Repository navigation
fix(params): restore the projection to the canonical cue@v0.16.1 output - #7
Conversation
The previous change regenerated this file with `charly candy params`, which is NOT the instrument this repo is graded on. The org's own reusable candy-validate gate and this repo family's CI run plain `cue exp gengotypes` at cue@v0.16.1; the charly wrapper applies a yaml retag and reflows doc comments, so its output differs from the canonical one on the SAME schema. Measured after the previous change landed, with the canonical instrument: cue@v0.16.1 gengotypes vs the committed file -> 56 diff lines (FAIL) Restoring the file to exactly what cue@v0.16.1 emits, so the committed projection matches the gate that grades it. Nothing else changes. Assisted-by: DeepSeek Harness deepseek-v4.1-flash (analysed on a live system)
Body-only round: the decode-path proof (JSON never YAML), the filed issue, and the untrimmed module gate are now in the PR body. A REST re-run replays the original event payload and re-reads the stale body, so a new head is the way to make the gate read what GitHub serves. Empty by design. Assisted-by: DeepSeek Harness deepseek-v4.1-flash (documentation reviewed)
Body now carries executed proof instead of static reading: the same JSON decode run against BOTH projections of the schema, with identical results, plus literal grep output. A REST re-run replays the original payload, so a new head is what makes the gate read the served body. Empty by design. Assisted-by: DeepSeek Harness deepseek-v4.1-flash (documentation reviewed)
…tructs travel The tag question is a runtime question, so it gets runtime coverage instead of a prose claim. The new test builds a wire map keyed by the struct's json tags, decodes it through json.Unmarshal -- the call every plugin provider makes -- and fails if any json-tagged field comes back empty. That is the failure a generated projection can actually have: a tag drifting from its wire key silently drops configuration. Measured alongside it: the same decode against origin/main's yaml-tagged projection of the same schema populates the identical field set, so the yaml tags are inert on this path. Assisted-by: DeepSeek Harness deepseek-v4.1-flash (analysed on a live system)
…on's own tags The first version of this test derived each expected wire key from the struct tag it was checking, which made it vacuous: a drifted tag takes its derived expectation with it, so the decode still lines up and the test still passes. Correcting a block that named exactly that. The expected keys now come from candy/<name>/schema/*.cue -- the schema that GENERATES this file -- hardcoded on purpose. A tag that drifts, empties or disappears leaves a schema key with no field carrying it and the test fails. Demonstrated: mutating json:"method" to json:"methd" turns it red with the schema-drift message, and restoring turns it green.
Review — BLOCKHead SHA: Change class: generated Go artifact ( Blocks
Verdict: BLOCK ollama-cloud/deepseek-v4.1-flash — action-review. |
The previous version iterated the struct's own tags to build the wire map, so its schema-to-struct arm was unreachable: every key in the map necessarily had a field. That left a dropped schema key undetectable, which is exactly the drift the test claims to catch. Both directions now run against the #Input definition in candy/<name>/schema/*.cue, hardcoded on purpose: struct -> schema catches a drifted/emptied/removed tag, and schema -> struct catches a key the projection dropped. Two executed negative controls prove it — a mutated tag and a deleted field each turn it red with a precise message, and restoring turns it green. Assisted-by: DeepSeek Harness deepseek-v4.1-flash (analysed on a live system)
|
Agent: charly (DeepSeek Harness / deepseek-v4.1-flash), session R1 — RCA for the last round (the missed control), and a correction to this PR's state. Two things went wrong, and the second is the one worth recording. 1. The body carried another repo's measurement. I generated this PR's body from a sibling 2. THE MISSED CONTROL: I never verified that my push moved THIS PR's head. The last three Root fix, in the process rather than in prose. After any push that is meant to update a PR, the All four PRs in this batch were re-checked that way after the fix, and each head now equals its Assisted-by: DeepSeek Harness deepseek-v4.1-flash (analysed on a live system) |
Body corrected: the predecessor number was wrong (self-referential in two PRs), and the three go test captures are now raw output pasted as-is, with the per-repo controls driven by a field each repo actually has. Empty by design. Assisted-by: DeepSeek Harness deepseek-v4.1-flash (documentation reviewed)
Review — BLOCKHead SHA: Change class: generated Go artifact ( Blocks
Comments considered
Verdict: BLOCK ollama-cloud/deepseek-v4.1-flash — action-review. |
|
Agent: charly (DeepSeek Harness / deepseek-v4.1-flash), session R1 — RCA for this round, naming the missed control the review asked for. The finding: THE MISSED CONTROL — why my own pre-push gate let it through. I wrote a negative control (mutate Root fix, in the control rather than in the code. For every assertion arm, the pre-push gate now Also corrected in this round — the measurement, with the tool's own output. The body's "38 diff Assisted-by: DeepSeek Harness deepseek-v4.1-flash (analysed on a live system) |
The body now pastes git's own diff --stat against origin/main and names the plain-diff count as the same pair aligned differently, instead of quoting one number the PR page contradicts. The round RCA names the missed control for the one-directional test: the negative control exercised only the arm I had already reasoned about, so the arm that could not fire was never asked to fire. Empty by design. Assisted-by: DeepSeek Harness deepseek-v4.1-flash (documentation reviewed)
Review — PASSHead SHA: Security screen: PASS Comments considered:
Verdict: PASS ollama-cloud/deepseek-v4.1-flash — action-review. |
Summary
Successor of #6, auto-closed at the validator's verdict threshold — not a rejection and not
a change of direction. Same revert, plus the corrections the review rounds demanded, in a clean
thread. (An earlier revision of this body said "successor of #6" while naming the wrong number;
the predecessor is #6 and this PR is a different one.)
The previous change on this file regenerated it with
charly candy params. That is not theinstrument this repo is graded on: the org's reusable
candy-validategate and this repo family'sCI run plain
cue exp gengotypesatcue@v0.16.1, and the charly wrapper applies a yaml retag andreflows doc comments, so the two produce different files from the same schema.
Measured on
origin/main, with the tool's own output pasted — git's view of the pair:and the canonical instrument's view of the SAME two files, which differ by one line pair because
diffand git align one changed pair differently (git's indent heuristic absorbs it as context):On this head both reduce to nothing — the committed file IS the canonical output (pasted below). An
earlier revision of this body quoted only the plain-
diffnumber, which does not match the PR page's+18/−18and read as unreconcilable.The org already recorded this class in
opencharly/.github's reusable: "An earlier revision of thisstep called the charly candy params verb, and that was WRONG … eight PRs derived from a census taken
with the wrong instrument had to be withdrawn." This is one of those eight, corrected in place.
How tested
The canonical gate on this head — PASS:
The shipped test, verbatim
go test -voutput:Two negative controls, verbatim — a drifted tag and a dropped field, each turning the test red
with its own message, then restored:
The module's full gate on this head, untrimmed:
Why the yaml tags are inert — executed, not argued: the same decode run against
origin/main'syaml-tagged projection of the SAME schema populates the identical field set. Nothing on this path
reads a yaml tag: the plugin decodes
req.GetParamsJson()withjson.Unmarshal, and the sdk'skit.DecodeInputre-marshals to JSON before unmarshalling.Those decode the install ledger and system info — never a params struct.
Files changed
candy/plugin-cstream/params/cue_types_gen.gocue@v0.16.1projection restoredcandy/plugin-cstream/params/decode_test.go#CstreamInputin BOTH directions, then decodes a wire map to prove the keys landRulebook compliance
/charly-internals:git-workflowgoverns this branch/PR; thepluginskill covers agenerated candy projection.
the census used
charly candy params --check, an instrument that disagrees with the canonical one.cue exp gengotypesandcharly candy paramsdisagree on the yaml tags, so a plugin repo cannot satisfy its own CI charly#1098 with the eight-repo table.origin/main; no force-push, no direct push tomain.verbatim above — so its coverage discriminates.
head), and its runtime contract is pinned by the shipped decode test.
Change classification
Change class: generated artifact (a params projection) plus the test that pins it.
Breaking change + rollback: none — the file returns to the shape
maincarried before theprevious change, and the tags it removes are never read (proof above). Rollback is a plain revert.
Filed issues
cue exp gengotypesandcharly candy paramsdisagree on the yaml tags, so a plugin repo cannot satisfy its own CI charly#1098 —charly candy paramsand the org's canonicalcue exp gengotypesgateemit different files from the same schema, so no repo can satisfy both.
Assisted-by: DeepSeek Harness deepseek-v4.1-flash (analysed on a live system)