tidy-dataplane-conn: one missing package, two stale promises - #217
Merged
Conversation
The layer map opened with "Nine packages, one layer each" above seven bullets, and internal/mrtr was the one with no bullet — though it has a section of its own further down. Neither number was right: counting the way the bullets themselves do, with toolsig and toonenc folded into their parents, there are eight. The missing bullet is the half that costs something. This list is where a reader goes to find which layer owns what, and MRTR input resolution is exactly the kind of thing someone arrives looking for; absent from the map, it reads as living inside downstream, which is where its retry loop is but not its resolution. The new bullet says which half is which. Docs only. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Two comments in admission.go cited machinery that is gone. The type's reason for existing was "the 7.2 argument self-heal re-issues the same call with repaired arguments" — that stage went with the rest of the argument-inspecting pipeline, and nothing in the tree re-issues a call today: pipeline.Execute invokes req.Call exactly once, at one line. And Wrap's nil case was explained by "the wiring can wrap both CallRequest fields", where CallRequest has one call field. A justification naming a mechanism that does not exist is worse than none: a reader who greps for the self-heal finds nothing and cannot tell whether the single-charge memo is load-bearing or a leftover to delete. It is neither today — it encodes a rule about agent intent that outlives whichever execute path is current — and that is what the comment now says. No behaviour change; the memo, the guard and the tests are untouched. make ci green. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
…ments Two claims in this file were false. stderrTail said "the FIRST line of the window is dropped when the window is full, because a 4 KiB cut lands mid-line and half a line in an error report is worse than no line" — no code does that, at either layer: transport's tailBuffer cuts on a byte boundary, and tailLines keeps the last n non-blank lines with no notion of a fragment. And tailLines called itself "unit-tested directly" while having no test at all; its only reference in the tree was its own definition and its one caller. The combination is the bad one. This projection is how every handshake failure and every respawn line reports a dead child, so a reader deciding whether a crash report can be trusted was reading a guarantee that was not there, backed by a test that was not there. tailLines now has the tests, including one that pins the fragment being reported whole — asserting what the code does, so it is the test that has to change when this is fixed. Both comments say what holds. dataplane.md carried the same false sentence and now carries the gap instead, with why closing it is not a tidy-night edit: tailLines is handed the window's contents and not its capacity, so plumbing the cap through changes the error text. No behaviour change. make ci green. Co-Authored-By: Claude Opus 5 (1M context) <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
1 participant
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.
Nightly-tidy slice over the connection half of the data plane —
internal/downstream,mrtr,router,pipeline,ratelimit— and the sections ofdocs/modules/dataplane.mdthat cover them.
Nothing here touches the frozen surface: the gate chain and its order,
RouteOf, thesingle execution path, or any failure direction.