Skip to content

docs(upstream): record the 1.4.1 to 1.5.0 audit and batch 1 - #180

Merged
sunerpy merged 1 commit into
mainfrom
docs/upstream-ledger-v15-audit
Aug 1, 2026
Merged

sunerpy merged 1 commit into
mainfrom
docs/upstream-ledger-v15-audit

Conversation

@sunerpy

@sunerpy sunerpy commented Aug 1, 2026

Copy link
Copy Markdown
Owner

Summary

The upstream-sync ledger had gone out of sync with reality. PR #173 (5d795a5, released in v0.41.0) landed a large slice of the colby 1.4.1 → 1.5.0 range with a 10 070-line evidence ledger at docs/upstream-sync/V1_5_PORTABLE_FIXES.md, but never wrote a disposition into UPSTREAM.md. On paper the sync had not happened — which is precisely the failure this file exists to prevent, since the next sync would have re-investigated all 86 commits from scratch.

This records the audit.

What the audit found

All 86 commits bucketed (11 + 2 + 47 + 26 = 86, arithmetic stated so it is checkable):

  • 11 COVERED by a feat: sync portable fixes from upstream colby v1.5.0 #173 batch — C++ operator calls, template-arg stripping, namespace prefixing, C attribute-macro blanking, imported-singleton and literal-receiver resolution, row-id ref cleanup, and two verified-equivalent watch behaviors.
  • 2 portable and still open — Go field-chain validated type inference, and a NO_COLOR/--no-color policy. The second may be "behavior naturally satisfied, flag absent" rather than a defect: the CLI has no colouring dependency at all. Flagged to verify before implementing.
  • 47 N/A — dominated by upstream's migration to a napi-rs kernel with vendored tree-sitter grammars as C, which violates this project's no-vendored-grammar invariant and is unnecessary anyway: all 11 languages in that series already have native extractors here.
  • 26 DEFER — performance work measured against Node and node:sqlite (pool sizing, cgroup cache credit, macOS vm_stat budgets). Byte-identical to graph output, so not parity blockers; re-derive from a Rust profile rather than transliterate. One of them becomes a correctness item if we ever adopt timer-triggered WAL truncation.

The 6 unreleased commits above v1.5.0, each with a disposition and rationale — one already ported (this batch), three queued, one partial, one N/A because we ship a native .exe and never a .cmd shim.

colby's open issues — 4 actionable with the code evidence for each, plus the already-handled claims we verified rather than assumed. Worth noting: the exact-name ranking shipped in v0.42.0 does not close #1473, because it changed search ranking while #1473 lives in the lookup family's separate symbol-resolution path.

Tracked parity stays 1.4.1

By this ledger's own definition — "all portable in-scope behavior landed" — parity advances only when every portable item has landed. Six remain. Advancing it now would make the file lie in the other direction.

Verification

  • make fmt idempotent; docs-only change, single file
  • The parity line is unchanged and asserted by gate F4

PR #173 landed a large slice of the 1.5.0 range with a ten-thousand-line evidence
ledger but never wrote a disposition here, so on paper the sync had not happened.
That is the failure this file exists to prevent: the next sync would have
re-investigated all eighty-six commits from scratch.

Bucket every one of them — eleven covered by a #173 batch, two portable and still
open, forty-seven inapplicable, twenty-six deferred — and say why for each group.
The inapplicable set is dominated by upstream's move to a napi-rs kernel with
vendored grammars, which this port cannot follow and does not need to: all eleven
languages in that series already have native extractors. The deferred set is
performance work measured against Node and node:sqlite, so it should be re-derived
from a Rust profile rather than transliterated.

Also record the six unreleased commits above the tag, the four actionable items
from colby's open issues with the code evidence for each, and which of their
already-handled claims we verified rather than assumed.

Tracked parity stays 1.4.1. By this ledger's own definition it advances only when
every portable in-scope behavior has landed, and six remain. Recording that
honestly is the point of the file.
@codecov

codecov Bot commented Aug 1, 2026

Copy link
Copy Markdown

Codecov Report

✅ All modified and coverable lines are covered by tests.

Impacted file tree graph

@@            Coverage Diff             @@
##             main     #180      +/-   ##
==========================================
- Coverage   95.08%   95.08%   -0.01%     
==========================================
  Files         136      136              
  Lines       70529    70529              
==========================================
- Hits        67063    67061       -2     
- Misses       3466     3468       +2     

see 1 file with indirect coverage changes

🚀 New features to boost your workflow:
  • ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
  • 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

@sunerpy
sunerpy merged commit a19e5f0 into main Aug 1, 2026
7 checks passed
@sunerpy
sunerpy deleted the docs/upstream-ledger-v15-audit branch August 1, 2026 17:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant