feat: split profiled and source-backed counts

scripts/generate_readme.py

@@ -292,10 +292,13 @@ def generate_readme(db: dict, platforms_dir: str) -> str:
             "",
             "Coverage is measured against the file list each platform declares,"
             " using that platform's own verification mode.",
-            "Source-backed counts the files whose expectations were also read"
-            " from an emulator's source code; the rest rely on the platform"
-            " list alone. A dash means no profiled emulator applies to the"
-            " platform, whose own source is then the only authority.",
+            "Source-backed counts the files whose content the emulator's own"
+            " code checks: a size or hash read from its source, reproduced at"
+            " verification. Files a profile documents without a content check"
+            " are counted separately on the"
+            f" [gap analysis]({SITE_URL}gaps/) page. A dash means no profiled"
+            " emulator applies to the platform, whose own source is then the"
+            " only authority.",
             "Where platform lists and emulator source code disagree, the"
             f" differences are tracked in the [gap analysis]({SITE_URL}gaps/).",
             "",

scripts/generate_site.py

@@ -1522,8 +1522,8 @@ def generate_gap_analysis(
     lines.extend([
         "## Verification by Platform",
         "",
-        "| Platform | Files | Verified | Untested | Missing | Mode | Source-backed |",
-        "|----------|------:|---------:|---------:|--------:|------|--------------:|",
+        "| Platform | Files | Verified | Untested | Missing | Mode | Profiled | Source-backed |",
+        "|----------|------:|---------:|---------:|--------:|------|---------:|--------------:|",
     ])
 
     for pname, cov in sorted(coverages.items(), key=lambda x: x[1]["platform"]):
@@ -1542,12 +1542,14 @@ def generate_gap_analysis(
         )
         gt = cov["ground_truth"]
         if not gt.get("applicable", True):
-            gt_cell = "-"
+            gt_cell = prof_cell = "-"
         elif gt["total"]:
             gt_pct = f"{gt['with_validation'] / gt['total'] * 100:.0f}%"
             gt_cell = f"{gt['with_validation']}/{gt['total']} ({gt_pct})"
+            prof_pct = f"{gt.get('with_profile', 0) / gt['total'] * 100:.0f}%"
+            prof_cell = f"{gt.get('with_profile', 0)}/{gt['total']} ({prof_pct})"
         else:
-            gt_cell = "0/0"
+            gt_cell = prof_cell = "0/0"
         lines.append(
             f"| [{display}](platforms/{pname}.md) "
             f"| {cov['total']} "
@@ -1555,6 +1557,7 @@ def generate_gap_analysis(
             f"| {untested_str} "
             f"| {missing_str} "
             f"| {cov['mode']} "
+            f"| {prof_cell} "
             f"| {gt_cell} |"
         )
     lines.extend([
@@ -1562,10 +1565,13 @@ def generate_gap_analysis(
         "Verification follows each platform's own runtime check "
         "([how each mode works](wiki/verification-modes.md)): the counts "
         "measure the repository against the file list each platform declares. "
-        "Source-backed counts the files whose expectations were also read "
-        "from an emulator's source code; the rest rely on the platform list "
-        "alone. A dash means no profiled emulator applies to the platform, "
-        "whose own source is then the only authority.",
+        "Profiled counts the files documented in an emulator profile (name or "
+        "alias read from source). Source-backed is stricter: the emulator's "
+        "own code checks the file's content (a size or hash read from its "
+        "source, reproduced at verification). A file can be profiled without "
+        "being source-backed when the code loads it but never checks it. "
+        "A dash means no profiled emulator applies to the platform, whose "
+        "own source is then the only authority.",
         "",
     ])
 

scripts/verify.py

@@ -835,6 +835,15 @@ def verify_platform(
 
     # Ground truth coverage
     gt_filenames = set(validation_index)
+    profiled_names: set[str] = set()
+    for profile in profiles.values():
+        if profile.get("type") in ("launcher", "alias"):
+            continue
+        for f in profile.get("files", []):
+            name = f.get("name", "")
+            if name:
+                profiled_names.add(name)
+            profiled_names.update(f.get("aliases") or [])
     dest_to_name: dict[str, str] = {}
     for sys_id, system in verify_systems.items():
         for fe in system.get("files", []):
@@ -845,6 +854,9 @@ def verify_platform(
     with_validation = sum(
         1 for dest in file_status if dest_to_name.get(dest, "") in gt_filenames
     )
+    with_profile = sum(
+        1 for dest in file_status if dest_to_name.get(dest, "") in profiled_names
+    )
     total = len(file_status)
 
     return {
@@ -858,6 +870,7 @@ def verify_platform(
         "details": details,
         "ground_truth_coverage": {
             "with_validation": with_validation,
+            "with_profile": with_profile,
             "platform_only": total - with_validation,
             "total": total,
             "applicable": bool(resolve_platform_cores(config, profiles)),