Repository navigation
D-VAS R1: E0 on the real mailbox (publication token, mailbox lane socket, route-bound admission) - #2061
Conversation
…mailbox lane socket; E0 tests before the run Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMMWgnUU9g7Ldyn6sbk
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMMWgnUU9g7Ldyn6sbk
…rectly (two disables were not load-bearing) Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMMWgnUU9g7Ldyn6sbk
…ventory delta; codegraph anchors follow the DT64 insertion Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMMWgnUU9g7Ldyn6sbk
Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMMWgnUU9g7Ldyn6sbk
|
Important Review skippedAuto reviews are disabled on base/target branches other than the default branch. Please check the settings in the CodeRabbit UI or the
|
…exes regenerated Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMMWgnUU9g7Ldyn6sbk
Codex Review Summary: ✅ CompletedThis comment shows the latest Codex review activity on this pull request.
ℹ️ About Codex in GitHubYour team has set up Codex to review pull requests in this repo. Reviews are triggered when you
Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings. |
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: c9317e15ef
ℹ️ About Codex in GitHub
Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".
…n's socket (review finding), before the fix Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMMWgnUU9g7Ldyn6sbk
…dress of another publication is refused (review finding) Co-Authored-By: Claude Fable 5.2 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XAqKMMWgnUU9g7Ldyn6sbk
ca0e423
into
claude/virtual-address-algebra-clean
Stacked on #2060 (research baseline). First implementation step of the roadmap
virtual-address-superpowers-roadmap-v1.md: stage E0,D-VAS-A1,D-VAS-A2,D-VAS-C2,D-VAS-D1, direct-classD-VAS-C1. Ledger:.claude/plans/d-vas-implementation-v1.md. Draft; not for merge without a merge decision.What E0 is
No guard, no effect, no selection, no
NodeGuid, no separate0xF###lens.Changes
cognitive-shader-driverMailboxSoA::published(&self) -> Published<'_, N>: a token that holds the shared borrow. No field added, no layout change.lance-graph-dt64BindStore::admit_routed(RouteAt, &MethodRef)andRefusal::NotRoutedClass.admitalone accepts a method of any registered class.lance-graph-mailbox-lanes(new)MailboxLanes<'_, N>: LaneSource. Lendsmeta,last_active_cycle,last_write_cycle(u32) andedges,temporal(u64) as the owner's own slices, populated prefix only. OneFrameStoreper publication.Why the frame is a borrow and not a counter
The lab plan (section 18, #2060) reads every mutation path of the mailbox: ten field setters and 19 public columns move neither
current_cyclenorlast_write_cycle; two writes in one cycle share a stamp; the cycle wraps onto the never-written sentinel. No stamp accounts for the writes a reader can see. What the owner can guarantee is that no column changes while a shared borrow is live.Publishedis that borrow, issued by the owner, and the socket's frame store lives exactly as long.Consequence: no frame, binding or lane address survives a publication. A frame that outlives the borrow needs a revision every write moves, which needs the public columns to stop being assignable. That is recorded as BLOCKED, not worked around.
Tests
crates/lance-graph-mailbox-lanes/tests/e0.rs, 7 tests on a realMailboxSoA<64>(40 rows populated, heterogeneous values), committed before the first run. Oracle: the owner's public arrays and the generator functions.set_metathat moves no stamp), two sockets over one publication sharing frames, another mailbox, another row order, another reader scope, a re-route, a foreignMethodRefwith the sameMethodId, another store's reference, a field outsidereads, a field outside the basis, a basis naming a column the socket cannot lend (f32,u16,u8,[u8; 12]are refused by name)Disable runs, each anchor asserted once and restored, all exit 101: route comparison skipped; all
Nrows lent; a lane number lending another column; the socket's own row-order/scope check; its own frame check. The last two passed at first because DT64 repeats both checks above the socket; a test that asks the socket directly was added.Local:
lance-graph-mailbox-lanes7 (debug and release),lance-graph-dt64117,cognitive-shader-driverlib +mailbox_cutover131. Clippy-D warningsclean onlance-graph-dt64,lance-graph-mailbox-lanes(all targets) andcognitive-shader-driver(lib).cognitive-shader-driver --all-targetshas three existingmanual_is_multiple_offindings inexamples/meta_prefilter_mask_probe.rs, a file this PR does not touch.Not established
Scopeissuer; the socket compares what the caller asserts.urn:mailbox:<column>is this crate's naming; no contract names the mailbox's columns.NodeGuid, guards, effects: R2 to R4.Generated by Claude Code