[Test, CI] (tools/trace_replay, workflows): give retrace-split the negative control it was cloned without, match ConfigLoader's whole INFO sentence rather than a KEY=VALUE an env dump also prints, read the three-arm names file the identity step was writing and discarding, and refuse an unknown manifest key that silently empties a CI matrix

tools/trace_replay/run_trace_case.cmake

@@ -252,17 +252,41 @@ if(DEFINED ENV{MOBILEGL_TRANSPORT} AND NOT "$ENV{MOBILEGL_TRANSPORT}" STREQUAL "
                 "retrace with no library log cannot be counted as a split retrace.")
     else()
         file(READ "${mobilegl_log}" split_log)
-        string(FIND "${split_log}" "MOBILEGL_TRANSPORT=inproc" split_armed_at)
+        # THE DISTINCTIVE PART OF ConfigLoader's INFO LINE, not the bare KEY=VALUE - review finding
+        # M-5. ConfigLoader.cpp:71 logs `Config: Accepted env variable: %s=%s` for EVERY MOBILEGL_*
+        # variable, unconditionally, in every build including the pull one. That line is MGLOG_D,
+        # so at the INFO level CI and the gate use it is compiled out - but at
+        # MOBILEGL_LOG_ACTIVE_LEVEL=..._DEBUG it reads `Config: Accepted env variable:
+        # MOBILEGL_TRANSPORT=inproc` and satisfied the old substring search. Observed GREEN on a
+        # crafted pull-build log. The person most likely to hit that is the one who rebuilds at
+        # DEBUG to debug a split failure. The sentence below exists only in
+        # ConfigLoader::InitTransport's InProcess arm, which exists only under
+        # MOBILEGL_BUILD_DISAGGREGATED.
+        set(split_expected_marker "MOBILEGL_TRANSPORT=inproc - the MGPipe record stream")
+        string(FIND "${split_log}" "${split_expected_marker}" split_armed_at)
         if(split_armed_at EQUAL -1)
+            # Say which transport was actually asked for. ConfigLoader REFUSES spawn / unix: /
+            # pipe: BY NAME and stays on monolith (they are P6's), so a run that set one of those
+            # has a different diagnosis from one that set inproc against a monolith library, and
+            # the old message named `inproc` either way.
+            set(split_refusal "")
+            if(NOT "$ENV{MOBILEGL_TRANSPORT}" STREQUAL "inproc")
+                set(split_refusal
+                        " NOTE: this run asked for '$ENV{MOBILEGL_TRANSPORT}', which P5 does not "
+                        "implement - ConfigLoader recognises spawn / unix: / pipe: and REFUSES them by "
+                        "name, staying on monolith. Only 'inproc' can resolve in P5.")
+            endif()
             message(FATAL_ERROR
                     "MOBILEGL_TRANSPORT=$ENV{MOBILEGL_TRANSPORT} is set for ${split_case} and the library "
-                    "never reported resolving it. ConfigLoader::InitTransport logs one line at INFO when "
-                    "it selects InProcess, and that line exists only in a build configured with "
+                    "never reported resolving it: ${mobilegl_log} carries no "
+                    "\"${split_expected_marker}\". ConfigLoader::InitTransport logs that line at INFO "
+                    "when it selects InProcess, and it exists only in a build configured with "
                     "-DMOBILEGL_BUILD_DISAGGREGATED=ON - in a build without it the whole parser is "
                     "compiled out and the variable is accepted and ignored, which is exactly the 'the "
                     "split lane ran monolith and went green' failure. Check that the SPLIT runtime "
                     "artifact is the one at ${MOBILEGL_LIBRARY}, and that MOBILEGL_LOG_ACTIVE_LEVEL "
-                    "admits INFO.")
+                    "admits INFO (at WARN or above the line is compiled out and this reds for no "
+                    "defect).${split_refusal}")
         endif()
         # The refusal census. Recorded on every split run, pass or fail.
         file(STRINGS "${mobilegl_log}" split_fatals REGEX "Fatal\\{")

tools/trace_replay/trace_cases.py

@@ -8,11 +8,48 @@ from pathlib import Path
 TRACE_CASES_JSON = Path(__file__).with_name("trace_cases.json")
 CI_BACKENDS = ("DirectGLES", "DirectVulkan")
 
+# Every key a case or the defaults block may carry. An UNKNOWN key is a hard error rather than a
+# silent no-op, which is review finding N-4: `"split": true` mistyped as `"splitt": true` loaded
+# clean, the split subset became [], the GitHub matrix became {"include":[]}, and `retrace-split`
+# was skipped with no red anywhere. Every other way of getting `split` wrong already raised
+# (`"ci": false`, a backend list without DirectGLES, a non-bool value) - the typo was the one hole,
+# and it is the shape that makes a whole CI job quietly stop existing.
+#
+# Adding a key means adding it here, deliberately, in the same commit. That is the point.
+KNOWN_CASE_KEYS = frozenset({
+    "name",
+    "trace_archive",
+    "trace_file",
+    "golden",
+    "alternate_golden",
+    "target_call",
+    "width",
+    "height",
+    "ssim_threshold",
+    "crop_x",
+    "crop_y",
+    "crop_width",
+    "crop_height",
+    "coherent_as_flush",
+    "timeout_seconds",
+    "ci",
+    "ci_backends",
+    "verify",
+    "split",
+    "avoid_angle_llvmpipe_explicit_lod_bias",
+})
+
 
 def load_trace_case_manifest(path=TRACE_CASES_JSON):
     with Path(path).open("r", encoding="utf-8") as file:
         manifest = json.load(file)
     defaults = manifest.get("defaults", {})
+    unknown_defaults = sorted(set(defaults) - KNOWN_CASE_KEYS)
+    if unknown_defaults:
+        raise ValueError(
+            f"unknown key(s) in the defaults block: {', '.join(unknown_defaults)}. "
+            f"Known keys are {', '.join(sorted(KNOWN_CASE_KEYS))}"
+        )
     cases = []
     seen = set()
     for case in manifest.get("cases", []):
@@ -20,6 +57,13 @@ def load_trace_case_manifest(path=TRACE_CASES_JSON):
         name = merged.get("name")
         if not name:
             raise ValueError("trace case is missing name")
+        unknown = sorted(set(case) - KNOWN_CASE_KEYS)
+        if unknown:
+            raise ValueError(
+                f"unknown key(s) for {name}: {', '.join(unknown)}. A mistyped flag loads clean and "
+                f"turns its whole CI subset into an empty matrix, which GitHub skips with no red. "
+                f"Known keys are {', '.join(sorted(KNOWN_CASE_KEYS))}"
+            )
         if name in seen:
             raise ValueError(f"duplicate trace case: {name}")
         seen.add(name)