November 2025 - ruby-core - ml.ruby-lang.org

[ruby-core:123414] [Ruby Bug#21629] Ruby-3.4.7 fails to build using clang / llvm
by debo (David Bohman) 22 Jan '26

22 Jan '26

Issue #21629 has been reported by debo (David Bohman). ---------------------------------------- Bug #21629: Ruby-3.4.7 fails to build using clang / llvm https://bugs.ruby-lang.org/issues/21629 * Author: debo (David Bohman) * Status: Open * ruby -v: ruby 3.4.6 (2025-09-16 revision dbd83256b1) +PRISM [x86_64-darwin16] * Backport: 3.2: UNKNOWN, 3.3: UNKNOWN, 3.4: UNKNOWN ---------------------------------------- Here is the error I encountered. This is using clang version 21.1.3 as the compiler toolchain on Darwin. ``` DYLD_LIBRARY_PATH=.:../../.. ASAN_OPTIONS=detect_leaks=0 "clang -I../../../.ext/include/x86_64-darwin16 -I../../../../ruby-3.4.7/include -I../../../../ruby-3.4.7/ext/-test-/public_header_warnings -I/usr/local/include -D_XOPEN_SOURCE -D_DARWIN_C_SOURCE -D_DARWIN_UNLIMITED_SELECT -D_REENTRANT -fstack-protector-strong -U_FORTIFY_SOURCE -D_FORTIFY_SOURCE=2 -fdeclspec -O3 -fno-fast-math -ggdb3 -Wall -Wextra -Wextra-tokens -Wdeprecated-declarations -Wdivision-by-zero -Wdiv-by-zero -Wimplicit-function-declaration -Wimplicit-int -Wpointer-arith -Wshorten-64-to-32 -Wwrite-strings -Wold-style-definition -Wmissing-noreturn -Wno-cast-function-type -Wno-constant-logical-operand -Wno-long-long -Wno-missing-field-initializers -Wno-overlength-strings -Wno-parentheses-equality -Wno-self-assign -Wno-tautological-compare -Wno-unused-parameter -Wno-unused-value -Wunused-variable -Wmisleading-indentation -Wundef -fno-common -pipe -D_TEST_OK -Werror -c conftest.c" In file included from conftest.c:1: In file included from ../../../../ruby-3.4.7/include/ruby.h:38: In file included from ../../../../ruby-3.4.7/include/ruby/ruby.h:28: In file included from ../../../../ruby-3.4.7/include/ruby/internal/arithmetic.h:24: In file included from ../../../../ruby-3.4.7/include/ruby/internal/arithmetic/char.h:29: ../../../../ruby-3.4.7/include/ruby/internal/core/rstring.h:398:24: error: default initialization of an object of type 'struct RString' with const member leaves the object uninitialized [-Werror,-Wdefault-const-init-field-unsafe] 398 | struct RString retval; | ^ ../../../../ruby-3.4.7/include/ruby/internal/core/rbasic.h:86:17: note: member 'klass' declared 'const' here 86 | const VALUE klass; | ^ 1 error generated. checked program was: /* begin */ 1: #include "ruby.h" 2: 3: int main(int argc, char **argv) 4: { 5: return !!argv[argc]; 6: } /* end */ ``` The configuration is the following: ``` CC=clang CXX=clang++ CPPFLAGS=-I/usr/local/include LDFLAGS=-L/usr/local/lib ../ruby-3.4.7/configure --enable-shared --with-rdoc=ri,html ``` -- https://bugs.ruby-lang.org/

5 18

[ruby-core:123923] [Ruby Bug#21716] YJIT panic: "all PendingBranchRefs should be unique" in net/http response reading (infinite unwind loop)
by angelim＠angelim.com.br (Alexandre Angelim) 22 Jan '26

22 Jan '26

Issue #21716 has been reported by angelim(a)angelim.com.br (Alexandre Angelim). ---------------------------------------- Bug #21716: YJIT panic: "all PendingBranchRefs should be unique" in net/http response reading (infinite unwind loop) https://bugs.ruby-lang.org/issues/21716 * Author: angelim(a)angelim.com.br (Alexandre Angelim) * Status: Open * ruby -v: ruby 3.4.7 * Backport: 3.2: UNKNOWN, 3.3: UNKNOWN, 3.4: UNKNOWN ---------------------------------------- We encountered a YJIT panic in production that caused the process to enter an infinite logging loop for several hours. The panic occurs within `yjit/src/core.rs` when handling branch references. The trigger appears to be in `net/http`, specifically when reading a response body (possibly involving chunked encoding or zlib) inside a Datadog trace transport flush. Key Observations: The Panic: YJIT fails with all PendingBranchRefs should be unique when ready to construct a Block. The Loop: After the panic, the process fails to exit and instead logs `rust_begin_unwind` and `try_fold` iterator errors repeatedly, filling logs for hours. System Configuration Ruby Version: ruby 3.4.7 (2025-10-08 revision 7a5688e2a2) +YJIT +PRISM [aarch64-linux] Environment: Linux aarch64 (AWS Graviton/ARM) Reproduction Context While we cannot provide a standalone reproduction script yet, the crash happens intermittently during a Net::HTTP request initiated by datadog-ruby (v2.17.0) to flush traces to a Unix socket, which is wrapped by sentry-ruby (v5.26.0). Ruby Location: `/gems/net-http-0.6.0/lib/net/http/response.rb:604` Rust Location: `./yjit/src/core.rs:2403:27` Backtrace & Logs ``` ruby: YJIT has panicked. More info to follow... thread '<unnamed>' panicked at 'all PendingBranchRefs should be unique when ready to construct a Block', ./yjit/src/core.rs:2403:27 stack backtrace: 0: rust_begin_unwind at /rustc/.../library/std/src/panicking.rs:578:5 ... 28: yjit::core::branch_stub_hit at /workspace/ruby-3.4.7/yjit/src/core.rs:3492:9 /gems/net-http-0.6.0/lib/net/http/response.rb:604: [BUG] YJIT: panicked at 'all PendingBranchRefs should be unique when ready to construct a Block', ./yjit/src/core.rs:2403:27 -- Control frame information ----------------------------------------------- c:0047 p:0049 s:0248 e:000247 BLOCK /gems/net-http-0.6.0/lib/net/http/response.rb:604 c:0046 p:0098 s:0243 e:000242 METHOD /gems/net-http-0.6.0/lib/net/http/response.rb:588 c:0045 p:0004 s:0235 e:000234 METHOD /gems/net-http-0.6.0/lib/net/http/response.rb:593 ... c:0039 p:0033 s:0196 e:000195 BLOCK /gems/sentry-ruby-5.26.0/lib/sentry/net/http.rb:44 ... c:0033 p:0023 s:0158 e:000157 METHOD /gems/datadog-2.17.0/lib/datadog/core/transport/http/adapters/unix_socket.rb:43 c:0032 p:0066 s:0152 e:000151 METHOD /gems/datadog-2.17.0/lib/datadog/core/transport/http/adapters/net.rb:77 ``` The stack suggests the crash happens during `branch_stub_hit` resolution. The presence of sentry-ruby and datadog suggests complex interaction with Net::HTTP patching. The infinite loop of logs following the crash looks like this (repeated): ``` /usr/lib/libruby.so.3.4(try_fold<alloc::vec::into_iter::IntoIter<alloc::rc::Rc<yjit::core::PendingBranch> ... `` This suggests the panic handler itself is tripping over the corrupted `PendingBranch` state. We encountered a YJIT panic in production that caused the process to enter an infinite logging loop for several hours. The panic occurs within `yjit/src/core.rs` when handling branch references. The trigger appears to be in `net/http`, specifically when reading a response body (possibly involving chunked encoding or zlib) inside a Datadog trace transport flush. -- https://bugs.ruby-lang.org/

5 6

[ruby-core:123878] [Ruby Bug#21702] `UNIXSocket` on Windows: suprising results in `#recvfrom` and `#remote_address`
by trinistr (Alexander Bulancov) 13 Jan '26

13 Jan '26

Issue #21702 has been reported by trinistr (Alexander Bulancov). ---------------------------------------- Bug #21702: `UNIXSocket` on Windows: suprising results in `#recvfrom` and `#remote_address` https://bugs.ruby-lang.org/issues/21702 * Author: trinistr (Alexander Bulancov) * Status: Open * Backport: 3.2: UNKNOWN, 3.3: UNKNOWN, 3.4: UNKNOWN ---------------------------------------- Support for `UNIXSocket` on Windows was added in #19135. Through testing in ruby/spec, I identified unexpected results in two methods: 1. `#remote_address.to_s` always returns 110 bytes, compared to `#local_address.to_s` which returns only needed bytes. Example: ```ruby # test (@addr is remote_address, @b is the socket created with UNIXServer#accept): @addr.to_s.should == @b.local_address.to_s # failure: "\x01\x00D:/a/spec/spec/rubyspec_temp/2032_3/unix.sock\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00" == "\x01\x00D:/a/spec/spec/rubyspec_temp/2032_3/unix.sock\x00" ``` Note how the address is correct, and `#unix_path`s are indeed equal, but there are NUL bytes filling the string representation up to 110 bytes. This number does not depend on the length of path name. 2. Much more worryingly, `UNIXSocket#recvfrom` **dumps 2047 bytes of memory** as the remote address: ```ruby # test: @server.recvfrom(5).should == ['hello', ['AF_UNIX', '']] # failure: ["hello",["AF_UNIX","\x14\xB3v\x02\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x00\x01\x01\x00\x00\x00\x00\x00\x00\xAB\xCD\x12C\xFF\x7F<snip>"]] == ["hello", ["AF_UNIX", ""]] ``` This seems to be a result of `union_sockaddr` having a `char place_holder[2048]` member inside, though it isn't clear why is this used instead of `sockaddr_un` member. These results are obtained from GitHub runners, as I don't have a machine to test directly. This is the PR in ruby/spec which brought this up: https://github.com/ruby/spec/pull/1300. -- https://bugs.ruby-lang.org/

3 4

[ruby-core:112045] [Ruby master Bug#19378] Windows: Use less syscalls for faster require of big gems
by aidog (Andi Idogawa) 12 Jan '26

12 Jan '26

Issue #19378 has been reported by aidog (Andi Idogawa). ---------------------------------------- Bug #19378: Windows: Use less syscalls for faster require of big gems https://bugs.ruby-lang.org/issues/19378 * Author: aidog (Andi Idogawa) * Status: Open * Priority: Normal * ruby -v: 3.2.0 * Backport: 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN ---------------------------------------- Hello 🙂 ## Problem require is slow on windows for big gems. (example: require 'gtk3'=> 3 seconds+). This is a problem for people who want to make cross platform GUI apps with ruby. ## Possible Reason As touched on in [#15797](https://bugs.ruby-lang.org/issues/15797) it seems like require uses realpath, which is emulated on windows. It checks every parent directory. The same syscalls run many times. ## Testfile C:\tmp\speedtest\testrequire.rb: ``` ruby require __dir__ + "/helloworld1.rb" require __dir__ + "/helloworld2.rb" ``` ``` shell ruby --disable-gems C:\tmp\speedtest\testrequire.rb ``` ### Syscalls per File/Directory: 1. CreateFile 2. QueryInformationVolume 3. QueryIdInformation 4. QueryAllInformationFile 5. QueryNameInformationFile 6. QueryNameInformationFile 7. QueryNormalizedNameInformationFile 8. CloseFile ### Files/Directories checked 1. C:\tmp 2. C:\tmp\speedtest 3. C:\tmp\speedtest\helloworld1.rb 4. C:\tmp 5. C:\tmp\speedtest 6. C:\tmp\speedtest\helloworld2.rb For two required files Ruby had to do 8*6 = **48** syscalls. The syscalls orginate from rb_w32_reparse_symlink_p / lstat Rubygems live in subfolders with 9+ parts: "C:\Ruby32-x64\lib\ruby\gems\3.2.0\gems\glib2-4.0.8\lib\glib2\variant.rb" Each file takes 8 * 9 = **72**+ calls. For variant.rb it is **80** calls. The result for the syscalls don't change in such a short time, so it should be possible to cache it. With require_relative it's twice as many calls. ## Other testcases Same result: ``` ruby File.realpath __dir__ + "/helloworld1.rb" File.realpath __dir__ + "/helloworld2.rb" ``` ``` ruby File.stat __dir__ + "/helloworld1.rb" File.stat __dir__ + "/helloworld2.rb" ``` It does not happen in $LOAD_PATH.resolve_feature_path(__dir__ + "/helloworld1.rb") ## Request Would it be possible to cache the stat calls when using require? I tried to implement a cache inside the ruby source code, but failed. If not, is there now a way to combine ruby files into one? I previously talked about require here: [YJIT: Windows support lacking.](https://bugs.ruby-lang.org/issues/19325#note-11) ## How to reproduce Ruby versions: At least 3.0+, most likely older ones too. Tested using Ruby Installer 3.1 and 3.2. [Procmon Software by Sysinternals](https://learn.microsoft.com/en-us/sysinternals/downloads/proc… -- https://bugs.ruby-lang.org/

6 7

[ruby-core:122552] [Ruby Feature#21442] Make tsort to bundled gems
by hsbt (Hiroshi SHIBATA) 08 Jan '26

08 Jan '26

Issue #21442 has been reported by hsbt (Hiroshi SHIBATA). ---------------------------------------- Feature #21442: Make tsort to bundled gems https://bugs.ruby-lang.org/issues/21442 * Author: hsbt (Hiroshi SHIBATA) * Status: Open ---------------------------------------- In last year, I gave up to make `tsort` to bundled gems because `set` depends `tsort`. https://github.com/ruby/set/blob/29a253d2259e8d6d9b313dec8418ec5ed0caab08/l… Fortunately, `set` is now core class. So, there is no library depends on `tsort` in ruby repository. I propose to make `tsort` to bundled gems at the next release(not Ruby 3.5/4.0, for 4.0(?)/4.1) -- https://bugs.ruby-lang.org/

1 3

[ruby-core:123453] [Ruby Bug#21634] Combining read(1) with eof? causes dropout of results unexpectedly on Windows.
by YO4 (Yoshinao Muramatsu) 08 Jan '26

08 Jan '26

Issue #21634 has been reported by YO4 (Yoshinao Muramatsu). ---------------------------------------- Bug #21634: Combining read(1) with eof? causes dropout of results unexpectedly on Windows. https://bugs.ruby-lang.org/issues/21634 * Author: YO4 (Yoshinao Muramatsu) * Status: Open * ruby -v: ruby 3.5.0dev (2025-10-03T08:59:54Z master 5b2ec0eb1b) +PRISM [x64-mingw-ucrt] * Backport: 3.2: UNKNOWN, 3.3: UNKNOWN, 3.4: UNKNOWN ---------------------------------------- On Windows, when reading a file containing EOF(\x1A), using read(1) with IO#eof? causes unexpected dropout of results. ```ruby irb(main):001> IO.binwrite("txt", "abcd\x1A") => 5 irb(main):002> open("txt", "r") { p _1.read(1) until _1.eof? }; # works fine "a" "b" "c" "d" "\x1A" irb(main):003> open("txt", "rt") { p _1.read(1) until _1.eof? }; # has failure "b" "d" irb(main):004> ``` The problem disappeared when I commented out one of the following lines (though this will break other things). * previous_mode = set_binary_mode_with_seek_cur(fptr); in io_read() * flush_before_seek(fptr, false); in set_binary_mode_with_seek_cur(() * io_unread(fptr, discard_rbuf); in flush_before_seek() Within io_unread(), rbuf.len should have changed as 5, 4, 3,... but instead changed as 4, 2,(end). Since inconsistencies already exist at this point, the problem appears to originate elsewhere. I found this in ruby master but the same issue was found at least in ruby-1.9.3-p551. -- https://bugs.ruby-lang.org/

3 5

[ruby-core:123829] [Ruby Bug#21691] On Windows some of binary read functions of IO are not functional
by YO4 (Yoshinao Muramatsu) 08 Jan '26

08 Jan '26

Issue #21691 has been reported by YO4 (Yoshinao Muramatsu). ---------------------------------------- Bug #21691: On Windows some of binary read functions of IO are not functional https://bugs.ruby-lang.org/issues/21691 * Author: YO4 (Yoshinao Muramatsu) * Status: Open * Backport: 3.2: UNKNOWN, 3.3: UNKNOWN, 3.4: UNKNOWN ---------------------------------------- When using the binary read method of IO on a file containing “\r\n”, it should read “\r”, but some methods do not function correctly on Windows. ```ruby File.binwrite("crlf.txt", "\r\n") methods = { readn: proc { |f| f.read(1) }, getbyte: proc { |f| f.getbyte.chr }, readbyte: proc { |f| f.readbyte.chr }, readpartial: proc { |f| f.readpartial(1) }, read_nonblock: proc { |f| f.read_nonblock(1) }, sysread: proc { |f| f.sysread(1) }, pread: proc { |f| f.pread(1, 0) }, } def test(mode, name, method) open("crlf.txt", "r") do |f| p [mode, name, method.call(f)] end end methods.each { |k, v| test("r", k, v) } methods.each { |k, v| test("rt", k, v) } ``` results: ```ruby ["r", :readn, "\r"] ["r", :getbyte, "\n"] # => shold be "\r" ["r", :readbyte, "\n"] # => shold be "\r" ["r", :readpartial, "\n"] # => shold be "\r" ["r", :read_nonblock, "\n"] # => shold be "\r" ["r", :sysread, "\n"] # => shold be "\r" ["r", :pread, "\r"] ["rt", :readn, "\r"] ["rt", :getbyte, "\n"] # => shold be "\r" ["rt", :readbyte, "\n"] # => shold be "\r" ["rt", :readpartial, "\n"] # => shold be "\r" ["rt", :read_nonblock, "\n"] # => shold be "\r" ["rt", :sysread, "\n"] # => shold be "\r" ["rt", :pread, "\r"] ``` On Ubuntu in WSL, all returns are “\r”. -- https://bugs.ruby-lang.org/

2 1

[ruby-core:121699] [Ruby Feature#21277] Windows ARM support
by hsbt (Hiroshi SHIBATA) 08 Jan '26

08 Jan '26

Issue #21277 has been reported by hsbt (Hiroshi SHIBATA). ---------------------------------------- Feature #21277: Windows ARM support https://bugs.ruby-lang.org/issues/21277 * Author: hsbt (Hiroshi SHIBATA) * Status: Open * Assignee: hsbt (Hiroshi SHIBATA) ---------------------------------------- I added the new CI job for Windows ARM support. * https://github.blog/changelog/2025-04-14-windows-arm64-hosted-runners-now-a… * https://github.com/ruby/ruby/pull/13118 We could build Ruby with Windows ARM platform. But `nmake check` is not working yet. I'm tracking that issues on this ticket. ## Can't generate `revision.h` `windows-11-arm` runner couldn't generate `revision.h` with `ruby tool/file2lastrev.rb --revision.h --output=revision.h`. I tried that with git provided by `scoop`. But that is also not working. I confirmed to work `git rev-parse HEAD` directly with `pwsh` of GitHub Actions. But `tool/file2lastrev.rb` returns the followings: ``` Run ruby tool/file2lastrev.rb --revision.h --output=revision.h file2lastrev.rb: git rev-parse failed fatal: unable to access 'NUL': Invalid argument ``` ## `nmake check` is not working https://github.com/ruby/ruby/actions/runs/14564201209/job/40851166866 ``` running file: C:/a/ruby/ruby/src/test/rubygems/test_gem_resolver_index_set.rb A test worker crashed. It might be an interpreter bug or a bug in test/unit/parallel.rb. Try again without the -j option. ``` https://github.com/ruby/ruby/actions/runs/14564667517/job/40852193517 ``` NMAKE : fatal error U1077: '@ .\ruby.exe -I../src/lib -I".ext/arm64-mswin64_140" -I. "../src/tool/runruby.rb" --extout=".ext" -- --disable-gems -r./arm64-mswin64_140-fake -r../src/tool/lib/_tmpdir ../src/spec/mspec/bin/mspec run -B ../src/spec/default.mspec ' : return code '0xc0000005' Stop. ``` They may be caused by Visual Studio 2022 issue same as #21255. -- https://bugs.ruby-lang.org/

1 1

[ruby-core:120488] [Ruby master Feature#21005] Update the source location method to include line start/stop and column start/stop details
by bkuhlmann (Brooke Kuhlmann) 30 Dec '25

30 Dec '25

Issue #21005 has been reported by bkuhlmann (Brooke Kuhlmann). ---------------------------------------- Feature #21005: Update the source location method to include line start/stop and column start/stop details https://bugs.ruby-lang.org/issues/21005 * Author: bkuhlmann (Brooke Kuhlmann) * Status: Open ---------------------------------------- ## Why 👋 Hello. After discussing with Kevin Newton and Benoit Daloze in [Feature 20999](https://bugs.ruby-lang.org/issues/20999), I'd like to propose adding line start/stop and column start/stop information to the `#source_location` method for the following objects: - [Binding](https://docs.ruby-lang.org/en/master/Binding.html) - [Proc](https://docs.ruby-lang.org/en/master/Proc.html) - [Method](https://docs.ruby-lang.org/en/master/Method.html) - [UnboundMethod](https://docs.ruby-lang.org/en/master/UnboundMethod.html) At the moment, when using `#source_location`, you only get the following information: ``` ruby def demo = "A demonstration." # From disk. method(:demo).source_location # ["/Users/bkuhlmann/Engineering/Misc/demo", 15] # From memory. method(:demo).source_location # ["(irb)", 3] ``` Notice, when asking for the source location, we only get the path/location as the first element and the line number as the second element but I'd like to obtain a much richer set of data which includes line start/stop and column start/stop so I can avoid leaning on the `RubyVM` for this information. Example: ``` ruby def demo = "A demonstration." # From disk. instructions = RubyVM::InstructionSequence.of method(:demo) puts [instructions.absolute_path, *instructions.to_a.dig(4, :code_location)] [ "/Users/bkuhlmann/Engineering/Misc/demo", # Source path. 15, # Line start. 0, # Column start. 15, # Line stop. 29 # Column stop. ] # From memory. instructions = RubyVM::InstructionSequence.of method(:demo) puts instructions.script_lines [ "def demo = \"A demonstration.\"\n", "" ] ``` By having access to the path (or lack thereof in case of IRB), line start/stop, and column start/stop, this means we could avoid using the RubyVM to obtain raw source code for any of these objects. This would not only enhance debugging situations but also improve Domain Specific Languages that wish to leverage this information for introducing new features and/or new debugging capabilities to the language. ## How Building upon the examples provided above, I'd like to see `Binding`, `Proc`, `Method`, and `UnboundMethod` respond to `#source_location` as follows: ``` ruby [ "/Users/bkuhlmann/Engineering/Misc/demo", # Source path. 15, # Line start. 15, # Line stop. 0, # Column start. 29 # Column stop. ] ``` Notice, for data grouping purposes, I changed the array structure to always start with the path as the first element, followed by line information, and ending with column information. Alternatively, it could might be nice to improve upon the above by answering a hash each time, instead, for a more self-describing data structure. Example: ``` ruby { path: "/Users/bkuhlmann/Engineering/Misc/demo", line_start: 15, line_stop: 15, column_start: 0, column_stop: 29 } ``` For in-memory, situations like IRB, it would be nice to answer the equivalent of `RubyVM::InstructionSequence#script_lines` which would always be an `Array` with no line or column information since only the source code is necessary. Example: ``` ruby [ "def demo = \"A demonstration.\"\n", "" ] ``` From a pattern matching perspective, this could provide the best of both worlds especially if information is answered as either a `Hash` or and `Array`. Example: ``` def demo = "A demonstration." case method(:demo).source_location in Hash then puts "Source information obtained from disk." in Array then puts "Source obtained from memory." else fail TypeError, "Unrecognized source location type." end ``` This above is only a simple example but there's a lot we could do with this information if the above pattern match was enhanced to deal with the extraction and formatting of the actual source code! ## Notes This feature request is related to the following discussions in case more context is of help: - [Feature 6012](https://bugs.ruby-lang.org/issues/6012) - [Feature 20999](https://bugs.ruby-lang.org/issues/20999) -- https://bugs.ruby-lang.org/

8 19

[ruby-core:120749] [Ruby master Feature#21084] Declare objects have weak references
by peterzhu2118 (Peter Zhu) 30 Dec '25

30 Dec '25

Issue #21084 has been reported by peterzhu2118 (Peter Zhu). ---------------------------------------- Feature #21084: Declare objects have weak references https://bugs.ruby-lang.org/issues/21084 * Author: peterzhu2118 (Peter Zhu) * Status: Open ---------------------------------------- # Summary The current way of marking weak references uses `rb_gc_mark_weak(VALUE *ptr)`. This presents challenges because Ruby's GC is incremental, meaning that if the `ptr` changes (e.g. realloc'd or free'd), then we could have an invalid memory access. This also overwrites `*ptr = Qundef` if `*ptr` is dead, which prevents any cleanup to be run (e.g. freeing memory or deleting entries from hash tables). This ticket proposes `rb_gc_declare_weak_references` which declares that an object has weak references and calls a cleanup function after marking, allowing the object to clean up any memory for dead objects. # Introduction In [[Feature #19783]](https://bugs.ruby-lang.org/issues/19783), I introduced an API allowing objects to mark weak references, the function signature looks like this: ```c void rb_gc_mark_weak(VALUE *ptr); ``` `rb_gc_mark_weak` is called during the marking phase of the GC to specify that the memory at `ptr` holds a pointer to a Ruby object that is weakly referenced. `rb_gc_mark_weak` appends this pointer to a list that is processed after the marking phase of the GC. If the object at `*ptr` is no longer alive, then it overwrites the object reference with a special value (`*ptr = Qundef`). However, this API resulted in two challenges: 1. Ruby's default GC is incremental, which means that the GC is not ran in one phase, but rather split into chunks of work that interleaves with Ruby execution. The `ptr` passed into `rb_gc_mark_weak` could be on the malloc heap, and that memory could be realloc'd or even free'd. We had to use workarounds such as `rb_gc_remove_weak` to ensure that there were no illegal memory accesses. This made `rb_gc_mark_weak` difficult to use, impacted runtime performance, and increased memory usage. 2. When an object dies, `rb_gc_mark_weak` only overwites the reference with `Qundef`. This means that if we want to do any cleanup (e.g. free a piece of memory or delete a hash table entry), we could not do that and had to defer this process elsewhere (e.g. during marking or runtime). # Declarative weak references In this ticket, I'm proposing a new API for weak references. Instead of an object marking its weak references during the marking phase, the object declares that it has weak references using the `rb_gc_declare_weak_references` function. This declaration occurs during runtime (e.g. after the object has been created) rather than during GC. After an object declares that it has weak references, it will have its callback function called after marking as long as that object is alive. This callback function can then call a special function `rb_gc_handle_weak_references_alive_p` to determine whether its references are alive. This will allow the callback function to do whatever it wants on the object, allowing it to perform any cleanup work it needs. This significantly simplifies the code for `ObjectSpace::WeakMap` and `ObjectSpace::WeakKeyMap` because it no longer needs to have the workarounds for the limitations of `rb_gc_mark_weak`. # Performance The performance results below demonstrate that `ObjectSpace::WeakMap#[]=` is now about 60% faster because the implementation has been simplified and the number of allocations has been reduced. We can see that there is not a significant impact on the performance of `ObjectSpace::WeakMap#[]`. Base: ``` ObjectSpace::WeakMap#[]= 4.620M (± 6.4%) i/s (216.44 ns/i) - 23.342M in 5.072149s ObjectSpace::WeakMap#[] 30.967M (± 1.9%) i/s (32.29 ns/i) - 154.998M in 5.007157s ``` Branch: ``` ObjectSpace::WeakMap#[]= 7.336M (± 2.8%) i/s (136.31 ns/i) - 36.755M in 5.013983s ObjectSpace::WeakMap#[] 30.902M (± 5.4%) i/s (32.36 ns/i) - 155.901M in 5.064060s ``` Code: ``` require "bundler/inline" gemfile do source "https://rubygems.org" gem "benchmark-ips" end wmap = ObjectSpace::WeakMap.new key = Object.new val = Object.new wmap[key] = val Benchmark.ips do |x| x.report("ObjectSpace::WeakMap#[]=") do |times| i = 0 while i < times wmap[Object.new] = Object.new i += 1 end end x.report("ObjectSpace::WeakMap#[]") do |times| i = 0 while i < times wmap[key] wmap[val] # does not exist i += 1 end end end ``` # Alternative designs Currently, `rb_gc_declare_weak_references` is designed to be an internal-only API. This allows us to assume the object types that call `rb_gc_declare_weak_references`. In the future, if we want to open up this API to third parties, we may want to change this function to something like: ```c void rb_gc_add_cleaner(VALUE obj, void (*callback)(VALUE obj)); ``` This will allow the third party to implement a custom `callback` that gets called after the marking phase of GC to clean up any dead references. I chose not to implement this design because it is less efficient as we would need to store a mapping from `obj` to `callback`, which requires extra memory. -- https://bugs.ruby-lang.org/

4 6