Synchronize current ttsc-graph semantics by samchon · Pull Request #97 · samchon/graph

samchon · 2026-07-19T04:48:17Z

Summary

Synchronize the standalone graph with the current canonical ttsc-graph snapshot schema and operation semantics (pinned at samchon/ttsc@2b724664e), and land the review-round correctness fixes discovered during that sync (issues #98–#115) as one coherent, per-issue commit stack.

make compiler-owned literals, enum members, object members, source digests, and resident generations authoritative instead of reconstructing them from signatures or mutable disk state
rank the full complete-set handle candidate list before applying limits; report trace truncation only when eligible results are omitted
port the matching benchmark viewer reducer, public structures, result audit, and pinned parity fixture

Per-issue commit stack (newest first)

Commit	Issue	Status
Isolate benchmark language tools and skip Flutter-only Dart packages	benchmark infra	Refs #96 / #60
Make closed-pipe lifecycle coverage deterministic across operating systems	#113	Closes
Settle every generic LSP request and own its retired child on failure	#100, #107	Closes #100; advances #107
Sanitize mixed canonical file identities without node-order dependence	#115	Closes
Refuse the published schema-3 ttscgraph snapshot as an unproved manifest	#108	see note
Reconnect cross-file Swift extension members to their owner	#112	Closes
Preserve the released overrideEdges compatibility export	#114	Closes
(7 earlier parity commits)	#96	umbrella

Closes #96, #100, #112, #113, #114, #115.

Partial / divergent (left open on purpose)

[Resident] Close unpublished sessions and retired native children #107 — gap 2 (retired native/LSP child exit-promise ownership + resident refresh cancellation) is delivered; gap 1 (buildLspGraph({keepAlive:true}) accumulate-then-close-on-throw cleanup) is a follow-up.
[Identity] Keep cross-project file identities stable and collision-free #104 — the graph-side node-order independence and path vocabulary are delivered (that is [Paths] Sanitize mixed canonical identities without node-order dependence #115); the full collision-free bijection (same-basename absolute siblings, UNC/POSIX roots, two checkout locations) and its upstream ttsc companion remain.
[Provider] Gate checker-owned member relations by ttsc schema #108 — resolved by a stronger decision than the written AC: schema 3 is refused wholesale as an unproved manifest (fall back honestly to LSP), rather than gated-and-retained. Left open for maintainer confirmation of the direction — see issue comment.
[Provider] Preserve canonical fileless compiler diagnostics #110 — ISamchonGraphDiagnostic.file now documents the "" global-finding identity (contract-accurate), but the adapter does not yet preserve fileless diagnostics; the adapter change is a follow-up.

Verification

pnpm --filter @samchon/graph build (no errors; pre-existing lint warnings only)
pnpm --filter @samchon/graph-test build
focused local suite (fake-server + static/pure): override_edges, swift ×2, dump_adapter_rejects, schema3_falls_back, viewer_reduce ×2, refresh_language_session, lsp_client_closes, resident_close ×3, provider_surfaces_process — all Success
required GitHub checks (100% coverage on Ubuntu/macOS/Windows + the 16-language experiment matrix) — the [CI] Make closed-pipe lifecycle coverage deterministic across operating systems #113 fix targets the prior POSIX coverage regression
zero-spend benchmark integrity suite; paid benchmark cells remain outside this PR

🤖 Generated with Claude Code

The canonical ttsc graph advanced across fact ownership, resident recovery, handle ranking, trace completeness, viewer paths, and snapshot-owned detail facts. Reserve one dependency batch so semantic identity work does not preserve superseded reader behavior. Constraint: ttsc origin/master 2b724664e is the user-designated structural reference. Rejected: Fold the synchronization into #65 | it would mix canonical parity with new multi-language identity policy. Confidence: high Scope-risk: broad Directive: Port reference ownership and algorithms before reconciling PR #95. Tested: Branch starts exactly at samchon/graph origin/master 84f5b6f. Not-tested: Implementation and regression verification are pending.

Port the current ttsc-graph snapshot, request lifecycle, ranking, trace, reduction, and source-provenance semantics so the standalone graph cannot reinterpret compiler-owned facts or live disk state. Constraint: D:/github/samchon/ttsc ttsc-graph is the canonical contract and operation reference. Rejected: Continue reconstructing object members and literals from signatures or live source | diverges from the compiler snapshot and races edits. Confidence: high Scope-risk: broad Directive: Regenerate and verify the pinned ttsc contract before changing dump schema or operation semantics. Tested: graph build; test build; 19 focused regression tests; all 18 parity contracts at ttsc 2b724664e; go test ./cmd/ttscgraph/... Not-tested: Full paid agent benchmark requires a separate exact-scope authorization. Related: #96

Close the review gaps where live disk reads, synthesized generic relationships, incomplete native shutdown, and bounded traces could make a result claim more than its producing snapshot proved. Preserve explicit compatibility with the latest published schema while keeping schema 5 as the complete compiler-owned contract. Constraint: Published ttsc 0.19.3 still emits schema 3; complete schema 5 manifests require samchon/ttsc#781 until the next release. Rejected: Infer member relationships from generic source text | name equality cannot prove checker dispatch. Rejected: Re-read live files for missing snapshot text | it mixes different program generations. Confidence: high Scope-risk: moderate Reversibility: clean Directive: Keep schema 3 as an explicit incomplete compatibility lane and require producer-owned facts for schema 5 details. Tested: @samchon/graph build; @samchon/graph-test build; 16 focused feature tests including coverage edges; 18-contract parity preview; schema 5 self-index of graph and graph-sitter with zero warnings. Not-tested: Full local coverage and all-platform suite; GitHub Actions remains authoritative for those surfaces.

socket-security · 2026-07-19T07:41:48Z

Review the following changes in direct dependencies. Learn more about Socket for GitHub.

Diff	Package	Supply Chain Security	Vulnerability	Quality	Maintenance	License
	@ttsc/unplugin@0.19.3
	@ttsc/lint@0.19.3
	ttsc@0.19.3

View full report

socket-security · 2026-07-19T07:41:49Z

Warning

Review the following alerts detected in dependencies.

According to your organization's Security Policy, it is recommended to resolve "Warn" alerts. Learn more about Socket for GitHub.

Action	Severity	Alert (click "▶" to expand/collapse)
Warn		Obfuscated code: npm `@ttsc/darwin-arm64` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/darwin-arm64@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/darwin-arm64@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/darwin-arm64` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/darwin-arm64@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/darwin-arm64@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/darwin-x64` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/darwin-x64@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/darwin-x64@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/darwin-x64` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/darwin-x64@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/darwin-x64@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/linux-arm` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/linux-arm@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/linux-arm@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/linux-arm` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/linux-arm@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/linux-arm@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/linux-arm64` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/linux-arm64@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/linux-arm64@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/linux-arm64` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/linux-arm64@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/linux-arm64@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/linux-x64` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/linux-x64@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/linux-x64@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/linux-x64` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/linux-x64@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/linux-x64@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/win32-arm64` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/win32-arm64@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/win32-arm64@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.
Warn		Obfuscated code: npm `@ttsc/win32-arm64` is 90.0% likely obfuscated Confidence: 0.90 Location: Package overview From: pnpm-lock.yaml → `npm/ttsc@0.19.3` → `npm/@ttsc/win32-arm64@0.19.3` ℹ Read more on: This package \| This alert \| What is obfuscated code? Next steps: Take a moment to review the security alert above. Review the linked package source code to understand the potential risk. Ensure the package is not malicious before proceeding. If you're unsure how to proceed, reach out to your security team or ask the Socket team for help at `support@socket.dev`. Suggestion: Packages should not obfuscate their code. Consider not using packages with obfuscated code. Mark the package as acceptable risk. To ignore this alert only in this pull request, reply with the comment `@SocketSecurity ignore npm/@ttsc/win32-arm64@0.19.3`. You can also ignore all packages with `@SocketSecurity ignore-all`. To ignore an alert for all future pull requests, use Socket's Dashboard to change the triage state of this alert.

View full report

samchon · 2026-07-19T07:41:51Z

Second implementation/review pass completed in 145f76a.

The pass closes snapshot-generation mixing, generic fact synthesis, resident close/abort races, bounded-trace completeness, resolver recomputation, schema compatibility, path validation, and stale regression expectations. Local evidence:

@samchon/graph build and @samchon/graph-test build pass.
16 focused feature tests pass, including the coverage edge suite.
All 18 contracts reproduce the clean canonical samchon/ttsc@2b724664e reference.
A schema v5 binary with the manifest fix indexes graph-sitter (558/1,522) and graph (1,325/4,600) through the strict lane with zero warnings.

Strict schema v5 exposed an upstream provenance defect: the producer omitted outside .d.ts and bundled:/// sources from sourceDigests. The general fix is tracked in samchon/ttsc#780 and submitted as samchon/ttsc#781. This PR remains fail-closed against older malformed schema v5 snapshots and keeps published schema v3 as an explicit incomplete compatibility lane.

Snapshot-carrying build helpers are useful beyond their original wrappers, but their placement broke the one-export-per-source convention and several document tests accidentally depended on the old live-disk default. Give each helper its canonical module and make document-reading tests choose either the produced snapshot or an explicit live reader. Constraint: Detached dumps must remain fail-closed for source-derived facts Rejected: Restore live disk reads as SamchonGraphMemory.from default | would reintroduce time-of-check/time-of-use drift Confidence: high Scope-risk: narrow Directive: Keep build result helpers in matching source files and require explicit provenance at memory construction Tested: @samchon/graph build; @samchon/graph-test build; focused export and engine boundary regressions Not-tested: Full cross-platform suite remains delegated to CI

Public source reads now require exact consumed text or a checker digest, so an unproved live disk read cannot counterfeit an audited snapshot. First-build failures close every accumulated bulk and LSP session, while the native client retains ownership of retired child generations through forced exit. Constraint: strict compiler facts may outlive the disk bytes that produced them Rejected: Freeze the first unproved disk read | it proves only read-time consistency, not snapshot identity Confidence: high Scope-risk: moderate Directive: Every session or child created before publication remains owned until returned or observed exiting Tested: graph/test builds; source, lifecycle, native-stall, fake-provider, and coverage regressions; full suite except one independently passing warmup-timeout flake Not-tested: POSIX SIGKILL branch locally; Linux/macOS CI owns it Related: #103 #107 #96

A member cap now retracts whole-member audit language, omitted dispatch-hub candidates set truncation only when they remain in scope, and schema-v3 input cannot smuggle schema-v5 member relations. Remove the unused public name-matching override heuristic so checker-owned relations remain the sole semantic source. Constraint: ttscgraph schema 5 is canonical; schema 3 is compatibility-only Rejected: Infer member overrides by same-name containment | spelling and ownership do not prove checker identity Confidence: high Scope-risk: moderate Directive: Any bounded response must report its eligible omissions through audit or truncated Tested: focused audit, trace, adapter, contract-parity, source-export tests; full suite with one independently passing warmup-timeout flake Not-tested: published ttsc 0.19.4 integration pending its authorized release Related: #105 #106 #108 #96

samchon · 2026-07-19T09:32:39Z

Review round 3 follow-up is now pushed in 1ff7c61 and e4329b8.

Resolved in this update:

source-backed audits no longer expose an unproved live-disk reader;
first-build abort/failure closes every accumulated bulk/LSP session;
retired ttscgraph child generations remain owned through SIGTERM/SIGKILL and actual exit;
schema 3 rejects schema-5-only member implements / overrides facts;
dispatch-hub suppression reports eligible omissions through truncated;
details(memberLimit) no longer claims whole-member coverage;
the unused public same-name override heuristic was removed;
the fake strict provider now attests its bundled library source.

Canonical boundary decisions:

ttsc manifest completeness was fixed upstream in fix(graph): attest every resident source ttsc#781 and merged as 4c5e11e;
portable cross-project wire identity remains upstream-first in fix(graph): give cross-project sources portable collision-free wire identities ttsc#783 / [Identity] Keep cross-project file identities stable and collision-free #104; the speculative downstream reducer rewrite was deliberately excluded from this PR;
TypeScript docs now name the compiler-owned ttscgraph lane and the pending ttsc@0.19.4 release.

Verification:

graph and test builds pass;
focused source/audit/trace/schema/native/lifecycle/provider/contract tests pass;
the ordinary full suite passed every test except one warmup-timeout fixture under the loaded run; that exact fixture passed immediately in isolation. GitHub Actions is the authoritative clean-machine rerun.

Carry the canonical ttsc source manifest without reopening compiler-owned files, bind universe roots and diagnostics to their proven snapshot, and let shutdown cancellation reach native and generic LSP work before unpublished sessions can leak. Static Scala and Swift ownership fixes close the language-specific gaps found by the independent review, while reducer and source utility cleanup remove unreachable or duplicated behavior. Constraint: Canonical ttscgraph output is the authority for compiler-owned TypeScript facts and source identity. Rejected: Re-read strict manifest files during generic finalization | external and bundled identities are not generic source capabilities and a second disk read violates snapshot ownership. Confidence: high Scope-risk: broad Directive: Do not pass bulk-provider manifest identities into source-derived generic export analysis; strict slices already carry compiler-resolved exports. Tested: pnpm coverage (all tests; statements, branches, functions, and lines 100%); focused generic/native cancellation, resident refresh, source identity, Scala, Swift, and dump-adapter tests. Not-tested: npm publication of ttsc 0.19.4 and paid agent benchmark runs remain separately authorized release/cost gates.

samchon · 2026-07-19T10:36:45Z

Review Round 4 remediation is pushed as 2be0d31. Local proof is green: the complete regression suite passes under c8 with statements/branches/functions/lines all at 100%, focused cancellation/source-identity/Scala/Swift tests pass, benchmark integrity tests pass, and git diff --check is clean. The new head also prevents complete strict manifests (including external and bundled identities) from being reopened by generic export analysis. GitHub CI and an independent clean Review Round 5 are now in progress.

Restore the previously shipped `@deprecated` `overrideEdges(nodes, edges)` root export and re-export it through the indexer barrel, so downstream code importing it from `@samchon/graph` keeps compiling. The core compiler-owned indexing path does not call it. Closes #114 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018G2eNarjGmcWmEjFTW2Qcz

Collect Swift `extension` declarations project-wide and emit conformance edges so members declared in a separate file rejoin their owning type; extend inheritance-edge evidence and accept `enum` conformances. Ambiguous receivers stay unassigned. Closes #112 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018G2eNarjGmcWmEjFTW2Qcz

Narrow `SUPPORTED_DUMP_SCHEMA_VERSIONS` to [5] and delete the schema-3 edge/objectMember compatibility gates: a schema-3 dump is refused before it can seed the graph, and the client falls back honestly to LSP while still returning project declarations. Also adds a canonical `C:/...` manifest fixture so Linux exercises the Windows-drive identity branch (#113). Closes #108 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018G2eNarjGmcWmEjFTW2Qcz

Reduce absolute paths outside the root to a normalized identity even when no shared root exists, and reroot only when every project file is absolute, so the same input yields the same identities regardless of node order. The package reducer and benchmark viewer reducer stay byte-aligned; structure docs adopt the shared relative / normalized-absolute / `bundled:///` identity vocabulary with matching parity substitutions. Closes #115 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018G2eNarjGmcWmEjFTW2Qcz

Settle each pending request exactly once on stdin/write/exit/malformed failure, cache the failure so later requests reject before the wire, and thread an abort signal so a resident `close()` cancels an in-flight readiness/symbol/reference refresh. Await each retired child's exit and terminate its exact process tree (SIGTERM before SIGKILL; `taskkill /T /F` on Windows), sparing unrelated processes. `refreshLanguageSession` keeps its three-argument form. Closes #100 Closes #107 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018G2eNarjGmcWmEjFTW2Qcz

…stems Publish the stdin-closed fence only after descriptor 0 is actually closed, so the client's next write deterministically reaches the closed-pipe path instead of racing the fake's teardown. Reorder the platform-specific transport probe after the independent closed-session assertions so a pipe failure cannot hide their coverage. Closes #113 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018G2eNarjGmcWmEjFTW2Qcz

Discover per-language benchmark tools under root/`bin`/`server/bin`/nested `bin` instead of a single `bin`, and skip Flutter-only Dart packages (`pubspecRequiresFlutter`) and their unit test so the sweep provisions only what it can honestly build. Benchmark infrastructure for the #96 parity sweep. Refs #96 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018G2eNarjGmcWmEjFTW2Qcz

The stubborn-tree regression spawns the fake LSP with --ignore-termination to prove the client escalates SIGTERM before SIGKILL. terminateOwnedProcess destroys the client's stdin write end before it signals, so without a ref'd handle the fake drained its event loop at stdin EOF and exited before its SIGTERM handler ran, leaving no SIGTERM evidence on POSIX (macOS/Ubuntu) while Windows takes the taskkill lane. Hold the loop open with a ref'd keep-alive timer (the same pattern the close-input fixture already uses) and swallow stdin errors, so SIGTERM is observed deterministically and only SIGKILL ends the child. Refs #107 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018G2eNarjGmcWmEjFTW2Qcz

Three per-OS gaps kept the coverage gate red: LspClient's stdin stream error listener was reached only through the Windows synthetic-destroy path, its terminate() forced-exit fail is Windows-only, and TtscGraphClient's superseded-child guard was exercised on no host. Cover the stdin error listener cross-platform by destroying the write stream with an error, ignore the Windows-only terminate branch that POSIX short-circuits, and drop a stale chunk into consume() while the client owns no child to prove the guard directly. Refs #100 Refs #107 Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_018G2eNarjGmcWmEjFTW2Qcz

This was referenced Jul 19, 2026

[Parity] Synchronize the current ttsc-graph snapshot and operation semantics #96

Closed

[Handoff] Bulk-index provider campaign — full context & next steps for a fresh agent #93

Closed

Introduce overload-safe semantic identities #95

Merged

samchon added 2 commits July 19, 2026 14:39

This was referenced Jul 19, 2026

[Source] Remove unproved live bytes from audited graph snapshots #103

Closed

[Provider] Gate checker-owned member relations by ttsc schema #108

Closed

samchon added 2 commits July 19, 2026 18:31

samchon mentioned this pull request Jul 19, 2026

[Benchmark] Run and publish the full 13-language @samchon/graph parity sweep #60

Open

11 tasks

samchon and others added 7 commits July 19, 2026 21:38

samchon marked this pull request as ready for review July 19, 2026 12:45

samchon and others added 2 commits July 19, 2026 22:00

samchon merged commit 6accdf8 into master Jul 19, 2026
21 checks passed

samchon deleted the feat/ttsc-current-parity branch July 19, 2026 13:54

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Synchronize current ttsc-graph semantics#97

Synchronize current ttsc-graph semantics#97
samchon merged 16 commits into
masterfrom
feat/ttsc-current-parity

samchon commented Jul 19, 2026 •

edited

Loading

Uh oh!

socket-security Bot commented Jul 19, 2026 •

edited

Loading

Uh oh!

socket-security Bot commented Jul 19, 2026 •

edited

Loading

Uh oh!

samchon commented Jul 19, 2026

Uh oh!

samchon commented Jul 19, 2026

Uh oh!

samchon commented Jul 19, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

Conversation

samchon commented Jul 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Per-issue commit stack (newest first)

Partial / divergent (left open on purpose)

Verification

Uh oh!

socket-security Bot commented Jul 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

socket-security Bot commented Jul 19, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

samchon commented Jul 19, 2026

Uh oh!

samchon commented Jul 19, 2026

Uh oh!

samchon commented Jul 19, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

samchon commented Jul 19, 2026 •

edited

Loading

socket-security Bot commented Jul 19, 2026 •

edited

Loading

socket-security Bot commented Jul 19, 2026 •

edited

Loading