Repository navigation
Merge train: 53 reviewed PRs (2026-11-06) - #1146
Merged
Merged
Conversation
…-namespace-class-nodes
… into pr/code-review-graph-998
… into pr/code-review-graph-1060
`class ::Root` inside `module Outer` qualified to `Outer.Root`, and a leading `::` left an empty segment behind (`Outer..A.B`, `.X.Y`). A leading `::` anchors the name at the top level, so the enclosing scope is now dropped instead of stacked, and the separator is stripped. Also correct the comment that still said "immediate scope only": the code stacks the full enclosing chain. Two test comments said the same.
… into pr/code-review-graph-1104 # Conflicts: # code_review_graph/tools/build.py
…king Translate the AspectJ '..' wildcard to \.(?:[^.]+\.)* instead of a bare '.*', so repeated '..' segments in an execution() pointcut can no longer stall build/update through regex backtracking. Add the edge-case tests from review/pr-916 and drop the xfail marker on test_pathological_double_dot_pointcut_resolves_in_bounded_time. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Start the filesystem observer before startup reconciliation so edits made during post-processing are captured. Serialize startup work and event batches through the same processor lock, and prevent the initial debouncer wake-up from being lost. Add regression coverage for startup edits, event serialization, and first-event delivery.
Double-quoted keys and values in Route::group options were not read as literals, so the "prefix" key was dropped and the endpoint got a path Laravel does not serve. Group options that are not a literal array (a variable, config(), array_merge(), a spread element or a dynamic key) were treated as empty; routes inside them are now skipped, as with prefix($var). Arrow-function group bodies now count as groups, and double-quoted route paths without interpolation produce endpoints. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Capture and clear affected flows before file node IDs churn, remap community assignments by qualified name, and normalize relative git paths against absolute graph file_path values. Fixes #569
Add maintainer-requested lifecycle parity cases for #569: relative paths without repo_root, underscore/percent filename literals, suffix collisions, deleted-community cleanup, and rename vs full-rebuild parity.
Escape the Windows backslash separator in expand_changed_file_paths so relative matches stay boundary-anchored (b.py no longer hits ab.py). Collapse duplicate lifecycle path expansion and flow ID queries, and assert real community-assignment equality vs a clean rebuild.
Align expand_changed_file_paths with #774 so incremental matching uses the same file_path identity as the graph store after rebase onto main.
Avoid clearing nested same-basename flows during incremental updates, fix the mypy batch type reuse, and align ambient-transaction handling with store_flows rollback convention.
Cover the reviewer repro where a failed incremental parse must keep existing flow and node state until a later successful replacement.
Skip the no-op early return when base matches the stored HEAD so empty commits still refresh git_head_sha. Add the PR #673 review lifecycle suite and drop the temporary freshness xfail marks.
semantic_search_nodes_tool takes include_source, expand_neighbors and max_source_lines. With include_source each hit carries its source lines; with expand_neighbors the top hits also carry their direct callers and callees with source. A shared line budget caps the response, and a stored path that resolves outside the repository is never read. Search ranks test code below production code unless the query asks about tests, so a short test stub no longer outranks the implementation. The installed instruction block points agents at the new options and still says not to change code from graph output alone. The previous block is kept in _legacy_instructions.py so upgrade and uninstall still recognise it. The README and docs/REPRODUCING.md publish the 2026-09-22 agent_baseline capture (median 6.65x across 18 questions on 6 repos). The 2026-05-25 impact_accuracy CSVs are removed; the 2026-08-02 captures replace them. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts: # code_review_graph/incremental.py
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Ruby methods always carry their full lexical scope (#933), so the leaf collision index from #1087 now covers Python only, and the docs and tests say so. Generators are not component wrappers' functions (#972 with #1131). Partial graph coverage counts tracked files only (#881 with #1094). A pending marker recomputes all derived data (#1105), so that test starts from a finished graph, the deleted-seed case now passes (#1062 with #673), and hooks.json keeps its events under a hooks key (#1102 with #1126). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…F-8 out of update and review (review of #1094) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…1108) Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Both sides fixed fork inheritance and the flock wrapper docs. Keep the author's versions, which also register the lock before the flock call. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…accepting plain directories (review of #1126) The new commit made a filesystem without flock crash every build with a traceback, and made update and postprocess refuse a plain directory that they accepted before. Neither is what the lock is for. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
# Conflicts: # code_review_graph/tools/build.py
The #1087 guard read a dotted scope as a leaf collision. Every nested Ruby class has one since #933, so same-file Ruby calls lost their targets: devise lost 23 call sites and sidekiq 71. A Ruby leaf now counts as ambiguous only when two scopes share it, and both repositories lose none against staging. Also: install leaves a pre-commit hook in another language alone, since a shell block appended to it made every commit fail; uninstall removes the empty Codex skills folder; the docs give the current schema version and say that Codex runs hooks only once they are trusted. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
code-review-graph reviewOverall risk: 0.85 (CRITICAL) — 500 changed function(s)/class(es), 130 affected flow(s), 282 test gap(s) (140 with no tested caller found, 142 reached only through a caller) Risk-scored changes
Affected execution flows
Test gaps
Reached only through a callerNo test names these directly, but a tested caller reaches them along the call path shown. That is a path in the call graph, not a record of execution -- the caller's tests may never take this branch. Treat it as where to look, not as coverage: these count as gaps and are scored as untested.
Token savings: this graph-backed report used ~1,186,192 fewer tokens (~66%) than reading every changed file in full (estimated, chars/4 approximation).
Powered by code-review-graph — local-first analysis; no code leaves the CI runner. |
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Contributor
|
Checked #1132 on the train head (da989c5): the commit is present unchanged, The four failing checks are in other areas. Failure names from the logs:
|
Contributor
|
Follow-up on the
The reason looks like the test's premise, not the feature. On 3.10/3.12 the same source evidently parses, so the parser does not take the "file defeats ast.parse" branch, the shim is not followed, and the assertion sees A deterministic version would force the failure (patch |
Legend icons were stretched to 104px by the page-wide svg rule, so the legend filled the screen and a sixth node kind pushed it under the filter panel. This is older than the train, and the icons are now 16px. Two tests also depended on their platform: Python 3.10 accepts the deep expression (#1133), and Windows paths have backslashes (#1108). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This was referenced Nov 6, 2026
Closed
Closed
gowrishacv
pushed a commit
to gowrishacv/code-review-graph
that referenced
this pull request
Nov 6, 2026
test_thousands_of_wrapped_components_in_one_scope took 24.6s against a 20s limit on the staging push run for tirth8205#1146, on the same tree that had passed on the PR. The bound is there to catch a blow-up, so 90s still does that. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This merges 53 reviewed PRs into
stagingin one CI run. Merge it with a merge commit, so each PR is marked as merged under its author's name.This is not a release. The version stays 2.3.9, with no tag and nothing sent to
testingormain.Included
Other tools and install
Imports and registry
Review and query
Languages
class Foo::Bar) #933 fix(ruby): emit Class nodes for compact namespace declarations (class Foo::Bar) (@idarfan)Build, update and watch
Dependencies, tests and speed
How it was checked
Each PR was read in full, merged with current
staging, tested and probed for faults, and Copilot's comments were checked against the code. Real faults were fixed on the train as separate commits. Those commits carry the maintainer's git name with a Claude co-author line.On the combined branch, on macOS only:
Full suite: 6,422 passed. 3 token-budget tests fail, and the same 3 fail on plain
staginghere. The check found the output is byte-identical tostagingand the cost comes from a long temp path, so CI is the real answer for those three.Packaging gate: 17 passed, run before the last fix commit.
Installed wheel on real repositories: this repo, devise, apollo and documenso. An incremental update matches a fresh build, except for criticality on 15 of 874 documenso flows. No file lost. Graphs from v2.3.9 and
stagingupgrade to schema 14.MCP through the real client: 30 tools and 5 prompts, cancel,
--multi-worktreeand--auto-watch.All 16 platforms: install, reinstall and uninstall in a throwaway HOME. Real
codexandclaudewere run.claude plugin validatepasses.Browser tests: 17 passed on real Chromium.
Upgrade suite with 2.3.9 added to the release list: 42 passed, 0 failed, 4 expected failures on 2.3.7 and 2.3.8.
CI on the first push found what the Mac run could not: a test tied to Python 3.11 or newer, a Windows path check, and legend icons stretched to 104px by a page-wide CSS rule. That last one is older than this train, and all three are fixed here.
The end-to-end run found one bug the test suite missed: #933 with #1087 lost same-file Ruby calls (23 sites on devise, 71 on sidekiq). It is fixed, and both now lose none against
staging.Not tested
Worth knowing
Auth.User.to_s. The schema 14 migration makes the nextupdatere-key Ruby files.buildonce to pick up generic YAML and workspace imports.post-checkout,post-mergeandpost-rewritehooks.--no-hooksskips them.main, and were retargeted tostaging.🤖 Generated with Claude Code