Repository navigation
Conversation
Co-Authored-By: Claude Opus 5.5 <[email protected]> Signed-off-by: adssib <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]> Signed-off-by: adssib <[email protected]>
|
Claude finished @adssib's task in 1m 29s —— View job ✅ Ready to merge: the metadata is complete, the golden and drop tests use real saved commits, it fetches only through Rules
BugsNone found. The change to Verified
Non-blocking
|
| Criticality string `json:"criticality"` | ||
| } | ||
|
|
||
| func (e upcoming) key() string { return e.Location + "@" + e.Date } |
There was a problem hiding this comment.
Non-blocking (precision): because the key includes Date, an entry GitHub postpones (same location, new future date) reads as a fresh announcement. And an entry that's removed by mistake and put back while its date is still in the future gets announced a second time. In both cases the event gets a new version, so it also gets a new fingerprint, and downstream sees a second deprecation for a location it was already told about. The page lists the postponement under Known gaps, which is fine. If you want it tighter, you could skip a location that any earlier commit in the pair's from already had (keyed on Location alone), and add a line to the page about the "removed and put back while still in the future" case.
| if !ok || typ == "" || rest == "" { | ||
| return event.Subject{}, false | ||
| } | ||
| if strings.HasSuffix(typ, "Input") || typ == "Query" || typ == "Mutation" || typ == "Subscription" { |
There was a problem hiding this comment.
Non-blocking: the doc comment and the provider page both say "a root field's argument", but this branch also matches a root field itself (Query.foo, with no argument part). That becomes a request_parameter whose property is Query.foo, even though removing a root field is really removing an operation. Either the page could say so, or you could require a second dot for the Query/Mutation/Subscription case (strings.Contains(rest, ".")) and let a bare root field fall through to schema_property (Query / foo). It might be worth a row in TestGraphQLLocations either way.
…t announced twice Co-Authored-By: Claude Opus 5.5 <[email protected]> Signed-off-by: adssib <[email protected]>
Co-Authored-By: Claude Opus 5.5 <[email protected]> Signed-off-by: adssib <[email protected]>
Stacked on #238
Part of #188 (I21.3,
github/graphql)What this changes
New source
github/graphql(kindcatalog, layerapi). As the audit recommends over the HTML page I21.3 proposed, it readssrc/graphql/data/fpt/upcoming-changes.jsonin github/docs, the file docs.github.com's GraphQL "Breaking changes" page is built from (re-checked 2026-11-21, 200), at every commit that changed it (29 since 2023-05).deprecationevent, itsversionthe removal date (the API has no versions; like a model shutdown date),previous_versionthe commit before. An entry dropped while still upcoming and restored later is the same event (same fingerprint).api_operation):request_parameterofPOST /graphql;schema_property(type →schema, field →property).owner(a GitHub team or handle) left out.npm:@octokit/graphql.Checklist
testdata/README.mdrowsproviders/github-graphql.mdmake checkpassesCalls I made (please check)
api_operation. The contract has no GraphQL subject, so a GraphQL location is mapped as above. If the client should treat GraphQL differently, the contract needs a subject first.breakingevent. The deprecation names the date, and an entry leaving the file doesn't say it shipped.Related PRs
🤖 Generated with Claude Code