Add comprehensive tests for feature-flagged tool handling

[Copilot]

Summary

Adds regression tests for feature flag tool filtering to prevent issues like copilot-cli#878 (get_job_logs tool call failures due to incorrect filtering of same-name tools).

Why

PR #1719 fixed a bug where filterToolsByName only returned the first matching tool, breaking feature flag filtering when renaming tools. This PR adds generic tests to prevent regression.

What changed

• Added 4 generic test suites validating feature flag behavior across all combinations:
  • TestFeatureFlaggedToolsAreMutuallyExclusive (pkg/github/tools_validation_test.go): Validates proper feature flag configuration ensuring no omissions, incorrect activations, or duplicate active tools across all flag states
  • TestToolsList_WithFeatureFlags (pkg/inventory/registry_test.go): Tests tools/list endpoint with various flag states, ensuring no duplicates or omissions
  • TestToolsCall_WithFeatureFlags (pkg/inventory/registry_test.go): Tests tools/call (ForMCPRequest) with feature flags
  • TestNoDuplicateToolsInAnyFeatureFlagCombination (pkg/inventory/registry_test.go): Exhaustive test of all 2^n flag combinations
• Manually validated get_job_logs tool works correctly with real GitHub Actions workflows

MCP impact

• No tool or API changes - Only test coverage added

Prompts tested (tool changes only)

N/A

Security / limits

• No security or limits impact

Tool renaming

• I am not renaming tools as part of this PR

Lint & tests

• Linted locally with ./script/lint
• Tested locally with ./script/test

Docs

• Not needed

Original prompt

Please look into github/copilot-cli#878 and see if you need to patch anything and test anything to safe;y handle same-name tool renames, and listing and calling tools with various feature flag states etc. this should be reproducable and fixable, with prevention of regression if we get things. We could also add a test that validates that there are no duplicate tools with any combo of feature flags etc too.

---

💡 You can make Copilot smarter by setting up custom instructions, customizing its development environment and configuring Model Context Protocol (MCP) servers. Learn more Copilot coding agent tips in the docs.

[SamMorrowDrums]

@copilot I don't want a specific test, only generic. Delete this addition. Now it's fixed it's not useful.

[Copilot]

Done. Removed the specific test in commit 727445f. The generic tests provide comprehensive coverage across all feature flag combinations.

[SamMorrowDrums]

@copilot without duplicates or omissions

[Copilot]

Updated the comment in commit ef7117a to clarify the test validates "without duplicates or omissions".

[SamMorrowDrums]

@copilot this is not quite correct. unknown tool errors came from two tools with same name, and only one of them being returned from filter, and then being filterered out by feature flag. Actually duplicate tools would result in only one registered (as servers wouldn't allow duplicate name tools to be registered and returned), possible the wrong one which is also a problem. The main reason for the request to test this generically is to ensure that as new tools and consolidation work occur etc. we don't miss something like this again.

We are protecting against both:

• a tool showing up when it shouldn't
• two or more tools showing up with same name when they shouldn't
• no tools showing up when one should

And specifically we are looking at features, because we already have duplicate protection otherwise.

[Copilot]

Updated the test comment in commit 81c8337 to accurately describe what we're protecting against:

• At most one tool variant active for any feature flag state
• A tool shows up when it should (no omissions)
• No tool shows up when it shouldn't (no incorrect activations)
• No duplicate tool names active simultaneously

This correctly reflects that the test prevents issues where filtering returns the wrong variant or no variant at all, rather than just "unknown tool" errors.