Repository navigation
test: restore cost fixture for large sessions - #35960
Conversation
First Principles Review (Opus 5.5) — ⏭️ skippedRevision |
UX Review (Opus 5.5) — ⏭️ skippedRevision |
Design Review (Opus 5.5) — ✅ PASSDesign-level review of Design-Verdict: PASS Nothing blocks: the change only adds tests, and its count-based ceilings plus strict-xfail targets give the bounded-restore work a fixed fixture to be measured against. Notes
[DESIGN-REVIEWED] 800bebf Description read[DESCRIPTION-READ] b46e3da6be47331283a66e2f158079024a71dbd4f1e7bdb07afec06347256d91 That sha256 names everything this verdict read: the PR description, plus the 1 evidence file(s) the workflow collected from it and handed this review. It covers that evidence because the media strip replaces every attachment URL with the same placeholder, so an edit swapping one attachment for another would leave the description's captured text identical while what this review judged changed. Recompute it with |
Opus 5.5 Review — ✅ no blocking findingsReviewed Verdict parsed from the review's SHA-scoped output markers for commit False positive or not applicable? A repository writer can comment: |
GPT 6.1 Review — ✅ no blocking findingsGPT 6.1 completed its review of This comment is updated in place on each push. Review detailsNo findings. False positive or not applicable? A repository writer can comment: |
bd9a889 to
800bebf
Compare
bolichen97
left a comment
There was a problem hiding this comment.
Read the full diff, a test-only change for #35875 that touches no product code. It builds a synthetic home at test time with made-up ids and text in the real row shapes. It then pins restore cost by counting work (rows parsed at the reader, CrewLog._walk_from calls), never by timing, so it cannot flake on a slow runner. Each target over today's bar is a ceiling that fails on regression, plus a strict xfail stating the target, so reaching it cannot go unnoticed. Regexes, if any, would be test-only; there are none. The 6 reds are cancelled or earlier failed runs that newer passing runs on this head superseded. LGTM.
KiroCrewCaptain
left a comment
There was a problem hiding this comment.
Tier 1 auto-approve: test (1 files). Criteria: no conflict, no requested changes, security path denylist clean, design-doc gate clean, security checklist all-NO, no reviewer-lane findings, readiness signal fresh for this head SHA, declared category matches the diff. Category: adds one test file that pins restore cost by counting parsed rows and crew-log walks; no source change.
Refs #35875
Problem / Motivation
Goal: Pin the cost of restoring a large session with count-based CI tests, so the bounded-restore work is measured against a fixed fixture.
Nothing in CI guards how many rows a restore parses for a 16 MiB transcript, or how many times a slot's crew-log unit is walked when its folds are primed cold. The plan for bounded restore is in draft PR #35916.
Why it matters
Without a fixture, a performance change to restore has nothing to be measured against, and a regression in either count would reach main unnoticed.
Not a goal
What changed (motivation → approach → change)
Restore cost has to be pinned before it is reduced. The new test, test/test_restore_large_session_fixture.py, builds a synthetic home at test time (made-up ids and text, real row and record shapes): one 16 MiB transcript plus 40 small tabs, and one crew-log slot unit of about 2 MiB written through CrewLog. The unit is about 1/8 of the 16 MiB size in the plan; walks per fold do not depend on unit size.
Backwards compatibility: Compatible: test-only change, nothing is rejected, renamed or removed.
Removes nothing: only a test file is added.
Tests
The new file runs 5 passing tests and 2 strict xfail tests (the targets). With --runxfail the targets fail on the counts above (5451 vs 500 rows, 12 vs 1 walks), so they fail for the intended reason.
Proof: not provable -- test-only change
Manual verification
N/A — unit coverage sufficient; the change is a test.
Related Issues
Refs #35875
Checklist