test(mcp): ask for JSON where the coverage test asserts a JSON field - #2248
Merged
Merged
Conversation
#1787 landed correct production code and a test that cannot pass on today's main. The test is the only thing wrong, and only in how it asks. check_index_coverage answers in the compact table by default, where a field name never appears: main.go main.go coverage_unavailable metadata_match read_source_and_reindex [] The test asserts on the JSON field spelling: strstr(response, "\"freshness\":\"metadata_match\"") Both halves are right; they were simply written against different response formats. The branch was cut on 2026-09-02, before that default changed, and was never rebased -- its CI green is from a main that no longer exists. The neighbouring coverage tests already pass "format":"json" for exactly this reason; this one predates them. So the request gains format=json and the assertions are untouched. Asserting the field rather than a bare strstr for "metadata_match" is what keeps the check exact: the value also appears in the table's recommended_action column and would be matched by any other path's row. Verified: build/c/test-runner mcp -> 318 passed, 0 failed, 4 skipped. Before this change the same suite reported 317 passed, 1 failed on identical production code. What this is NOT: a production defect. coverage_path_freshness on main uses cbm_path_info_utf8, exactly as #1787 intended, and the Windows second-vs- nanosecond truncation it fixes is gone. I checked that by reading origin/main rather than the local checkout, which was stale at 59a05eb and still carried the old helper -- that misreading briefly convinced me the merge had dropped the fix, and it had not. Signed-off-by: Martin Vogel <martin.vogel.tech@gmail.com>
This was referenced Sep 20, 2026
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Restores
mainto a fully greenmcpsuite. One test, one request parameter, no assertion changed.What broke
#1787 merged correct production code together with a test that cannot pass on today's
main.check_index_coverageanswers in the compact table by default, where a field name never appears:The test asserts on the JSON field spelling:
Both halves are correct — they were written against different response formats. That branch was cut on 2026-09-02, before the default changed, and never rebased. The neighbouring coverage tests already pass
"format":"json"for exactly this reason; this one predates them.The fix
The request gains
format=json; the assertions are untouched. Asserting the JSON field rather than a barestrstrformetadata_matchis what keeps it exact — that value also appears in the table'srecommended_actioncolumn and would be satisfied by any other path's row.Verification
build/c/test-runner mcp→ 318 passed, 0 failed, 4 skipped.metadata_changed) executes and passes, so the test still exercises the nanosecond-exactness it was written to pin.What this is not
Not a production defect.
coverage_path_freshnessonmainusescbm_path_info_utf8exactly as #1787 intended, and the Windows second-vs-nanosecond truncation it fixes is gone.How it reached main
I merged #1787 on a CI green that was 18 days old, without re-running it against the current base. The contract calls green perishable; I applied that reasoning to another PR the same day and skipped it here. Nothing about @umi008's work was at fault — it passes on its own head.