gate: fail loudly when python_gate copy can't be made

SonicField · SonicField · commit f3271f58a560 · 2026-04-21T17:43:52.000-07:00
Pushes 36, 37, 38 all gated against the stale python_gate from push 33 (8382896), not their respective HEAD binaries. Root cause: 'cp "$PYTHON" "${PYTHON}_gate" 2>/dev/null || true'. When a prior gate left python_gate held open (or otherwise busy), cp failed with 'Text file busy', stderr was discarded, '|| true' swallowed the exit code, and PYTHON pointed to the leftover stale copy. Subsequent test runs reported 981 PASS — against a binary from 3 commits ago. Fix: rm -f the destination first, then cp without || true. If cp fails for any reason, set -e aborts the gate. No more silent fallback to a stale binary. Discovery: testkeeper investigated SIGSEGV during G1.6 LIR-capture prep on the actual HEAD binary (carrying generalist's in-flight LOAD_ATTR_SLOT stash). The crash didn't reproduce on python_gate because python_gate was 3 commits old and predated the SLOT work. The version-string mismatch ('79890e7b73-dirty' vs '8382896c85') made the stale-binary substitution visible. Also adds catch python#4 to docs/wiring_catches.md per supervisor 2026-04-22 00:42:38Z directive.
diff --git a/docs/wiring_catches.md b/docs/wiring_catches.md
@@ -16,6 +16,7 @@ caught it> <root cause class> <test-bug | emit-bug | gate-script bug>`.
 | 1 | 2026-04-21 | 0e62d9108b   | wiring smoke `load_attr_generic`       | `import sys` in body → IMPORT_NAME → JIT PYJIT_RESULT_UNKNOWN_ERROR at force_compile                | test-bug         |
 | 2 | 2026-04-21 | 15feea07c9   | ARM64 gate (push 36 stash bracket)     | `git stash push -u` captured SCP'd bundle inside cpython tree → fetch failed silently               | gate-script bug  |
 | 3 | 2026-04-21 | 91d5b60f5d   | ARM64 commit-match check               | `grep -oP 'ARM64_COMMIT=\K\S+'` matched heredoc body literal `$(git rev-parse...)` not runtime echo | gate-script bug  |
+| 4 | 2026-04-22 | 79890e7b73   | testkeeper G1.6 prep investigation     | `cp "$PYTHON" "${PYTHON}_gate" \|\| true` silently tolerated 'Text file busy' → gate ran against stale `python_gate` (8382896c85) for pushes 36/37/38; masked emitGetIter validation AND latent LOAD_ATTR_SLOT segfault in generalist's in-flight stash | gate-script bug  |
 
 ## Meta-observation (supervisor 2026-04-22 00:17:13Z)
 
diff --git a/scripts/gate_phoenix.sh b/scripts/gate_phoenix.sh
@@ -73,8 +73,15 @@ if ! "$SCRIPT_DIR/build_phoenix.sh" $BUILD_FLAGS; then
 fi
 echo "Build: PASS" | tee -a "$RESULTS_FILE"
 
-# Copy binary to avoid "Text file busy" if another process holds the original
-cp "$PYTHON" "${PYTHON}_gate" 2>/dev/null || true
+# Copy binary to avoid "Text file busy" if another process holds the original.
+# rm before cp so a busy ${PYTHON}_gate from a prior run cannot leave a
+# stale binary in place — that masquerades as PASS while the gate actually
+# tests the OLD binary. Pre-2026-04-22 the cp was '|| true' which silently
+# tolerated 'Text file busy' and ran the gate against whatever stale
+# ${PYTHON}_gate happened to already exist (caught at 00:41Z, push 38).
+# Now: rm forces a fresh copy; cp failure aborts the gate.
+rm -f "${PYTHON}_gate"
+cp "$PYTHON" "${PYTHON}_gate"
 PYTHON="${PYTHON}_gate"
 
 # Step 1b: _reg usage policy gate (no-FS DeoptBase factories banned in simplify_c.c)
