] [Ruby master Feature#19588] Allow Comparable#clamp(min, max) to accept nil as a specification
by kyanagi (Kouhei Yanagita) 29 Nov '23

29 Nov '23

Issue #19588 has been reported by kyanagi (Kouhei Yanagita). ---------------------------------------- Feature #19588: Allow Comparable#clamp(min, max) to accept nil as a specification https://bugs.ruby-lang.org/issues/19588 * Author: kyanagi (Kouhei Yanagita) * Status: Open * Priority: Normal ---------------------------------------- `Comparable#clamp(min, max)` (with two arguments) accepts `nil`. This behaves the same as beginless/endless Range. ~~~ruby 5.clamp(nil, 0) # => 0 5.clamp(10, nil) # => 10 5.clamp(..0) # => 0 5.clamp(10..) # => 10 ~~~ This behavior is not documented. Presumably, this was not introduced intentionally. The old Rubies did not accept accept a `Range` argument. In Ruby 2.7, accepting `Range` as an argument was introduced. At that time, the approach of passing `nil` as a two-argument method was also discussed but not adopted, and using Range was chosen instead. https://bugs.ruby-lang.org/issues/14784 However, in Ruby 3.0, the behavior of `clamp` has changed to accept `nil`. This change is not documented in the NEWS or the documentation for `clamp`, and I believe that it was not an intentional change. ~~~ % docker run -it --rm rubylang/all-ruby env ALL_RUBY_SINCE=ruby-2.4.0 ./all-ruby -e "p 5.clamp(0, nil)" ruby-2.4.0 -e:1:in `clamp': comparison of Integer with nil failed (ArgumentError) from -e:1:in `<main>' exit 1 ... ruby-2.7.8 -e:1:in `clamp': comparison of Integer with nil failed (ArgumentError) from -e:1:in `<main>' exit 1 ruby-3.0.0-preview1 5 ... ruby-3.2.2 5 ~~~ It seems that https://github.com/ruby/ruby/commit/a93da4970be44a473b7b42e7516eb2663dece2c3 brought about this change. How about making the current behavior a specification? It has been three years since the behavior changed, and I don't see much point in prohibiting `nil` now. -- https://bugs.ruby-lang.org/

5 6

[ruby-core:111074] [Ruby master Bug#19160] cmp_clamp arguments
by kaiquekandykoga 13 Nov '23

13 Nov '23

Issue #19160 has been reported by kaiquekandykoga (Kaíque Koga). ---------------------------------------- Bug #19160: cmp_clamp arguments https://bugs.ruby-lang.org/issues/19160 * Author: kaiquekandykoga (Kaíque Koga) * Status: Open * Priority: Normal * ruby -v: ruby 3.1.2p20 (2022-04-12 revision 4491bb740a) [x86_64-freebsd13.1] * Backport: 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN ---------------------------------------- If clamp receives min higher than max, it will raise an exception. The message says *min argument must be smaller than max argument* , but min can actually be equal to max too. Patch https://github.com/ruby/ruby/pull/6802. -- https://bugs.ruby-lang.org/

2 1

[ruby-core:113107] [Ruby master Bug#19576] Backport request: Gemfile.lock resolving is broken with bundler shipped with Ruby 3.1.4
by jprokop (Jarek Prokop) 09 Nov '23

09 Nov '23

Issue #19576 has been reported by jprokop (Jarek Prokop). ---------------------------------------- Bug #19576: Backport request: Gemfile.lock resolving is broken with bundler shipped with Ruby 3.1.4 https://bugs.ruby-lang.org/issues/19576 * Author: jprokop (Jarek Prokop) * Status: Open * Priority: Normal * ruby -v: ruby 3.1.3p185 (2022-11-24 revision 1a6b16756e) [x86_64-linux] * Backport: 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN ---------------------------------------- This is a backport request for bundler, that regressed in 2.3.36 in a specific situation. Newer and older bundler versions that ship with Ruby are not problematic, only the version that ships with Ruby version >= 3.1.3. A few weeks ago, we discovered a bug in resolving in bundler shipped with Ruby 3.1.3 and 3.1.4: Bundler version: ``` $ bundler --version Bundler version 2.3.26 ``` Affected rubies `ruby -v`: First: ``` $ ruby -v ruby 3.1.4p223 (2023-03-30 revision 957bb7cb81) [x86_64-linux] ``` Second: ``` $ruby -v ruby 3.1.3p185 (2022-11-24 revision 1a6b16756e) [x86_64-linux] ``` Initial bug report with reproducer and more in-depth description can be found here: https://github.com/sclorg/s2i-ruby-container/issues/469 Using the following Gemfile for a rails app: https://github.com/sclorg/rails-ex/blob/67b7a61eae9efa1088ff3f634ae316e1022… bundler locks up in trying to resolve Nokogiri for Ruby 3.1, but keeps failing because it keeps using incompatible built binary gem instead of falling back to installing and building the binary extension of Nokogiri locally. We craft this Gemfile to be usable from Ruby 2.5 up to Ruby 3.1, as the app is used mainly for testing. I have created a patch to fix the situation, see the attached files. There are 2 of them, one contains the fix and the other one contains the test from the rubygems repo PR#6225. full commit available here: https://src.fedoraproject.org/fork/jackorp/rpms/ruby/c/5ef600a8f40b76de5636… The patches are created from the following upstream changes in bundler: https://github.com/rubygems/rubygems/pull/6225 and adapted: https://github.com/rubygems/rubygems/commit/7b64c64262a7a980c0eb23b96ea56cf… for the PR#6225. With the fix applied I no longer have issues doing `bundle install` with our Gemfile.lock. ---Files-------------------------------- rubygem-bundler-2.3.26-Tests-from-bundler-PR-6225.patch (1.82 KB) rubygem-bundler-2.3.26-Provide-fix-for-bundler-Gemfile-resolving-regression.patch (5.21 KB) -- https://bugs.ruby-lang.org/

5 7

[ruby-core:111953] [Ruby master Bug#19362] #dup on Proc doesn't call initialize_dup
by zverok (Victor Shepelev) 25 Oct '23

25 Oct '23

Issue #19362 has been reported by zverok (Victor Shepelev). ---------------------------------------- Bug #19362: #dup on Proc doesn't call initialize_dup https://bugs.ruby-lang.org/issues/19362 * Author: zverok (Victor Shepelev) * Status: Open * Priority: Normal * Backport: 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN ---------------------------------------- In #17545, `#dup` had changed to create an instance of the subclass. It, though, doesn't invoke `initialize_dup` of the subclass, unlike other standard classes. ```ruby class MyAry < Array def initialize_dup(...) p(self.class, ...) super end end class MyString < String def initialize_dup(...) p(self.class, ...) super end end class MyProc < Proc def initialize_dup(...) p(self.class, ...) super end end MyString.new('test').dup # prints MyString, "test" MyAry.new(['test']).dup # prints MyAry, ["test"] MyProc.new { 'test' }.dup # doesn't print anything ``` This makes the change in #17545 useless: while inheriting from core classes is indeed marginal, one of author's intention might be carrying additional information with the Proc instance, and bypassing `#initialize_dup` makes it impossible to maintain this information. It seems that actually `#initialize_dup` is also invoked on the core classes themselves, but ignored on `Proc`. ```ruby class Array def initialize_dup(...) p(self.class, ...) super end end class String def initialize_dup(...) p(self.class, ...) super end end class Proc def initialize_dup(...) p(self.class, ...) super end end 'test'.dup # prints String, "test" ['test'].dup # prints Array, ["test"] Proc.new { 'test' }.dup # doesn't print anything ``` Which is an even more marginal problem but still an inconsistency. -- https://bugs.ruby-lang.org/

3 7

[ruby-core:112944] [Ruby master Feature#19541] Proposal: Generate frame unwinding info for YJIT code
by kjtsanaktsidis (KJ Tsanaktsidis) 19 Oct '23

19 Oct '23

Issue #19541 has been reported by kjtsanaktsidis (KJ Tsanaktsidis). ---------------------------------------- Feature #19541: Proposal: Generate frame unwinding info for YJIT code https://bugs.ruby-lang.org/issues/19541 * Author: kjtsanaktsidis (KJ Tsanaktsidis) * Status: Open * Priority: Normal ---------------------------------------- ## What is being propsed? Currently, Ruby crashes with yjit generated code on the stack, `rb_print_backtrace()` is unable to actually show any frames underneath the yjit code. For example, if you send SIGSEGV to a Ruby process running yjit, this is what you see: ``` /ruby/miniruby(rb_print_backtrace+0xc) [0xaaaad0276884] /ruby/vm_dump.c:785 /ruby/miniruby(rb_vm_bugreport) /ruby/vm_dump.c:1093 /ruby/miniruby(rb_bug_for_fatal_signal+0xd0) [0xaaaad0075580] /ruby/error.c:813 /ruby/miniruby(sigsegv+0x5c) [0xaaaad01bedac] /ruby/signal.c:919 linux-vdso.so.1(__kernel_rt_sigreturn+0x0) [0xffff91a3e8bc] /ruby/miniruby(map<(usize, yjit::backend::ir::Insn), (usize, yjit::backend::ir::Insn), yjit::backend::ir::{impl#17}::next_mapped::{closure_env#0}>+0x8c) [0xaaaad03b8b00] /rustc/897e37553bba8b42751c67658967889d11ecd120/library/core/src/option.rs:929 /ruby/miniruby(next_mapped+0x3c) [0xaaaad0291dc0] src/backend/ir.rs:1225 /ruby/miniruby(arm64_split+0x114) [0xaaaad0287744] src/backend/arm64/mod.rs:359 /ruby/miniruby(compile_with_regs+0x80) [0xaaaad028bf84] src/backend/arm64/mod.rs:1106 /ruby/miniruby(compile+0xc4) [0xaaaad0291ae0] src/backend/ir.rs:1158 /ruby/miniruby(gen_single_block+0xe44) [0xaaaad02b1f88] src/codegen.rs:854 /ruby/miniruby(gen_block_series_body+0x9c) [0xaaaad03b0250] src/core.rs:1698 /ruby/miniruby(gen_block_series+0x50) [0xaaaad03b0100] src/core.rs:1676 /ruby/miniruby(branch_stub_hit_body+0x80c) [0xaaaad03b1f68] src/core.rs:2021 /ruby/miniruby({closure#0}+0x28) [0xaaaad02eb86c] src/core.rs:1924 /ruby/miniruby(do_call<yjit::core::branch_stub_hit::{closure_env#0}, *const u8>+0x98) [0xaaaad035ba3c] /rustc/897e37553bba8b42751c67658967889d11ecd120/library/std/src/panicking.rs:492 [0xaaaad035c9b4] ``` (n.b. - I compiled Ruby with `-fasynchronous-unwind-tables –rdynamic –g` in cflags to make sure gcc generates appropriate unwind info & keeps the symbol tables). Likewise, if you attach gdb to a Ruby process with yjit enabled, gdb can't show thread backtraces through yjit-generated code either. My proposal is that YJIT generate sufficient unwinding and debug information on all platforms to allow both `rb_print_backtrace()` and the platform's debugger (gdb/lldb/WinDbg) to show: * Full stack traces all the way back to `main`. That is, it should be possible to see frames _underneath_ `[0xaaaad035c9b4]` from the backtrace above. * Names for the dynamically generated yjit blocks (e.g. instead of `[0xaaaad035c9b4]`, we should see something like `yjit$$name_of_ruby_method`, where `name_of_ruby_method` is the `label` for the iseq this is JIT'd code for). ## Motivation I have a few motivations for wanting this. Firstly, I feel this functionality is independently useful. When Ruby crashes, the more information we can get, the more likely we are to find the root cause. Likewise, the same principle applies to debugging with gdb - you can get a fuller understanding of what the process is doing if you see the whole stack. I have often found attaching gdb to the Ruby interpreter helps in understanding problems in Ruby code or C extensions and is something I do relatively frequently; yjit breaking that will definitely be inconvenient for me! ## Implementation I have a draft implementation here on how I'd implement this: . It's currently missing tests & platform support (it only works on Linux aarch64). Also, it implements unwind info generation, so unwinding can work _through_ yjit code, but it does not currently emit symbols to give _names_ to those yjit frames. My PR contains a document which explains how the Linux interfaces for registering unwind info for JIT'd code work, so I won't duplicate that information here. The biggest implementation question I had is around the use of Rust crates. Currently, I prototyped my implementation using the gimli & object crates, for generating DWARF info and ELF binaries. However, the yjit build does purposefully does not use cargo & external crates for release builds. There are a few different ways we could go here: * Don't use the gimli & object crates; instead, re-implement all debug info & object file generation code in yjit. * Don't use the crates; instead, link againt C libraries to provide this functionality & call them from Rust (perhaps some combination of libelf, libdw, libbfd, or llvm might do what we need) * Use cargo after all for the release build & download the crates at build-time * Use cargo for the release build, but vendor everything, so the build doesn't need to download anything * Only make unwind info generation available in dev mode where cargo is used, and so mark the gimli/object dependencies as optional in Cargo.toml. We'd need to decide on one of these approaches for this proposal to work. I don't really have a strong sense of the pros/cons of each. (Side note - my PR actually depends on a _fork_ of gimli - I've been discussing adding the needed interfaces upstream here: https://github.com/gimli-rs/gimli/issues/648) ## Benchmarks I ran the yit-bench suite on my branch and compared it to Ruby master: * My branch: https://gist.github.com/KJTsanaktsidis/5741a9f64e5cd75cdf5fedd846091a4f * Ruby master: https://gist.github.com/KJTsanaktsidis/592d3ebcf98f6745dfa3efbd30a25acf This is a (very simple) comparison: ``` -------------- ------------ ------------ --------------- bench yjit (ms) branch (ms) branch/yjit (%) activerecord 97.5 98.5 101.03% hexapdf 2415.3 2458.2 101.78% liquid-c 61.9 63.1 101.94% liquid-render 135.3 135.0 99.78% mail 104.6 105.5 100.86% psych-load 1887.1 1922.0 101.85% railsbench 1544.4 1556.0 100.75% ruby-lsp 88.4 89.5 101.24% sequel 147.5 151.1 102.44% binarytrees 303 305.6 100.86% chunky_png 1075.8 1079.4 100.33% erubi 392.9 392.3 99.85% erubi_rails 14.7 14.7 100.00% etanni 792.3 791.4 99.89% fannkuchredux 3815.9 3813.6 99.94% lee 1030.2 1039.2 100.87% nbody 49.2 49.3 100.20% optcarrot 4142 4143.3 100.03% ruby-json 2860.7 2874.0 100.46% rubykon 7906.6 7904.2 99.97% 30k_ifelse 348.7 345.4 99.05% 30k_methods 828.6 831.8 100.39% cfunc_itself 28.8 28.9 100.35% fib 34.4 34.5 100.29% getivar 115.5 109.7 94.98% keyword_args 37.7 38.0 100.80% respond_to 26 26.1 100.38% setivar 33.8 33.5 99.11% setivar_object 208.7 194.3 93.10% str_concat 52.6 52.2 99.24% throw 23.8 24.1 101.26% -------------- ------------ ------------ --------------- ``` It seems like the performance impact of generating and registering the debug info is marginal. -- https://bugs.ruby-lang.org/

7 12

[ruby-core:113381] [Ruby master Bug#19624] Backticks - IO object leakage
by pineman 20 Sep '23

20 Sep '23

Issue #19624 has been reported by pineman (João Pinheiro). ---------------------------------------- Bug #19624: Backticks - IO object leakage https://bugs.ruby-lang.org/issues/19624 * Author: pineman (João Pinheiro) * Status: Open * Priority: Normal * ruby -v: ruby 3.2.2 (2023-03-30 revision e51014f9c0) [arm64-darwin22] * Backport: 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN ---------------------------------------- Hi, This code works on ruby 3.0.6: ```ruby `echo` ObjectSpace.each_object(IO) do |io| if ![STDIN, STDOUT, STDERR].include?(io) io.close end end ``` but raises `IOError` on 3.2.2: ``` minimal-repro-case.rb:8:in `close': uninitialized stream (IOError) ``` I found it started failing on ruby 3.1.0 and after, on macOS and Linux. This code is useful for closing unneeded IO objects in forked processes. It looks like backticks is 'leaking' IO objects, waiting for GC, and it didn't used to in 3.1.0. In ruby 3.1.0, inside `rb_f_backquote`, `rb_gc_force_recycle` was removed in favor of `RB_GC_GUARD`. I wonder if this has something to do with the problem. Is this code incorrect since ruby 3.1.0 or is it a bug in ruby? Thanks. ---Files-------------------------------- minimal-repro-case.rb (109 Bytes) -- https://bugs.ruby-lang.org/

4 9

[ruby-core:111572] [Ruby master Bug#19297] Don't download content from internet to execute Ruby test suite
by vo.x (Vit Ondruch) 06 Sep '23

06 Sep '23

Issue #19297 has been reported by vo.x (Vit Ondruch). ---------------------------------------- Bug #19297: Don't download content from internet to execute Ruby test suite https://bugs.ruby-lang.org/issues/19297 * Author: vo.x (Vit Ondruch) * Status: Open * Priority: Normal * ruby -v: ruby 3.2.0 (2022-12-25 revision a528908271) [x86_64-linux] * Backport: 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN ---------------------------------------- Trying to build Ruby 3.2.0 for Fedora and execute its test suite via `make check` as we always did [1], the test suite suddenly fails (while it was working with commit:git|c5eefb7f37): ~~~ ... snip ... C-API Util function ruby_strtod - converts a string to a double and returns the remaining string - returns 0 and the full string if there's no numerical value Finished in 45.737677 seconds 3827 files, 31635 examples, 177877 expectations, 0 failures, 0 errors, 0 tagged ./miniruby -I/builddir/build/BUILD/ruby-3.2.0/lib -I. -I.ext/common /builddir/build/BUILD/ruby-3.2.0/tool/runruby.rb --extout=.ext -- --disable-gems -C "/builddir/build/BUILD/ruby-3.2.0" bin/gem install --no-document \ --install-dir .bundle --conservative "bundler" "rake" "rspec:~> 3" #"ruby-prof" ERROR: Could not find a valid gem 'bundler' (>= 0), here is why: Unable to download data from https://rubygems.org/ - SocketError: Failed to open TCP connection to rubygems.org:443 (getaddrinfo: Temporary failure in name resolution) (https://rubygems.org/specs.4.8.gz) ERROR: Could not find a valid gem 'rspec' (~> 3), here is why: Unable to download data from https://rubygems.org/ - SocketError: Failed to open TCP connection to rubygems.org:443 (getaddrinfo: Temporary failure in name resolution) (https://rubygems.org/specs.4.8.gz) make: Leaving directory '/builddir/build/BUILD/ruby-3.2.0/redhat-linux-build' make: *** [uncommon.mk:1464: yes-test-syntax-suggest-prepare] Error 2 ~~~ This is obviously due to the test suite trying to download `rspec` from the internet, while Fedora builders does not have internet access (and won't ever have for security reasons). If I am not mistaken, this is caused by commit:git|cae53842735237ccf71a13873fd0d1ae7f165582. Now 1) Can this be fixed? 2) Can the tarball be always self contained? [1]: https://src.fedoraproject.org/rpms/ruby/blob/631163e3b8a51ed610528181aabe0d… -- https://bugs.ruby-lang.org/

8 11

[ruby-core:111499] [Ruby master Bug#19281] SyntaxError if first argument of command call has semicolon inside parenthesis
by tompng (tomoya ishida) 04 Sep '23

04 Sep '23

Issue #19281 has been reported by tompng (tomoya ishida). ---------------------------------------- Bug #19281: SyntaxError if first argument of command call has semicolon inside parenthesis https://bugs.ruby-lang.org/issues/19281 * Author: tompng (tomoya ishida) * Status: Open * Priority: Normal * Backport: 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN ---------------------------------------- These are syntax error ~~~ruby p (1;2),(3),(4) p (;),(),() a.b (1;2),(3),(4) a.b (;),(),() ~~~ I expect it to be syntax ok because the code below is syntax ok. ~~~ruby p (1),(2;3),(4;5) p (),(;),(;) a.b (1),(2;3),(4;5) a.b (),(;),(;) ~~~ It will be easy to traverse sexp if the sexp of first argument is same as others ~~~ruby Ripper.sexp "p (),(),()" # => [:program, [[:command, [:@ident, "p", [1, 0]], [:args_add_block, [[:paren, false], # [:paren, [[:void_stmt]]] [:paren, [[:void_stmt]]], [:paren, [[:void_stmt]]]], false]]]] Ripper.sexp "p (1),(2),(3)" # => [:program, [[:command, [:@ident, "p", [1, 0]], [:args_add_block, [[:paren, [:@int, "1", [1, 3]]], # [:paren, [[:@int, "1", [1, 3]]]] [:paren, [[:@int, "2", [1, 7]]]], [:paren, [[:@int, "3", [1, 11]]]]], false]]]] ~~~ -- https://bugs.ruby-lang.org/

3 2

[ruby-core:111269] [Ruby master Bug#19230] The openssl backend of securerandom is no longer needed
by mame (Yusuke Endoh) 30 Aug '23

30 Aug '23

Issue #19230 has been reported by mame (Yusuke Endoh). ---------------------------------------- Bug #19230: The openssl backend of securerandom is no longer needed https://bugs.ruby-lang.org/issues/19230 * Author: mame (Yusuke Endoh) * Status: Open * Priority: Normal * ruby -v: ruby 3.1.3p185 (2022-11-24 revision 1a6b16756e) [x86_64-linux] * Backport: 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN ---------------------------------------- securerandom first checks if Random.urandom is available ([Line 77](https://github.com/ruby/securerandom/blob/5bfe7d6c163f7a8a45af8d2fc377f…) and if not available, it uses the openssl backend as a degeneration. However, the openssl backend does not work because it internally uses Random.urandom ([Line 55](https://github.com/ruby/securerandom/blob/5bfe7d6c163f7a8a45af8d2fc377f…) to create a seed. This issue is found by @hanachin. ``` $ ruby -ve 'def Random.urandom(*); raise; end; require "securerandom"; p SecureRandom.bytes(10)' ruby 3.1.3p185 (2022-11-24 revision 1a6b16756e) [x86_64-linux] -e:1: warning: method redefined; discarding old urandom -e:1:in `urandom': unhandled exception from /home/mame/local/lib/ruby/3.1.0/securerandom.rb:75:in `singleton class' from /home/mame/local/lib/ruby/3.1.0/securerandom.rb:42:in `<module:SecureRandom>' from /home/mame/local/lib/ruby/3.1.0/securerandom.rb:41:in `<top (required)>' from <internal:/home/mame/local/lib/ruby/3.1.0/rubygems/core_ext/kernel_require.rb>:85:in `require' from <internal:/home/mame/local/lib/ruby/3.1.0/rubygems/core_ext/kernel_require.rb>:85:in `require' from -e:1:in `<main>' ``` There has been this bug since commit:abae70d6ed63054d7d01bd6cd80c1b5b98b93ba3, which made the urandom backend as default and left the openssl backend just for degeneration. I think no one need the openssl anymore because no one has reported this bug for such a long time. How about removing it? ```diff diff --git a/lib/securerandom.rb b/lib/securerandom.rb index 07ae048634..32b76a2137 100644 --- a/lib/securerandom.rb +++ b/lib/securerandom.rb @@ -14,7 +14,6 @@ # # It supports the following secure random number generators: # -# * openssl # * /dev/urandom # * Win32 # @@ -46,21 +45,6 @@ def bytes(n) private - def gen_random_openssl(n) - @pid = 0 unless defined?(@pid) - pid = $$ - unless @pid == pid - now = Process.clock_gettime(Process::CLOCK_REALTIME, :nanosecond) - OpenSSL::Random.random_add([now, @pid, pid].join(""), 0.0) - seed = Random.urandom(16) - if (seed) - OpenSSL::Random.random_add(seed, 16) - end - @pid = pid - end - return OpenSSL::Random.random_bytes(n) - end - def gen_random_urandom(n) ret = Random.urandom(n) unless ret @@ -77,13 +61,7 @@ def gen_random_urandom(n) Random.urandom(1) alias gen_random gen_random_urandom rescue RuntimeError - begin - require 'openssl' - rescue NoMethodError - raise NotImplementedError, "No random device" - else - alias gen_random gen_random_openssl - end + raise NotImplementedError, "No random device" end public :gen_random ``` -- https://bugs.ruby-lang.org/

6 5

[ruby-core:112399] [Ruby master Bug#19436] Call Cache for singleton methods can lead to "memory leaks"
by byroot (Jean Boussier) 23 Aug '23

23 Aug '23

Issue #19436 has been reported by byroot (Jean Boussier). ---------------------------------------- Bug #19436: Call Cache for singleton methods can lead to "memory leaks" https://bugs.ruby-lang.org/issues/19436 * Author: byroot (Jean Boussier) * Status: Open * Priority: Normal * Backport: 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN ---------------------------------------- Using "memory leaks" with quotes, because strictly speaking the memory isn't leaked, but it can nonetheless lead to large memory overheads. ### Minimal Reproduction ```ruby module Foo def bar end end def call_bar(obj) # Here the call cache we'll keep a ref on the method_entry # which then keep a ref on the singleton_class, making that # instance immortal until the method is called again with # another instance. # The reference chain is IMEMO(callcache) -> IMEMO(ment) -> ICLASS -> CLASS(singleton) -> OBJECT obj.bar end obj = Object.new obj.extend(Foo) call_bar(obj) id = obj.object_id obj = nil 4.times { GC.start } p ObjectSpace._id2ref(id) ``` ### Explanation Call caches keep a strong reference onto the "callable method entry" (CME), which itself keeps a strong reference on the called object class and in the cache of a singleton class, it keeps a strong reference onto the `attached_object` (instance). This means that any call site that calls a singleton method, will effectively keep a strong reference onto the last receiver. If the method is frequently called it's not too bad, but if it's infrequently called, it's effectively a (bounded) memory leak. And if the `attached_object` is big, the wasted memory can be very substantial. ### Practical Implications Once relative common API impacted by this is [Rails' `extending` API](https://api.rubyonrails.org/classes/ActiveRecord/QueryMethods.html#met…. This API allow to extend a "query result set" with a module. These query results set can sometimes be very big, especially since they keep references to the instantiated `ActiveRecord::Base` instances etc. ### Possible Solutions #### Only keep a weak reference to the CME The fairly "obvious" solution is to keep a weak reference to the CME, that's what I explored in https://github.com/ruby/ruby/pull/7272, and it seems to work. However in debug mode It does fail on an assertion during compaction, but it's isn't quite clear to me what the impact is. Additionally, something that makes me think this would be the right solution, is that call caches already try to avoid marking the class: ```c # vm_callinfo.h:275 struct rb_callcache { const VALUE flags; /* inline cache: key */ const VALUE klass; // should not mark it because klass can not be free'd // because of this marking. When klass is collected, // cc will be cleared (cc->klass = 0) at vm_ccs_free(). ``` So it appears that the class being also marked through the CME is some kind of oversight? #### Don't cache based on some heuristics If the above isn't possible or too complicated, an alternative would be to not cache CMEs found in singleton classes, except if it's the the singleton class of a `Class` or `Module`. It would make repeated calls to such methods slower, but the assumption is that it's unlikely that these CME would live very long. #### Make `Class#attached_object` a weak reference Alternatively we could make the `attached_object` a weak reference, which would drastically limit the amount of memory that may be leaked in such scenario. The downside is that `Class#attached_object` was very recently exposed in Ruby 3.2.0, so it means changing its semantic a bit. cc @peterzhu2118 @ko1 -- https://bugs.ruby-lang.org/

8 24