You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
Tracking spec change: Implementations may not deprecate a field that the interface hasn't deprecated #3766
There's a branch ready that adds this check (spec step 2.6) to TypesImplementInterfaces. It covers object→interface and interface→interface, for both SDL-built and code-built schemas. Before opening a PR, though, a question about rollout.
If it's on by default, it rejects several of the anonymized production schemas in src/test/resources:
fixture
fields deprecated on the implementation but not on the interface
large-schema-3
11
large-schema-4
44
large-schema-5
135
large-schema-federated-1
92
large-schema-federated-2
162
Only large-schema-4 is built by a unit test. The others are used by the JMH benchmarks.
Should this be on by default, with the fixtures updated in the same PR? Or would you prefer it opt-in, or held for a major version?
Activity
dondonz commentedon Apr 5, 2025
This has been approved and will soon get merged into the spec
feat: validate that implementations cannot deprecate non-deprecated i…
ybai08 commentedon Oct 18, 2026
There's a branch ready that adds this check (spec step 2.6) to
TypesImplementInterfaces. It covers object→interface and interface→interface, for both SDL-built and code-built schemas. Before opening a PR, though, a question about rollout.If it's on by default, it rejects several of the anonymized production schemas in
src/test/resources:Only large-schema-4 is built by a unit test. The others are used by the JMH benchmarks.
Should this be on by default, with the fixtures updated in the same PR? Or would you prefer it opt-in, or held for a major version?