Enter DDL wait loop before canceling RPC test

[apstndb]

Refs #941. Follow-up to #943.

Summary

Test-only change in internal/mycli/execute_ddl_rpc_test.go. After the fake accepts UpdateDatabaseDdl and serves the first GetOperation, the wait-loop test waits until executeDdlStatements has left op.Poll and blocked in the wait-loop select before canceling. Canceling at handler return was racing the in-flight poll and skipping the wait-loop ctx.Done() branch.

No production code changes. .octocov.yml remains acceptable: 80%.

HEAD 27ba1b6c5e87b44824a1df21bc6064ab38515fea includes additive merges of origin/main through #944 (a54dbdfe34b581679cb42b84af4b5abfb6d6b0f9). Exclusive file vs main remains execute_ddl_rpc_test.go.

Hosted coverage (pre-#944 merge head)

• Head: bb0ce44e4a1876891c1d7e37ea5ba1f1657e5cac
• CI run: https://github.com/apstndb/spanner-mycli/actions/runs/34705647574 (success)
• Job: test (103585083339)
• Step: run tests with coverage (make test-coverage)
• go tool cover -func total: 84.8%
• Coverage HTML artifact: coverage-report id 10301329604
• octocov (comment) on merge SHA 8fe38b37a47942c9063b89dedba90c0f8a39ec10: 85.1% vs then-main 873d8ee 85.1% (+0.0%, +3 covered, 7853 → 7856 / 9223)
• PR-scope: execute_ddl.go 80.7% (+2.6%; restores the 78.0% wait-loop miss on merged Test DDL execution and EXPLAIN/DESCRIBE paths #943)

Hosted CI/octocov for current HEAD 27ba1b6 (after additive #944 merge) is pending.

Remaining gaps in this lane

• executeDdlStatements ticker / second poll (5s ticker; no production seam)
• executeExplainAnalyzeDML remainder
• writePlanExportFile remaining I/O error branches
• ExplainAnalyzeDmlStatement.isMutationStatement marker method

  

[github-actions]

Code Metrics Report

📊 View detailed coverage report (available for 7 days)

	main (873d8ee)	#947 (707fa3e)	+/-
Coverage	85.1%	85.8%	+0.7%
Code to Test Ratio	1:2.1	1:2.1	+0.0
Test Execution Time	1m38s	1m38s	0s

Details

  |                     | main (873d8ee) | #947 (707fa3e) |  +/-  |
  |---------------------|----------------|----------------|-------|
+ | Coverage            |          85.1% |          85.8% | +0.7% |
  |   Files             |            105 |            105 |     0 |
  |   Lines             |           9223 |           9223 |     0 |
+ |   Covered           |           7853 |           7920 |   +67 |
+ | Code to Test Ratio  |          1:2.1 |          1:2.1 |  +0.0 |
  |   Code              |          20926 |          20926 |     0 |
+ |   Test              |          43998 |          44846 |  +848 |
  | Test Execution Time |          1m38s |          1m38s |    0s |

Code coverage of files in pull request scope (73.0% → 82.4%)

Files	Coverage	+/-	Status
internal/mycli/execute_ddl.go	80.7%	+2.6%	affected
internal/mycli/execute_sql.go	94.7%	+13.0%	affected
internal/mycli/formatter_utils.go	93.7%	+31.2%	affected
internal/mycli/row_processor.go	95.4%	+70.4%	affected
internal/mycli/statements.go	75.0%	+0.7%	affected
internal/mycli/statements_transaction.go	70.2%	+1.3%	affected

---

Reported by octocov

[apstndb]

Reviewed bb0ce44e4a1876891c1d7e37ea5ba1f1657e5cac.

P2, source-backed: internal/mycli/execute_ddl_rpc_test.go:220 says this waits until the client has left op.Poll, but the new time.After(100 * time.Millisecond) provides no synchronization with that client-side event. A slow client can still be in the poll after 100ms; a fast client merely pays an arbitrary delay. This makes the claimed wait-loop coverage scheduler-dependent and does not establish the condition described by the comment/PR. No runtime failure is claimed to be reproduced.

The merged test already verifies cancellation after operation acceptance with explicit synchronization. Please do not add a delay solely to raise coverage of a particular select arm. Keep this follow-up draft and pause implementation while the integrated #941 result is verified; if the target is reached, retain this as a deferred candidate rather than creating additional coverage work. A future proposal should use a deterministic existing observation point or focus on the externally observable cancellation contract.