feat: complete runtime isolation and cloning phases 10-16 - #926
Merged
Conversation
Implement Perl threads plan phases 10, 11, and 12 as one maintainer-requested change while keeping both compiler backends functional and thread feature flags disabled. Phase 10 moves lexical hints, warning registries, feature flags, source maps, source filters, and end-of-scope hook state into the bound PerlRuntime. The global compilation lock continues to protect compiler-only parser scratch. Phase 11 completes the mutable runtime-state inventory across lifecycle and reachability tracking, weak references and DESTROY dispatch, signals and alarms, filesystem/stat/random/data-section state, debugger state, process and I/O registries, native/module registries, and blessing identity. Background callbacks that access Perl state capture and bind their owning runtime, while immutable caches and synchronized process services remain process-global. Phase 12 adds idempotent initialization and close plus exclusive reentrant managed execution. Runtime close releases alarms, signals, handles, native registries, and lifecycle roots. Direct bind remains the explicit context propagation primitive; managed execute serializes ownership of one runtime. Tests cover nested and concurrent compilation state, conflicting source locations, runtime-owned miscellaneous state, independent JVM/interpreter programs, close cleanup, and alarm callback binding. The full make passes. make test-all reaches 83.1% Perl core and 81.0% bundled modules. Scalar::Util 1.70 and lifecycle smokes pass on both backends; the JVM Moo constructor smoke passes, while the known interpreter Moo syntax limitation remains unchanged. Same-base three-run benchmark medians remain within the 5% gate: global +3.8%, lexical +7.2%, method -3.1%, closure +7.1%, regex -1.8%, and eval-string +1.1%. Profile-guided runtime lookup batching and idle lifecycle fast paths recover the initially observed hot-path regressions without weakening state isolation. Update dev/design/concurrency.md with the validation record and Phase 13 handoff. Completed implementation history remains recoverable from commit and pull-request messages rather than a growing completed-phases section. Generated with [Codex](https://openai.com/codex) Co-Authored-By: Codex <223018245+openai-codex[bot]@users.noreply.github.com>
Keep runtime-isolation tests platform-native, execute string-form Windows pipe commands through cmd.exe so built-ins remain available, validate relative Windows PATH entries in taint mode, and preserve lexical Windows SYSTEM IDs after Xerces URI normalization. Generated with [Codex](https://openai.com/codex) Co-Authored-By: Codex <223018245+openai-codex[bot]@users.noreply.github.com>
Keep File::Spec::Unix path operations host-independent, use stable cmd command parsing, and support the portable bare type output-pipe idiom on Windows. This completes the fixes for the remaining PR 926 CI failures. Generated with [Codex](https://openai.com/codex) Co-Authored-By: Codex <223018245+openai-codex[bot]@users.noreply.github.com>
Parse quoted Windows one-string commands into argv when shell syntax appears only inside arguments. This keeps quoted jperl.bat invocations out of cmd.exe's nested quote handling while retaining cmd for real shell expressions. Generated with [Codex](https://openai.com/codex) Co-Authored-By: Codex <223018245+openai-codex[bot]@users.noreply.github.com>
Record the measured Phase 13 benchmark gate as complete and advance the threads implementation plan to the identity-preserving graph cloner. The six same-base benchmark medians are all within the required five percent regression limit. Historical implementation and validation details remain in the phase commit and pull-request history. Generated with [Codex](https://openai.com/codex) Co-Authored-By: Codex <223018245+openai-codex[bot]@users.noreply.github.com>
Implement the Phase 14 ithread graph-cloning foundation with a single identity map across roots. Preserve scalar-slot identity, aggregate aliases and cycles, blessings, weak edges, readonly values, and tie wrappers without Storable hooks. Define the initial conservative resource policy: nonportable Java I/O handles become undef in the cloned graph. Register aggregate shells before recursion and defer weak-edge installation until the strong graph is complete, leaving the source graph untouched. Validation: full make gate and focused graph tests pass in all five shards. Generated with [Codex](https://openai.com/codex) Co-Authored-By: Codex <223018245+openai-codex[bot]@users.noreply.github.com>
Add explicit generated capture metadata for JVM closures and clone them without reflecting over Java field order. Rebuild interpreter closures from immutable bytecode metadata while cloning their constants, captures, state variables, recursive self references, and lexical maps through the Phase 14 identity map. Cross-backend tests prove cloned scalar, array, and hash captures evolve independently from their source closures. The full make gate passes. Generated with [Codex](https://openai.com/codex) Co-Authored-By: Codex <223018245+openai-codex[bot]@users.noreply.github.com>
Compose runtime package state and the identity-preserving graph cloner into PerlRuntime.snapshotClone(). Materialize lazy hook definitions and run CLONE_SKIP in the parent before copying, then invoke CLONE in deterministic package order after the child graph is installed. Clone globals and CODE slots through one graph map while leaving execution, lifecycle, signal/alarm, native, classloader, cache, and I/O state fresh in the child. Preserve provider-initialized runtimes without resetting their live package graph when they enter the managed lifecycle API. Cross-backend snapshot tests cover global alias isolation, skipped blessed objects, hook context, parent immutability, and fresh execution stacks. The full make gate passes; Config thread capability flags remain disabled. Generated with [Codex](https://openai.com/codex) Co-Authored-By: Codex <223018245+openai-codex[bot]@users.noreply.github.com>
fglock
changed the title
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
CLONE_SKIP, child-sideCLONE, and fresh child execution/resource stateConfigthread flags disabled; this PR does not expose the PerlthreadsAPIValidation
make— pass after each significant phase; final gate passCLONE_SKIP,CLONE, parent immutability, and fresh execution stacks pass on JVM and interpreter backendsPerformance
Three fresh-JVM runs per tree, comparing median rates against the exact merged base:
Design notes
CloneablePerlSubroutinedev/design/concurrency.md