Skip to content

routing-overlay: merge develop - #3444

Merged
timohueser merged 146 commits into
routing-overlayfrom
claude/overlay-sync-develop
Nov 13, 2026
Merged

timohueser merged 146 commits into
routing-overlayfrom
claude/overlay-sync-develop

Conversation

@timohueser

Copy link
Copy Markdown
Owner

Mode: skip: a merge of develop into the integration branch of #3227
Requirements: none (not checked: the verification token is absent in this container)
Public docs: as on develop

Why

CI tracks the newest stable Rust. The clippy of Rust 1.100 reports needless_range_loop, so each pull request into routing-overlay fails the job clippy. develop has the fix (07128e5d4) and its CI is green. routing-overlay did not have develop since 1e2f0091.

What

  • A merge commit of develop at a4fd2ae34 into routing-overlay at 9be8cc4df. Git merged it with no conflict.
  • One more commit: the same lint in code that only the overlay has, planner/router-build/src/partition.rs (a range loop over self.level becomes an iteration of the slice).

No commit of develop in this merge changes planner/router/src/ or planner/router-build/src/, so the routing bake does not change.

Checks

  • cargo clippy -p planner-router -p planner-router-build -p planner-service -p obc-data -p obc-data-steps --all-targets --all-features --locked -- -D warnings with Rust 1.100, and cargo fmt --all --check, in the cloud container.
  • Left out: the clippy of the whole workspace and the test suites. CI runs them.

Generated by Claude Code

claude and others added 30 commits November 11, 2026 10:54
The Rules table named one test without its file.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
… of the Rules tables

The weekly report keeps the skips, the ratchet and the list of pull requests
that touch an approved system since its doc changed. The size estimates and
the drift step with its Anthropic API batch are removed, with their settings.

R3 is removed: a shell's line count is not a violation. The baseline holds the
IDs of the known violations with no sizes, so more files under a known R1 or
R4 entry are not new.

check now fails when a name in the last column of a doc's Rules table is not
in a file that the same cell links.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
- The number of changed lines alone never gives L. An L design runs the critic
  and readiness. A sub-issue of an approved epic runs only the critic and does
  not repeat its epic. An issue body holds only the current design.
- The Status line has no done value: a closed issue is done.
- The code reviewer gets the approved system docs of the touched systems and
  reports the statements that the diff makes false.
- document writes four sections and no list of parts, shows the draft in one
  message, skips the hay for 5 leaves or fewer, and writes a short doc for a
  shell or a system with one leaf.
- The router no longer lists commands that have no reference, or a glossary.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
Seven of its eleven systems were flagged only for files under tools/, which
R4 reports. The other four span build roots that stay as they are. The
runtimes table of the config goes with it.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
build and fix open a draft. After the fixes of the own review, the pull
request is marked ready, and the bot reviews the final diff.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
A new adapter gives one edge for each relative import between two tracked
web files. A file edge counts for the owners of the two files, and an edge
from a test file is a test edge. The scan resolves 3,016 of 3,029 relative
imports; the rest name generated WASM packages.

The baseline grows by 42 entries that the check could not see before: 39
cycle edges and 3 undeclared dependencies of a system on a shell or in a test.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
The three shells of the web app declare the systems that they use. The map
shows one cycle of 16 systems: web imports join the two Rust cycles through
the operator page of obc data, the shared code of the builder shell, and the
imports between TrailView and route planning.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
The data engine has no knowledge of maps, and the page draws one. The builder
shell lists its own folders, because two shells cannot claim one file. A web
folder that no doc claims is now a file with no owner.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
The data engine and the systems around it are outside the cycle, now that
the operator page belongs to its shell.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
- The Rules check matches the whole name, and reads only the rows of the table.
- The report still prints when the history cannot be read.
- The help text of obc cairn no longer names --drift.
- The README and the suite comment list the faults that fail for every doc.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
A new adapter finds violations that the check could not see before. In that
change, check --base allows the baseline to add the violations of HEAD, and
no other entry.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
- The report raises when git cannot read the history of an approved doc.
- review.md marks ready only a pull request that the agent opened as a draft.
- document.md states which names of a Rules table the check tests.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
Limits::machine budgets three quarters of the physical memory and does
not read the cgroup limit of the process. Under a scope with a lower
MemoryMax, the engine starts more heavy steps than the scope allows, and
the OOM killer of the scope stops the bake.

Limits::machine now takes three quarters of the smaller of the physical
memory and the memory.max of the cgroup v2 of the process or a parent.
The run log prints the job and memory limits before the first step.
The job count stays on available_parallelism, which already honours the
affinity mask and the cgroup cpu.max.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvHzQoZca9zdxx3yoLa
A personal Apple development team cannot sign the Associated Domains
capability, so `obc ios-companion` fails to build for a device. The
entitlement has no effect yet: the domain does not serve the
apple-app-site-association file.

project.yml no longer declares the entitlements, and the generated
OBCCompanion.entitlements file is gone. The universal-link handler and
LaunchFlowModel.acceptsPairingLink stay. The entitlement comes back with
the paid team and the domain file.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvHzQoZca9zdxx3yoLa
The test compares the terrain shadows with a reference ray march and
allows a mismatch at 2 % of its points. That limit has no source. With
the GLO-30 fixture terrain the test measured 1.7 %; with the
swissALTI3D fixture terrain it measures 2.3 % and fails, with no change
of the shadow code. The difference is the coarse terrain levels of the
far horizon march, which the limit was never chosen for.

The flat-ground helper of the harness goes with it: only this test
used it. The sun probe of Terrain.check stays for now; no test reads
it.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvHzQoZca9zdxx3yoLa
cairn: a smaller check, lighter designs, and the dependencies of the web code
The display contracts had a generic frame trait, associated damage and
region types and neutral constructors that only the conformance suite
called. The board built its own damage and windows, wrapped the frame
in FbDevice64 at five sites, and recovered from a failed push with
Ls021Flpr::reset_diff, which no test called.

Presenter and OverlayPresenter now take the Device64Frame, the excluded
rows and a RowWindow. NativeFrame, RowDamage and the neutral
constructors are gone. Device64Frame::draw_target replaces the five
FbDevice64 wrappers.

The presenter repairs itself: a failed present or overlay push resets
its row store, so the next present pushes every row outside exclude.
reset_diff and its two calls are gone. The host double fails after the
diff as the board does, and a new test proves the recovery after a
failed present and after a failed overlay push.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvHzQoZca9zdxx3yoLa
The system doc named NativeFrame, RowDamage and reset_diff, which are
gone.

Step 1 names Device64Frame::draw_target. Step 9 says that the presenter
resets its row store after a failed present or overlay push. The parts
table, the contracts decision and the word list follow the concrete
traits.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvHzQoZca9zdxx3yoLa
The reset after a failed present lived in the board and in the test
double, and no test reached the board's copy.

RowDiff now keeps one confirmed flag. Spans from diff_clipped leave the
store unconfirmed until the presenter calls confirm after a push that
reached glass. While unconfirmed, the next diff_clipped returns every
row outside exclude. A diff with no spans keeps a confirmed store
confirmed. The board and the double confirm a push that reached glass
and no longer reset on a failed present; a failed overlay push still
resets. RowDiff::diff folds into diff_clipped.

The rowdiff unit tests cover unconfirmed, confirmed and empty diffs.
Two comments get their reasons right.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvHzQoZca9zdxx3yoLa
Step 9 said the presenter resets its store after a failed present. The
presenter now confirms a push to RowDiff only when the rows reach
glass, so an unconfirmed push is pushed again. A failed overlay push
still resets the store.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvHzQoZca9zdxx3yoLa
A name in backticks, then " in ", then a link must be in the file of that
link. So a Swift test name and a JavaScript test title are checked too, and
each against its own file.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
Each doc says what the shell or system does in at most 150 words, with the
rules that a test enforces and the decisions that the code shows. The owner
approved the eleven docs and eight new purpose lines.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
Limits::machine takes the memory.max of the cgroup that
cgroup::memory returns. That reader gives the cgroup with the least
free memory, so a parent with a high limit and much memory in use hides
the lower limit of its child, and the budget is too large.

cgroup::limit gives the least memory.max of the cgroup of the process
and of its parents, and Limits::machine uses it. The two readers share
the walk up to the root.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvHzQoZca9zdxx3yoLa
…ains entitlement

The README names the cause, what a pairing link does without the
entitlement, and where a paid team finds the value to add.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvHzQoZca9zdxx3yoLa
… title

A link with an anchor hid its name from the check. A title of several words
passed when it was only the start of a longer title.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
…eps for the check

The web bundle rule, the task list of the board image, the readers of the
data registry and the flag of the parts list now say what the code does.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvG9jaRQXPnLGp5hbbr
docs: short docs for the shells, the data registry and the hardware; the Rules check tests any name with its file
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011wmjvKzamiTf3Th8ePmUoi
iOS companion: remove the Associated Domains entitlement
timohueser and others added 25 commits November 12, 2026 22:14
# Conflicts:
#	builder/web/src/Shell.svelte
#	docs/design/web-planner.md
A stopped run adds no file to a new list, and the listing stops at once.
Inflation stops past the declared size and keeps large entries in Blob
parts. A specific bike of the file stays under the Strava type Ride, and
a Strava sport that is not by bike or on foot refuses the file. A ride
with no usable name gets a name from its date. Names in a ZIP with no
UTF-8 flag are UTF-8 when they can be. Cancel and Close use the close
function of the window.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…y change

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- The session takes the stored revision of its plan with the name, so it
  can delete its plan after a rename.
- The cover record keeps the credits of the still. The picture shows them
  with the credits of the basemap behind a button (Credits.svelte, shared
  with TrailView); an older cover shows the credits of the release.
- The page asks for a cover that is not settled for the line of the
  record, also when it shows the cover of an earlier line.
- #/planner/ID for the open plan keeps its Undo history and map view; a
  plan that fails to open shows the error in the status line.
- At most 3 place searches for climbs at a time.
- An empty name does not rename.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
renameRoute raises the revision that the tab knows only when the tab knew
the stored one, and removeRoute checks against the newest revision that
the tab knows, as saveRoute does. The session takes only the name of its
plan. The picture shows its credits in both layouts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A table gives the speed on the flat of each activity at Moderate:
touring 19, gravel 21, road 25, mountain bike 15.6, on foot 4.5 km/h.
The factor on the riding time of the router is its speed on the flat
over the speed of the activity at the pace. The pace now also scales a
plan on foot; in a bike plan a pushed stretch keeps its time. The hint
beside the control goes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
nearest takes a limit in pixels of the image; Cover passes NEAR_PX (24 CSS
pixels) and gives no place further away. The browser test of a cover
points near the line.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Route page: the picture, the numbers and the buttons
The planner converts each distance and height it shows with distanceIn and
heightIn and writes the unit of the settings. With kilometres no text
changes. Short notes keep metres with kilometres and show tenths of a mile,
at least 0.1 mi, with miles.

Co-Authored-By: Claude Opus <noreply@anthropic.com>
# Conflicts:
#	builder/web/src/lib/library/store.test.ts
#	builder/web/src/lib/library/store.ts
heightRange steps 100 m with kilometres and 100 ft with miles, so the
profile labels stay round. noteDistance uses formatDistance for miles.

Co-Authored-By: Claude Opus <noreply@anthropic.com>
A card whose record gets no cover shows the line on a still picture of
the planner map with no labels. The cover job makes the picture one time
for each theme and stores it in the field `map` of the cover record.
`Cover` takes `place` (row, card or page) in place of `ask`.

Co-Authored-By: Claude Opus <noreply@anthropic.com>
A planned tour of Komoot has a time at each point. The rider chooses the
form on its row; a file that starts later than now is a planned route.
Either stored form makes the file known.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A failed load of the picture code counts no try. Only a card loads again
after a theme change. No map is made for a hidden page or an aborted job.
Tests: pictures continue after a still timeout and with a software renderer.

Co-Authored-By: Claude Opus <noreply@anthropic.com>
Covers: a still picture of the map for a card with no cover
planner: distances and heights in the units of the rider
Add files: GPX, FIT and ZIP; a ride or a route; no file two times
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DTHu3HYv6LXGFQ8VecfWNK
Clippy of Rust 1.100 reports the range loop. develop has the same fix for its own code.

Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01DTHu3HYv6LXGFQ8VecfWNK
@coderabbitai

coderabbitai Bot commented Nov 13, 2026

Copy link
Copy Markdown
Contributor

Important

Review skipped

Auto reviews are disabled on base/target branches other than the default branch.

Please check the settings in the CodeRabbit UI or the .coderabbit.yaml file in this repository. To trigger a single review, invoke the @coderabbitai review command.

Run configuration

  • Source mode: Auto
  • Repository: Unavailable
  • Branch: Unavailable
  • File: No configuration file from this repository applied
  • Commit: Unavailable
  • Layers: defaults
  • Review profile: CHILL
  • Plan: Advanced
  • Run ID: 637277cf-3054-45f6-a33f-d380fddf8aa8

You can disable this status message by setting the reviews.review_status to false in the CodeRabbit configuration file.

Use the checkbox below for a quick retry:

  • Trigger review
  • Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

Share

@timohueser
timohueser merged commit 55100c5 into routing-overlay Nov 13, 2026
28 checks passed
@timohueser
timohueser deleted the claude/overlay-sync-develop branch November 13, 2026 08:03
@github-actions github-actions Bot mentioned this pull request Nov 16, 2026
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.

2 participants