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
(Optional) Language server protocol tests without VS Code #135
Why: when an integration test fails, it isn't clear whether the server or the client broke. Talking to the server over LSP directly separates the two, runs faster, and needs no display. It overlaps with #131, so only do this if integration failures turn out hard to diagnose.
What: start the server JAR, connect with vscode-jsonrpc, send initialize, didOpen and liquidjava/verify, then check the liquidjava/diagnostics notification for the passing and failing fixtures.
I recommend keeping this optional and deferring it for now. The integration tests added in #144–#146 already exercise initialization, opening a document, Verify, and explicit passing/failing diagnostic notifications through the bundled extension in real VS Code.
A direct JSON-RPC suite would mainly help isolate server failures from client failures and run without an editor or display. That is useful, but adds another process-launch, connection, event-waiting, and cleanup harness to maintain. The failures encountered while implementing the current suites were diagnosable using logs and targeted inspection.
Revisit this if integration failures repeatedly leave us unsure whether the server or client broke, as the issue already proposes.
Activity
rcosta358 commentedon Oct 2, 2026
I recommend keeping this optional and deferring it for now. The integration tests added in #144–#146 already exercise initialization, opening a document, Verify, and explicit passing/failing diagnostic notifications through the bundled extension in real VS Code.
A direct JSON-RPC suite would mainly help isolate server failures from client failures and run without an editor or display. That is useful, but adds another process-launch, connection, event-waiting, and cleanup harness to maintain. The failures encountered while implementing the current suites were diagnosable using logs and targeted inspection.
Revisit this if integration failures repeatedly leave us unsure whether the server or client broke, as the issue already proposes.
Generated by Codex.