🐛🧑‍💻 - Bugbot cannot fully review legitimate long SDD clarifications in a small consumer PR

Is there an existing issue for this?

I have searched the existing issues.

Searches for human-conversation and partial review found #724 (automated Codecov reports misclassified as human) and #831/#832 (finding grounding). This report concerns genuine long SDD clarifications, not bot reports or a false-positive finding.

Which actions are affected?

Pull Request, Commits

Which platforms are affected?

Linux

Description

Bugbot cannot complete review of a normal 18-file consumer increment because the human-conversation source omits five of eight legitimate specification clarifications. All code analysis partitions complete, but the review remains partial and cannot establish the consumer's required full-context acceptance.

Consumer: Domus PR #15, head 695e5c31a466b8a42c8840c46f708141979b35e4. This first entry-model increment preserves all eight requirements and 48 published companion cases. No full UI/owner adoption is claimed.

The fail-closed partial result is correct given the omitted context; this report does not request a green status on incomplete evidence. The workflow limitation is that ordinary, authoritative SDD answers accepted during pre-branch planning cannot all participate in the later review. Please investigate bounded complete handling (for example, independently partitioning genuine conversation evidence while preserving source provenance), rather than dropping requirements, deleting review history, truncating source or merely increasing an unchecked cap.

Expected: either review all required human source under the existing safe per-request bounds or provide a concrete source-preserving path to a complete review. Retain partial status until that evidence is complete.

Reproducing the issue

Consumer issue #8 contains eight substantive human-authored SDD clarification comments, all from efraespada. They concern authoritative APIs, session/journal recovery, delivery, quality gates and accepted English/Spanish behavior.
Their original comment IDs/character lengths are:
- 6547492565: 4,355
- 6547600490: 2,250
- 6547969134: 5,585
- 6548654723: 4,128
- 6549534976: 6,512
- 6549593567: 6,001
- 6549996864: 1,584
- 6550215922: 13,278
The Action subsequently publishes and verifies the companion on the managed branch; immutable SDD publication commit is 82519ee4e547dc347b8d0a18352bd1acc7783d70, develop base a63b2315cd5121dde9585ba8875e286fdef5330e.
Analyze the first implementation commit above in review run 41295062113 / job 124814060445.
The original telemetry reports 18 changed files, two analysis partitions, both complete, zero candidate findings, but overall context coverage partial:
human-conversation: status=partial, pagesFetched=1, itemsFetched=8, itemsRetained=3, omittedItems=5, truncatedItems=0, limitReached=true.
Copilot / Review check 124814608482 remains neutral/partial and lifecycle blocked. This is not a complete zero-findings review. Telemetry does not identify the five omitted IDs; their precise selection and the applicable underlying limit are not inferred here.
Separately, all nine project execution groups and aggregate CI Check 124814684215 / run 41295062246 pass, including both container architectures and the new real-Chromium application gate. Forty model tests and all four complete-runtime coverage metrics are 100%. That passing execution is not used to waive missing review context.

copilot Version

vypdev/copilot@v3 resolved in the hosted job to 3e7ceea08e6131360e026937d1c0cc1b120472e0; codex-cli 0.166.0, reviewer gpt-6-luna, smart effort.

Relevant Log Output

analysisPartitions=2
completedAnalysisPartitions=2
analysisAssignedFiles=18
contextCoverageStatus=partial
human-conversation.itemsFetched=8
human-conversation.itemsRetained=3
human-conversation.omittedItems=5
human-conversation.truncatedItems=0
human-conversation.limitReached=true
candidateFindings=0
publishedFindings=0
outcome=partial

Additional context and comments

The original thread, owning SDDs and acceptance are retained; the consumer continues implementation while reporting this limitation. No Copilot source, installed/generated resource, configuration, model, review bound, acceptance gate or gateway was modified. No repeated unchanged recheck was attempted to force this partial review clean.

The separate Commit progress failure on this head is already reported in #828 (scope-clause-limit, reference 56263937-b3b4-45d5-b50c-e7c56ecb838b); it must not be conflated with this source-coverage limit.

@vypbot

Hi! I’m @vypbot, the Copilot assistant for this repository.

I can answer questions, explain the codebase, propose implementation and test plans, review issues and pull requests for potential bugs or security problems, and help authorized maintainers apply changes.

Try /copilot help to see the available commands, or mention @vypbot with your question.

Additional changed-source observation on Domus PR #15, head e51556f1ec571b1d25111051a0bb22aa0947995e (2026-11-01). This is the normal strict HTTP-adapter increment after ca55cdb, not an unchanged workflow retry.

Pull-request run 41308556484, job 124853938719, completes successfully as an Action job, while Copilot / Review check 124854415280 is neutral with the explicit title Copilot review has partial coverage. No full clean review is claimed.

Actual telemetry reports reviewId=vypdev:domus:pr-15:e51556f1ec57, contextCoverageStatus=partial, outcome=partial. All 8/8 planned code partitions complete (45 assigned files, 66 fragments); candidate/published/overflow findings are zero. Finding states: open=0, reopened=0, fixed=1, verification-required=0, unknown=0. Historical Unicode finding was genuinely fixed with regressions; it was not manually dismissed.

The sole partial component remains human-conversation: pagesFetched=1, itemsFetched=9, itemsRetained=4, omittedItems=5, truncatedItems=0, limitReached=true. Selection, issue comments (19/19), PR comments (2/2), review threads (1/1), full diff (45/45), previous-findings and rules are complete. Thus full code partition completion still cannot establish that the omitted human context was reviewed.

Source tests independently pass: 379 actual Chromium cases, every statement/branch/function/line across all seven frontend runtime files at 100%, 37 Node policy/graph cases and two actual native/standalone provider conformance cases. Current-head hosted frontend also passes. These results do not replace missing review context. The full job log and exact telemetry are retained with the consumer checkpoint; product UI and integrated browser acceptance are still incomplete.

Please investigate upstream completeness/retrieval handling under the unchanged bounded contract. We preserved conversation history, credentials/configuration/resources/limits and did not retry the unchanged run or suppress partial status. The separately repeated progress-scope failure at original source L97 is recorded under #828; the unrelated Unicode C1 prose false positive is #845. Consumer implementation continues without treating partial review as a merge approval.

Additional changed-source observation: Domus PR #15, exact head 709994192eef0a3b2eaf0a8ac638cc8d9cc28ec6, parent e51556f1ec571b1d25111051a0bb22aa0947995e, develop base a97294385798b31e98c200e7ed93e355ffc7c9a7 (2026-11-01). This normal contributor increment adds browser capabilities/lifecycle composition and the owned Lit EN/ES view/catalog; it is not an unchanged workflow retry.

Actual review run 41312573975, job 124865853683, succeeds as an Action job, while Copilot / Review check 124866582988 remains neutral with title Copilot review has partial coverage. Native review history has one genuinely fixed/outdated historical finding and zero unresolved threads; no full-context clean acceptance is inferred.

Telemetry reviewId=vypdev:domus:pr-15:709994192eef reports all 10/10 code partitions complete, 60 assigned/retained diff files and 85 fragments. New candidate/published/overflow findings are zero. States: open=0, reopened=0, fixed=1, verification-required=0, unknown=0. Overall contextCoverageStatus=partial and outcome=partial.

The sole partial component is still human-conversation: pagesFetched=1, itemsFetched=9, itemsRetained=4, omittedItems=5, truncatedItems=0, limitReached=true. Selection (1/1), issue comments (19/19), PR comments (2/2), review threads (1/1), full diff (60/60), previous findings and rules report complete. Thus completing every code partition still does not establish review of the five omitted human-context items. The log does not identify their exact rejected contents or the internal reason for this bound; those are not guessed.

CI Check run 41312574000, aggregate 124866589069, is green including both CPU containers. Actual frontend job 124865853613 passes all490 Chromium cases, every authored runtime file/four-metric inventory at100%, and18 built-catalog Playwright cases with actual Chromium156.0.8078.4. Local full148 tooling and38 Node cases pass; product bootstrap/static delivery and browser-to-backend/owner acceptance remain explicitly incomplete.

Unlike the preceding e51556f source, this head's Copilot / Verification 124866489935 and Copilot - Commit 124865836156 succeed. This observation therefore records the partial-review limitation only; it is not another instance of the separate #828 scope failure. The previous failures and their original diagnostics remain preserved.

Complete original review/frontend logs, telemetry, source/SDK identities, actual failure regressions, source coverage, runnable fictional catalog and screenshots are retained in the consumer checkpoint. No Copilot source, configuration, workflow, credentials, limits or human history changed; no partial result was suppressed and no unchanged retry was requested. Please investigate upstream retrieval/completeness while preserving bounded contracts and truthful partial status. Consumer implementation continues without treating this as merge approval.

Additional verified recurrence on Domus PR #15 at source 04fc72da07d5510c269179538f0b7a3ac4b983c2. This is the same conversation-coverage problem, not a new issue and not a dispute of the valid product finding.

Action: vypdev/copilot@v3, exact resolved source 3e7ceea.
Original execution: https://github.com/vypdev/domus/actions/runs/41319945673/job/124887798695
Review check: https://github.com/vypdev/domus/runs/124888525393 (neutral).
Review identity: vypdev:domus:pr-15:04fc72da07d5; started 2026-11-01T15:11:45.545Z.
All 12/12 code partitions completed; 81 assigned files, 109 fragments. Diff coverage retained all 81 files with no omission/truncation. One candidate was published, with open=1 and fixed=1.
Human-conversation coverage is still partial: one page, 9 fetched, 4 retained, 5 omitted, zero truncated, limitReached=true. Every other reported context coverage component is complete. Overall contextCoverageStatus=partial and outcome=partial.
Telemetry estimated 290,712 input tokens from 1,162,848 prompt characters across the bounded execution; elapsed 149,519ms. This records actual emitted measurements, not a request to raise limits or change runtime configuration.

The new finding about logout erasing an in-flight command journal is valid. We reproduced it in eight real Chromium controller/runtime regressions and two real backend/SQLite journeys, and are fixing it in Domus. Its validity is independent of the incomplete conversation fetch. We do not interpret zero remaining candidates, a completed code scan or a neutral check as complete clean acceptance.

Expected: consume the complete supported current conversation through bounded continuation, or report a specific actionable coverage boundary without conflating a complete diff scan with full review. Please preserve findings and review history while addressing retrieval. No consumer-side Copilot source, workflow, credentials, context limit or branch lifecycle has been modified; implementation and test work continues.

Verified follow-up at 9f5cabe9cb4ffa5156f0fa01d8a3699a04f498af:

All required CI suites and aggregate124891940509 passed: https://github.com/vypdev/domus/actions/runs/41321048352/job/124891940509.
Original review execution: https://github.com/vypdev/domus/actions/runs/41321048391/job/124891144597; review check124892004233 remains neutral.
Native review history now marks both findings fixed: open=0,reopened=0,fixed=2,verification-required=0,unknown=0. The in-flight logout correction was accepted; no review thread was manually resolved.
All12/12 analysis partitions again completed for81files/109fragments, with zero new findings published. The emitted human-conversation coverage still has9fetched/4retained/5omitted/0truncated and limitReached=true; overall coverage/outcome remains partial.
Thus a corrected product finding and a fully successful project CI do not repair the independent incomplete conversation fetch. This update preserves the earlier evidence and records the same problem at the verified corrected source. Implementation continues with exact image packaging, without modifying Copilot or changing its limits/configuration.

Final entry checkpoint at consumer 3da78a77d5d9b4c45422af10b3a7da12b404d189: review job124928564910 completes14/14analysis partitions over94files/11751changed lines. Candidate/published findings0; native states open0/reopened0/fixed6/obsolete1/verification-required0/unknown0. All7native threads are resolved. Human conversation remains partial:14fetched/7retained/7omitted/0truncated/limitReachedtrue; diff and other named sources are complete. The check124929410871 is neutral, not a complete-context passing review.

The independent CI Check124929735579 succeeds, including backend2987cases/107modules100%, frontend577cases/15modules100% in all four metrics,45composed product journeys,18catalog journeys and real Edge.5 image tests on both CPU families. The owner explicitly allows reported Copilot infrastructure/context failures alongside merge once full product tests/coverage and no unresolved findings are verified. We are proceeding under that explicit exception, not claiming complete conversation coverage or changing any Copilot limit/resource/credential. The full product goal remains active; this checkpoint covers entry/setup/join only.

New exact-head occurrence with independently verified truncation as well as omission: Domus PR#19, head 930c2fc1f68e715f2c4ed39a2511fab1ae7fb235, review run41344718433/job124961974826, Review check124963167821.

The Action source is 3e7ceea08e6131360e026937d1c0cc1b120472e0, codex-cli0.166.0, gpt-6-luna/smart. Original completed telemetry and the native status card agree:

changedFiles=15
analysisPartitions=2
completedAnalysisPartitions=2
contextCoverageStatus=partial
issue-comments: complete, itemsFetched=12, itemsRetained=12
human-conversation: partial, itemsFetched=6, itemsRetained=5,
  omittedItems=1, truncatedItems=1, limitReached=true
candidateFindings=0
publishedFindings=0
outcome=partial

The native card additionally identifies issue:6562490003 as truncated. The full original human clarification was fetched independently and currently contains38305 Unicode characters. It supplies the reviewed recurrence companion proposal and concrete scope, not an automated coverage report. The specific omitted item's identity and the exact retained/truncated fragment are not exposed in the inspected telemetry and are not inferred. All12 issue comments being fetched completely does not imply the derived human-conversation source was retained completely.

Project CI Check and all applicable groups succeed on this head. The100% local backend result does not turn partial review into complete review, and the two completed code partitions do not establish complete conversation coverage. The PR is ready for early review but remains an incomplete implementation; no merge is attempted.

Please investigate a bounded source-preserving complete-context path, including both omission and truncation of legitimate admitted SDD clarification, while keeping the partial status until evidence is complete. No original comment, approved requirement, review history, Copilot configuration/resource or bound has been removed or relaxed. The independent Commit scope-clause-limit occurrence/reference869f6f2e is separately recorded in#828; no common root cause is asserted.

Implementation plan

Current status: Ready to start. No action is required from maintainers before implementation.

Trace conversation collection and bounds
- Read the repository guidance and documentation, then locate human-conversation fetching, retention limits, provenance handling, and partial-coverage reporting.
- Identify which bound caused omission and how source items are selected, without assuming telemetry reveals omitted IDs.
Define bounded complete coverage
- Specify a source-preserving strategy to process all relevant human-authored clarifications within existing per-request safety bounds, such as independently partitioning evidence.
- Define how provenance, pagination, truncation, and incomplete partitions contribute to coverage status; preserve partial status whenever evidence is missing.
Implement conversation partitioning
- Partition eligible conversation evidence deterministically and retain source identifiers and attribution through each partition.
- Aggregate results only after all required partitions complete; report omitted or truncated evidence as partial.
Add focused regression coverage
- Cover the eight-comment case, per-request bounds, pagination, provenance retention, and incomplete or failed partitions.
- Verify that all-complete coverage can proceed while any missing evidence remains partial.
Document the coverage behavior
- Document partition and bound semantics, provenance expectations, and how partial outcomes are surfaced.

Acceptance: A regression fixture with eight eligible clarifications exceeding the former retention cap processes every item within the configured per-request bounds, preserves each source identifier, and reports complete coverage only after all partitions succeed; an omitted or failed partition reports partial coverage.

Need something else? Mention the bot with a question or use /copilot help.

Issue closed — historical progress

Current status: closed as completed.

The last estimate is retained below as history. Closure does not certify a new percentage, tests, or branch alignment.

Action required: none for this closed issue. If work continues, reopen the issue; a fresh assessment can update this same card.

Last assessment before closure (historical)

Progress: 75% — in progress

> Current status: [R1] Bugbot cannot complete review of a normal 18-file consumer increment because the human-conversation source omits five of eight legitimate specification clarifications. All code analysis partitions complete, but the review remains partial and cannot establish the consumer's required full-context acceptance. — The source adds bounded complete-discussion delivery to every source-bound partition, but the supplied synthetic fixtures do not establish a complete review of the consumer’s actual discussion.
[R2] The fail-closed partial result is correct given the omitted context; this report does not request a green status on incomplete evidence. The workflow limitation is that ordinary, authoritative SDD answers accepted during pre-branch planning cannot all participate in the later review. Please investigate bounded complete handling (for example, independently partitioning genuine conversation evidence while preserving source provenance), rather than dropping requirements, deleting review history, truncating source or merely increasing an unchecked cap. — The source implements bounded pages, preserves source identity and attribution, preflights request and invocation limits, and retains partial behavior for provider omissions and callers without the source capability.
[R3] Expected: either review all required human source under the existing safe per-request bounds or provide a concrete source-preserving path to a complete review. Retain partial st
[untrusted content truncated]

Next: [R1] Bugbot cannot complete review of a normal 18-file consumer increment because the human-conversation source omits five of eight legitimate specification clarifications. All code analysis partitions complete, but the review remains partial and cannot establish the consumer's required full-context acceptance. — Verify complete coverage for the required consumer discussion using its actual source and review receipts.
[R3] Expected: either review all required human source under the existing safe per-request bounds or provide a concrete source-preserving path to a complete review. Retain partial status until that evidence is complete. — Verify that the required discussion reaches every partition within the bounds and that incomplete evidence remains partial.

Issue closed: synchronization recommendation retired

No branch synchronization is required for this closed issue.

The previous recommendation is historical. Issue closure does not certify that the branches are synchronized.

Navigation Menu

Repository navigation

🐛🧑‍💻 - Bugbot cannot fully review legitimate long SDD clarifications in a small consumer PR #843

Is there an existing issue for this?

Which actions are affected?

Which platforms are affected?

Description

Reproducing the issue

copilot Version

Relevant Log Output

Additional context and comments

6 remaining items

Implementation plan

Issue closed — historical progress

Last assessment before closure (historical)

Progress: 75% — in progress

Issue closed: synchronization recommendation retired

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

🐛🧑‍💻 - Bugbot cannot fully review legitimate long SDD clarifications in a small consumer PR #843

Description

Is there an existing issue for this?

Which actions are affected?

Which platforms are affected?

Description

Reproducing the issue

copilot Version

Relevant Log Output

Additional context and comments

Activity

vypbot commented on Nov 1, 2026

efraespada commented on Nov 1, 2026

efraespada commented on Nov 1, 2026

efraespada commented on Nov 1, 2026

efraespada commented on Nov 1, 2026

6 remaining items

vypbot commented on Nov 3, 2026

Implementation plan

vypbot commented on Nov 3, 2026

Issue closed — historical progress

Last assessment before closure (historical)

Progress: 75% — in progress

vypbot commented on Nov 3, 2026

Issue closed: synchronization recommendation retired

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions