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
Generated proto code is one generator version behind go.mod after #10 #18
go-grpc/proto/echopb/echo.pb.go and go-grpc/proto/pricingpb/pricing.pb.go on main carry a header naming protoc-gen-go 1.36.11, while go.mod has google.golang.org/protobuf v1.36.12. Regenerating changes one line in each file:
The generated code is otherwise identical. Nothing is broken; the committed files are stale by a comment.
Cause
#10 bumped google.golang.org/protobuf from 1.36.11 to 1.36.12. scripts/build.sh builds protoc-gen-go from the module version in go.mod (line 138) and regenerates the proto code before every image build, so the first build after that merge produced files that differ from what is committed.
Found during the local cluster cycle on #11, which was the first make up after #10 landed. #11 itself does not touch the protos.
Why CI did not catch it
The checks workflow does not run scripts/build.sh, which needs podman, so no job compares regenerated output with the committed files.
Fix
Regenerate and commit the two files. Ideally this rides in the Dependabot pull request that bumps the module, so the generated code and the generator move together.
Every future bump of google.golang.org/protobuf or google.golang.org/grpc/cmd/protoc-gen-go-grpc will do the same. A CI step that regenerates the protos with the pinned protoc and fails on git diff --exit-code would catch it on the Dependabot pull request itself; protoc is already pinned by checksum in scripts/build.sh, and the generation step does not need podman.
Activity