Issue #19554 has been updated by mdalessio (Mike Dalessio). I'm just here to support Peter's take: Valgrind is a very useful tool for C extension maintainers to find and prevent illegal memory access. It's been an incredibly valuable tool for me over the ~15 years I've been a maintainer, and is a key tool in Nokogiri's CI pipelines. ASan is also useful! But because it's a build-time option, it's slightly less useful for debugging production builds. If it's not too difficult, I think Ruby should support both tools. ---------------------------------------- Bug #19554: Invalid memory access detected by Valgrind when using Fibers https://bugs.ruby-lang.org/issues/19554#change-102594 * Author: peterzhu2118 (Peter Zhu) * Status: Open * Priority: Normal * Backport: 2.7: UNKNOWN, 3.0: UNKNOWN, 3.1: UNKNOWN, 3.2: UNKNOWN ---------------------------------------- This issue was originally reported here: https://github.com/Shopify/ruby_memcheck/issues/14 Running the script shown below using `valgrind --trace-children=yes --num-callers=500 ruby test.rb` outputs a lot of invalid memory access errors. I've shown a few sample errors below. I am able to reproduce this issue on the master branch (commit [1e9a218ade](https://github.com/ruby/ruby/commit/1e9a218ade3af90c18f42e3fea08e2fcea81222a)), 3.2, 3.1, 3.0. ```ruby require "bundler/inline" gemfile do source "https://rubygems.org" gem "graphql" end module Example class FooType < GraphQL::Schema::Object field :id, ID, null: false end class FooSource < GraphQL::Dataloader::Source def fetch(ids) ids end end class QueryType < GraphQL::Schema::Object field :foo, Example::FooType do argument :foo_id, GraphQL::Types::ID, required: false, loads: Example::FooType end def foo(foo: nil); end end class Schema < GraphQL::Schema query Example::QueryType use GraphQL::Dataloader def self.object_from_id(id, ctx) ctx.dataloader.with(Example::FooSource).request(id) end end end Example::Schema.execute(<<-GRAPHQL) { foo(fooId: "Other") { id } } GRAPHQL ``` ``` ==203957== Use of uninitialised value of size 8 ==203957== at 0x3453FD: vm_exec_core (vm.inc:4411) ==203957== by 0x357EFB: rb_vm_exec (vm.c:2366) ==203957== by 0x354E44: invoke_block (vm.c:1384) ==203957== by 0x355759: invoke_iseq_block_from_c (vm.c:1440) ==203957== by 0x355759: invoke_block_from_c_proc (vm.c:1538) ==203957== by 0x355759: vm_invoke_proc (vm.c:1568) ==203957== by 0x355DF4: rb_vm_invoke_proc (vm.c:1589) ==203957== by 0x48F695: rb_fiber_start (cont.c:2513) ==203957== by 0x48CCF8: fiber_entry (cont.c:831) ==203957== ==203957== Invalid write of size 8 ==203957== at 0x48C407: fiber_pool_stack_reset (cont.c:325) ==203957== by 0x48C4E9: fiber_pool_vacancy_reset (cont.c:364) ==203957== by 0x48CBB0: fiber_pool_stack_release (cont.c:752) ==203957== by 0x48CECF: fiber_stack_release (cont.c:874) ==203957== by 0x48FC9F: fiber_switch (cont.c:2726) ==203957== by 0x4901F9: fiber_resume_kw (cont.c:2906) ==203957== by 0x490235: rb_fiber_resume_kw (cont.c:2912) ==203957== by 0x4903B7: rb_fiber_m_resume (cont.c:2973) ==203957== by 0x3337D6: ractor_safe_call_cfunc_m1 (vm_insnhelper.c:3166) ==203957== by 0x33440A: vm_call_cfunc_with_frame_ (vm_insnhelper.c:3357) ==203957== by 0x3345E1: vm_call_cfunc_with_frame (vm_insnhelper.c:3385) ==203957== by 0x3398E5: vm_sendish (vm_insnhelper.c:5225) ==203957== by 0x341203: vm_exec_core (insns.def:835) ==203957== by 0x357EFB: rb_vm_exec (vm.c:2366) ==203957== by 0x354E44: invoke_block (vm.c:1384) ==203957== by 0x355759: invoke_iseq_block_from_c (vm.c:1440) ==203957== by 0x355759: invoke_block_from_c_proc (vm.c:1538) ==203957== by 0x355759: vm_invoke_proc (vm.c:1568) ==203957== by 0x355DF4: rb_vm_invoke_proc (vm.c:1589) ==203957== by 0x48F695: rb_fiber_start (cont.c:2513) ==203957== by 0x48CCF8: fiber_entry (cont.c:831) ==203957== Address 0x9bad008 is in a rw- anonymous segment ``` -- https://bugs.ruby-lang.org/