Fix/pointer gaps - #281
Merged
Merged
Conversation
Several checks walk a map looking for fault: definitions, paths, status codes, response headers, properties. Go randomises map iteration, so the same document could be reported differently from one run to the next — in which order the findings came out, and, where a check stops on the first fault it meets, which of two equally faulty definitions was named at all. Those walks now take their keys in sorted order. "First" therefore means first in definition-name (or path, or status-code) order. How many findings a document yields, and when a check stops, are unchanged. The circular-ancestry check in validateDuplicatePropertyNames returned on the first offender whatever the options said, so a second circular definition could never be heard from. It now returns only when ContinueOnErrors is false, and otherwise moves on to the next definition without descending into the loop it just found. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Signed-off-by: Frederic BIDON <fredbi@yahoo.com>
Located.Pointer is documented as a JSON pointer into the validated document, and in four families of finding it addressed nothing. Values checked by a borrowed validator. The parameter and header validators are the ones a generated client uses at runtime, where the name of the parameter is all a caller has, so they locate a finding by that name. Spec validation borrows them to check a default or an x-example, where the name addresses nothing: those findings now sit on the value's own node. A body parameter is entered through its schema, as a response already was. The path template. A parameter missing from a path template was reported under the variable name rather than the path holding it. Findings that knew their site but did not say it. A duplicate operationId is counted over the operations rather than a flat list of identifiers, so it can name the first operation declaring it. An unresolved reference is located at the first local $ref that addresses nothing, and the separate message raised while expanding a response carries that response. Faults reached through a $ref. Checks walk an expanded document, so a finding below a reference came out with a pointer descending into a node holding nothing but "$ref". Such a pointer is now followed to what the reference leads to, as many times as it takes; one that stops at the reference itself is left where it is. Three smaller gaps close with them: a document holding no paths is reported at the root rather than under /paths, the format validator records the location it holds, and a parameter too broken to be identified keeps its name in the message while the pointer stops on the array that holds it — pathToken gains a cosmetic kind, the converse of structural, for that. Finally, a location is trimmed to the deepest node the document holds. This runs once every check has had its say and only ever shortens, which is what makes the pointer always resolve. Over the fixture corpus, findings whose pointer addresses nothing go from 132 to 0, and exact locations from 1955 to 2161 of 2213. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Signed-off-by: Frederic BIDON <fredbi@yahoo.com>
A definition is not the only place a document says an object must hold a property: so does the schema of a property, of an array item, of an additionalProperties. Each is a self-contained object definition, and a required entry naming something it never declares is the same modelling slip wherever it sits. Only the top of a definition was checked, so the slip went unreported one level down. The check now walks the schemas a definition holds inline, and locates each finding on the offending entry of the required array it walked into, e.g. /definitions/A/properties/inner/required/0. A finding about such a schema names it by the way down to it — schema "A.inner" — rather than naming the definition holding it, which would send a reader to the wrong place. A definition still names itself, in the words it always used. Two boundaries keep the walk honest. A schema written as a $ref is left alone: it is checked where it is defined, and following it would report the same slip twice and would not terminate on a recursive definition. And inside allOf, anyOf, oneOf or not, a member is a fragment of a constraint rather than a complete definition — its required entries speak of the instance the whole composition describes, are legitimately met by a sibling member or by no declaration at all, and are enforced when data is validated. Their own member schemas are still walked. That second boundary also fixes a false positive that predates the walk: a definition requiring a property contributed by one of its allOf members, or by a base definition it refers to, was reported as requiring something it does not declare. Over the fixture corpus, which includes the Kubernetes and Bitbucket APIs, this reports neither more nor less than before: the specifications that do carry nested required entries declare them correctly. Signed-off-by: Frederic BIDON <fredbi@yahoo.com>
fredbi
force-pushed
the
fix/pointer-gaps
branch
from
August 9, 2026 12:14
2d2c7a3 to
7cdf3d9
Compare
No -race with that flag, as the extra resource consumption with pool debug breaks CI. Signed-off-by: Frederic BIDON <fredbi@yahoo.com>
fredbi
force-pushed
the
fix/pointer-gaps
branch
from
August 9, 2026 13:02
15e5233 to
2bf3c27
Compare
Codecov Report❌ Patch coverage is
Additional details and impacted files@@ Coverage Diff @@
## master #281 +/- ##
==========================================
+ Coverage 93.13% 95.32% +2.19%
==========================================
Files 27 31 +4
Lines 3395 3618 +223
==========================================
+ Hits 3162 3449 +287
+ Misses 156 106 -50
+ Partials 77 63 -14 ☔ View full report in Codecov by Harness. |
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.
Change type
Please select: 🆕 New feature or enhancement|🔧 Bug fix'|📃 Documentation update
Short description
Fixes
Full description
Checklist