Bind query evidence and Python imports to proven index state #264

Merged

buildagent merged 3 commits from fix/review-followthrough-20260911 into master 2026-09-11 15:57:11 +02:00

Conversation 0 Commits 3 Files changed 68 +2834 -536

buildagent commented 2026-09-11 11:17:22 +02:00

Member

Copy link

Queries could combine different SQLite snapshots while reporting only Git provenance, and negative answers did not disclose working-tree changes beyond indexed content. This change adds transactional content identities, evidence from the snapshots actually read, RPC propagation, content-bound cursors, compound-query consistency checks, and bounded freshness observations.

Python imports resolve by proven module and export origin, including aliases, relative imports and conservative re-exports. Ambiguous or shadowed origins remain unresolved. Schema 67 repairs existing Python bindings; actual Flask and Django v65 databases upgrade to the same bindings as fresh indexes. The untyped storage.request.COOKIES chain in #250 remains outside proven type flow.

Executable build identities are unified, archive member versions and installer staging are verified, and routing timing assertions use deterministic work counts. Windows smoke stages the four release binaries in a fresh directory; unexpected executables within an archive are still rejected.

Corpus effects

All nine pinned repositories were compared against clean master 017c6d8 using occurrence-aware joins and source adjudication. Reference populations are unchanged; all seven non-Python repositories are bind-for-bind unchanged.

Corpus	Correct gains	False bindings removed	Correct losses	Conditionally correct losses	Corrected targets
Django	1715	1126	74	76	2
Flask	13	10	2	0	0

Correct losses include composed exports, extra import roots and object/ancestry flows not modeled here. Conditional losses involve runtime proxies, GIS configuration and task decorators. These accepted coverage losses are recorded alongside the gains. Historical #247 residuals remain tracked in #263; no rerun of that historical census is claimed.

The approved structural, resolver-stage and tier-3 records were generated by the existing measurement writers, with generation identities, categorized diffs and positive controls. Flask resolves one additional reference overall; Django resolves 439 more. The structural exception and complete target-movement review are documented in the records.

Measurement records

Corpus response budgets record snapshot evidence costs of +816/+912/+840 tokens for C#/Flask/Rust and freshness costs of +71/+0/+70. Attribution preserves the previous history and separates inherited record drift from these exact field costs. Questions, calls, answer quality, fallback counts and competitor populations are unchanged. The new recorded totals are 10645/10311/9976; fresh verification measured 10651/10304/9970.

Ruby package cost was measured in isolation under schema 67 / package 0.6.0: vm_step 26439909, fullscan_step 584758, sort 297, autoindex 20745, wall ratio 135%. These remain inside the previous bands. A subsequent verification also passed. All tolerance multipliers, the 60-token response drift allowance, reserve policy and existing plugin benchmark records are unchanged.

Validation

• On 296cc58, native Windows CI passed workspace tests, all eight staged shipped-set smoke checks and all six release-worker smoke checks. Linux CI passed fmt, clippy, MSRV, windows-gnu, ABI32/wasm, deny, private-item docs, complete workspace and full daemon transport.
• On 022d665, release structural/stage/tier-3/Ruby gates passed: 26 tests. Real-corpus benchmarks and attribution checks passed: 7 tests. Formatting and diff checks passed. Full PR CI passed on 022d665: native Windows (including staged shipped-set smoke 8/8 and release-worker smoke 6/6), complete Linux workspace, full daemon transport, corpus, formatting, lint, compatibility, dependencies and documentation. Corpus CI measured 10654/10312/9970 response tokens, inside the unchanged tolerances.
• Complete earlier local workspace: 3762 passed, 43 ignored, 347 test groups. Full daemon-backed MCP suite passed.
• Release precision gate passed 7/7 against its declared synthetic decoys; it covers zero corpus repositories.
• 32 mutation checks killed their intended tests; source restoration was verified by checksum.
• Linux release archive smoke passed 8/8; installer source/checksum audit passed.
• Cold-index cost passed without changing its baseline after statement-level attribution and SQL optimization.
• Tier-3 scale passed on Django and rust-analyzer (890445 references, six positive controls); both are below the separate one-million-edge daemon cap.
• Both plugin payload benchmarks passed: 12472/13058 tokens, all 20 expected items correct across 19 questions; the +555-token change is attributed to snapshot/freshness evidence and cursor hints.
• Package baseline changed only for schema conditions, with zero projection changes.

Refs #245, #246, #250, #251, #253, #260, #261, #262.

Queries could combine different SQLite snapshots while reporting only Git provenance, and negative answers did not disclose working-tree changes beyond indexed content. This change adds transactional content identities, evidence from the snapshots actually read, RPC propagation, content-bound cursors, compound-query consistency checks, and bounded freshness observations. Python imports resolve by proven module and export origin, including aliases, relative imports and conservative re-exports. Ambiguous or shadowed origins remain unresolved. Schema 67 repairs existing Python bindings; actual Flask and Django v65 databases upgrade to the same bindings as fresh indexes. The untyped storage.request.COOKIES chain in #250 remains outside proven type flow. Executable build identities are unified, archive member versions and installer staging are verified, and routing timing assertions use deterministic work counts. Windows smoke stages the four release binaries in a fresh directory; unexpected executables within an archive are still rejected. ### Corpus effects All nine pinned repositories were compared against clean master 017c6d8 using occurrence-aware joins and source adjudication. Reference populations are unchanged; all seven non-Python repositories are bind-for-bind unchanged. | Corpus | Correct gains | False bindings removed | Correct losses | Conditionally correct losses | Corrected targets | |---|---:|---:|---:|---:|---:| | Django | 1715 | 1126 | 74 | 76 | 2 | | Flask | 13 | 10 | 2 | 0 | 0 | Correct losses include composed exports, extra import roots and object/ancestry flows not modeled here. Conditional losses involve runtime proxies, GIS configuration and task decorators. These accepted coverage losses are recorded alongside the gains. Historical #247 residuals remain tracked in #263; no rerun of that historical census is claimed. The approved structural, resolver-stage and tier-3 records were generated by the existing measurement writers, with generation identities, categorized diffs and positive controls. Flask resolves one additional reference overall; Django resolves 439 more. The structural exception and complete target-movement review are documented in the records. ### Measurement records Corpus response budgets record snapshot evidence costs of +816/+912/+840 tokens for C#/Flask/Rust and freshness costs of +71/+0/+70. Attribution preserves the previous history and separates inherited record drift from these exact field costs. Questions, calls, answer quality, fallback counts and competitor populations are unchanged. The new recorded totals are 10645/10311/9976; fresh verification measured 10651/10304/9970. Ruby package cost was measured in isolation under schema 67 / package 0.6.0: vm_step 26439909, fullscan_step 584758, sort 297, autoindex 20745, wall ratio 135%. These remain inside the previous bands. A subsequent verification also passed. All tolerance multipliers, the 60-token response drift allowance, reserve policy and existing plugin benchmark records are unchanged. ### Validation - On 296cc58, native Windows CI passed workspace tests, all eight staged shipped-set smoke checks and all six release-worker smoke checks. Linux CI passed fmt, clippy, MSRV, windows-gnu, ABI32/wasm, deny, private-item docs, complete workspace and full daemon transport. - On 022d665, release structural/stage/tier-3/Ruby gates passed: 26 tests. Real-corpus benchmarks and attribution checks passed: 7 tests. Formatting and diff checks passed. Full PR CI passed on 022d665: native Windows (including staged shipped-set smoke 8/8 and release-worker smoke 6/6), complete Linux workspace, full daemon transport, corpus, formatting, lint, compatibility, dependencies and documentation. Corpus CI measured 10654/10312/9970 response tokens, inside the unchanged tolerances. - Complete earlier local workspace: 3762 passed, 43 ignored, 347 test groups. Full daemon-backed MCP suite passed. - Release precision gate passed 7/7 against its declared synthetic decoys; it covers zero corpus repositories. - 32 mutation checks killed their intended tests; source restoration was verified by checksum. - Linux release archive smoke passed 8/8; installer source/checksum audit passed. - Cold-index cost passed without changing its baseline after statement-level attribution and SQL optimization. - Tier-3 scale passed on Django and rust-analyzer (890445 references, six positive controls); both are below the separate one-million-edge daemon cap. - Both plugin payload benchmarks passed: 12472/13058 tokens, all 20 expected items correct across 19 questions; the +555-token change is attributed to snapshot/freshness evidence and cursor hints. - Package baseline changed only for schema conditions, with zero projection changes. Refs #245, #246, #250, #251, #253, #260, #261, #262.

buildagent added 1 commit 2026-09-11 11:17:22 +02:00

fix: bind query evidence and Python imports to proven index state ... a2a2b3f8bd

Add transactional content identities and query-scoped SQLite snapshot evidence,
propagate it over RPC, bind cursors to content, and disclose bounded freshness
for negative answers. Resolve Python imports by exact module/export origin and
repair existing databases on upgrade.

Unify executable build identities, verify release archive versions and installer
staging, and replace routing timing assertions with deterministic work counts.

Local workspace, daemon MCP, docs, precision, cost and release smoke checks pass.
The protected corpus baseline remains unchanged: its Flask structural gate is
red and requires an explicit reviewed exception before merge.

Refs: #245, #246, #250, #251, #253, #260, #261, #262

buildagent referenced this pull request from a commit 2026-09-11 12:07:49 +02:00

ci: stage shipped Windows binaries before version smoke

buildagent added 1 commit 2026-09-11 12:07:49 +02:00

ci: stage shipped Windows binaries before version smoke ... 296cc583a4

The build directory also contains helper executables left by tests. Copy the
four release binaries into a fresh directory before the shipped-set smoke, so
version discovery grades the archive population. Keep executable enumeration
strict so an unexpected executable in a staged archive still fails.

Validated the workflow scope guard and a real-process Linux reproduction:
dirty build directory rejected, clean staged set 8/8, contaminated archive
rejected. Native Windows CI will validate the PowerShell staging path.

Refs: #262, #264

buildagent added 1 commit 2026-09-11 15:06:28 +02:00

Record reviewed corpus and query-evidence measurements ... 022d6653d0

Apply the approved five-record proposal with source-adjudicated Python coverage tradeoffs, measured snapshot and freshness token costs, and schema-67 Ruby cost conditions. Preserve tolerance multipliers, corpus populations, plugin benchmarks, and prior attribution history.

buildagent merged commit 0ee980d152 into master 2026-09-11 15:57:11 +02:00

buildagent referenced this pull request from a commit 2026-09-11 15:57:13 +02:00

Merge PR #264: Bind query evidence and Python imports to proven index state

buildagent referenced this pull request 2026-09-11 18:04:02 +02:00

Prepare v0.28.1 maintenance release #265

buildagent referenced this pull request 2026-09-11 20:24:42 +02:00

Two calls seconds apart served DIFFERENT index generations while answer_provenance was byte-identical: nothing says which generation answered #245

buildagent referenced this pull request 2026-09-11 20:24:42 +02:00

claim_domain_scale moved ONE wall-clock ratio to #[ignore] and left its sibling live — the sibling is a FLOOR, which contention breaks in the direction a floor cannot survive #253

buildagent referenced this pull request 2026-09-11 20:24:42 +02:00

answer_provenance names a commit the INDEX has not reached, so a watcher-lag miss and a measured absence are indistinguishable #260

buildagent referenced this pull request 2026-09-11 20:24:42 +02:00

Publish install.sh as a release asset, so an installer can be pinned to a release instead of tracking master #261

buildagent referenced this pull request 2026-09-11 20:24:42 +02:00

code-index-plugin-host cannot say which build it is — three of four shipped binaries answer --version, the fourth exits 21 with usage #262

buildagent referenced this pull request 2026-09-11 20:24:59 +02:00

#175 residual: 13 phantoms where a PROJECT module shadows a STDLIB module name (django/utils/{inspect,html}.py) #251

Sign in to join this conversation.

Reviewers

No reviewers

Labels

Clear labels   

code-review

  

correctness

  

dos

  

performance

  

security

  

severity/high

  

severity/low

  

severity/medium

  

tech-debt

  

Kind/Breaking

Breaking change that won't be backward compatible

  

Kind/Bug

Something is not working

  

Kind/Documentation

Documentation changes

  

Kind/Enhancement

Improve existing functionality

  

Kind/Feature

New functionality

  

Kind/Security

This is security issue

  

Kind/Testing

Issue or pull request related to testing

  

Priority

Critical

The priority is critical

  

Priority

High

The priority is high

  

Priority

Low

The priority is low

  

Priority

Medium

The priority is medium

  

Reviewed

Confirmed

Issue has been confirmed

  

Reviewed

Duplicate

This issue or pull request already exists

  

Reviewed

Invalid

Invalid issue

  

Reviewed

Won't Fix

This issue won't be fixed

  

Status

Abandoned

Somebody has started to work on this but abandoned work

  

Status

Blocked

Something is blocking this issue or pull request

  

Status

Need More Info

Feedback is required to reproduce issue or to continue work

No labels

code-review

correctness

dos

performance

security

severity/high

severity/low

severity/medium

tech-debt

Kind/Breaking

Kind/Bug

Kind/Documentation

Kind/Enhancement

Kind/Feature

Kind/Security

Kind/Testing

Priority

Critical

Priority

High

Priority

Low

Priority

Medium

Reviewed

Confirmed

Reviewed

Duplicate

Reviewed

Invalid

Reviewed

Won't Fix

Status

Abandoned

Status

Blocked

Status

Need More Info

Milestone

Clear milestone

No items

No milestone

Projects

Clear projects

No items

No project

Assignees

Clear assignees

No assignees

1 participant

Notifications

Subscribe

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

h-dv/code-index!264

No description provided.