BuildTestEmbedded's local kompile test harness recompiles the full 546-script workspace repeatedly

Reported (UTC): 2026-07-15 18:26

BuildTestEmbedded's local kompile test harness recompiles the full 546-script workspace repeatedly when many --test selectors are supplied in one scripts/test.bash invocation. A 31-scenario deletion regression selection stayed CPU-bound for 21 minutes, produced more than 134,000 lines of repeated buildMaven/PackageWarning diagnostics, and had not begun the selected runtime scenarios; a six-test selection performs one roughly 11-minute catalog compile, while the larger selection appeared to begin a fresh full-catalog pass per selector batch. Impact: the documented local targeted-test workflow becomes slower than the sharded full CI and is not a viable regression gate. The run was stopped after the six new scenarios had already been verified in smaller invocations. Durable fix: compile each test script once per source/build-rule artifact hash, reuse that compiled catalog across all selectors in the same invocation, and suppress handled BuildRuleExecutedEffect stack traces so final failures remain visible.

Production verification — 2026-07-17

Status: STILL EXISTS. Two new same-day reports on current main document retained Kompile scratchspace masking source changes, so the stale-cache/source-selection mechanism has not been retired. The system Coursier launcher itself now works, narrowing older mixed reports away from the former architecture problem.

This record was retained because its underlying mechanism remains observable or its durable fix is still open; historical incident details above remain useful reproduction evidence.