tools/lgdl.py: precheck <f_id> <edition_id> [pages] — mislinked-file gate (reverse-map editions + locator batch path + size sanity; exit 0/1/2); dl with 4th arg auto-prechecks and aborts on RED. Tested: caught 3 bad files (Одержимый магией / Laso / Captain Action comic); OK path verified

tools/lgdl.py

@@ -11,12 +11,22 @@ Pipeline (curl engine — Python's DNS is flaky on this box, curl is not):
 
 Usage:
     tools/lgdl.py ed   <edition_id> [more_ids...]
-    tools/lgdl.py dl   <file_id> <out_dir> <target_name>
+    tools/lgdl.py dl   <file_id> <out_dir> <target_name> [edition_id]
     tools/lgdl.py md5  <file_id>
+    tools/lgdl.py precheck <file_id> <edition_id> [expected_pages]
+
+precheck (no download, ~2 libgen calls): file record vs the target edition.
+  RED   (exit 2) — reverse-map `editions` non-empty and does NOT contain the
+        target edition (file belongs to another book), or locator path is a
+        foreign-language batch (spa/English/fiction) for a Cyrillic book.
+  YELLOW(exit 1) — reverse map empty, or size implausible vs expected pages.
+  OK    (exit 0) — reverse map matches target, or empty with plausible locator.
+`dl` with the optional 4th arg runs precheck first and aborts on RED.
 
 Verification on dl: final size must equal libgen filesize; magic-byte sniff
 (fb2/pdf/epub/doc) must not be HTML. Card+key are re-fetched up to 3x.
-Politeness: ~4s between libgen requests.
+Content-verify after dl (AGENTS.md lesson 8): pdfinfo pages vs RSL + first-page
+text must contain the RU title/author. Politeness: ~4s between libgen requests.
 """
 import json, os, re, subprocess, sys, time
 
@@ -164,6 +174,72 @@ def download(f_id, out_dir, name, tries=16):
     return None
 
 
+def precheck(f_id, e_id, expected_pages=None):
+    """Mislinked-file gate. Returns 0=OK, 1=SUSPECT, 2=BAD (see module doc)."""
+    fi = file_info(f_id)
+    time.sleep(2)
+    ext, size = fi.get("extension"), int(fi.get("filesize") or 0)
+    locator = fi.get("locator") or ""
+    rev = {eid: int(x.get("e_id") or 0) for eid, x in (fi.get("editions") or {}).items()}
+    rev = {k: v for k, v in rev.items() if v}
+    verdict, reasons = "OK", []
+
+    # 1) reverse map: who does this file BELONG to?
+    rev_match = None
+    if rev:
+        if int(e_id) in rev.values():
+            rev_match = True
+            reasons.append(f"reverse-map matches target e_id {e_id}")
+        else:
+            verdict = "BAD"
+            owners = []
+            for owner_eid in set(rev.values()):
+                try:
+                    oe = jget(f"json.php?object=e&addkeys=title,author,year&ids={owner_eid}")
+                    o = oe.get(str(owner_eid), {})
+                    owners.append(f"e_id {owner_eid} «{o.get('title')}» {o.get('author')} {o.get('year')}")
+                    time.sleep(2)
+                except Exception:
+                    owners.append(f"e_id {owner_eid} (?)")
+            reasons.append("reverse-map points to ANOTHER edition: " + "; ".join(owners))
+    else:
+        rev_match = False
+        verdict = "SUSPECT" if verdict == "OK" else verdict
+        reasons.append("reverse-map EMPTY (old file) — rely on locator + post-verify")
+
+    # 2) locator: foreign-language batch path. Only meaningful when the reverse
+    #    map does NOT confirm the target (matching map = the edition itself comes
+    #    from that batch — an origin note, not a mislink).
+    low = locator.lower()
+    foreign = re.search(r"(spa\\|spanish|english|epublibre|0day|\\fiction\\)", low)
+    cyrillic = re.search(r"[\u0400-\u04FF]", locator)
+    if locator and foreign and not cyrillic:
+        if rev_match is True:
+            reasons.append(f"origin note: edition comes from a foreign batch: {locator[:120]}")
+        else:
+            verdict = "BAD" if verdict == "OK" else verdict
+            reasons.append(f"locator is a foreign batch path: {locator[:120]}")
+    elif locator and rev_match is not True:
+        reasons.append(f"locator ok: {locator[:120]}")
+
+    # 3) size sanity vs expected pages (text formats are tiny: ~200B/pp;
+    #    scans: 50KB..15MB/pp)
+    if expected_pages and size:
+        per_lo = 200 if ext in ("fb2", "epub", "txt", "html") else 10_000
+        lo, hi = expected_pages * per_lo, expected_pages * 15_000_000
+        if size < lo:
+            verdict = "SUSPECT" if verdict == "OK" else verdict
+            reasons.append(f"size {size}b < {lo}b for ~{expected_pages} pp (too small — fragment?)")
+        elif size > hi:
+            verdict = "SUSPECT" if verdict == "OK" else verdict
+            reasons.append(f"size {size}b > {hi}b for ~{expected_pages} pp (too big — scan mismatch?)")
+
+    print(f"precheck f_id={f_id} vs e_id={e_id}: {verdict} [{ext}, {size}b]")
+    for r in reasons:
+        print(f"  - {r}")
+    return {"OK": 0, "SUSPECT": 1, "BAD": 2}[verdict]
+
+
 def main():
     if len(sys.argv) < 2:
         sys.exit(__doc__)
@@ -179,7 +255,19 @@ def main():
             time.sleep(1)
     elif cmd == "dl":
         f_id, out_dir, name = sys.argv[2], sys.argv[3], sys.argv[4]
+        if len(sys.argv) > 5 and sys.argv[5].isdigit():
+            rc = precheck(f_id, sys.argv[5])
+            if rc == 2:
+                print("  ABORT: precheck RED (mislinked file) — pick another f_id/edition or flib")
+                sys.exit(2)
+            if rc == 1:
+                print("  precheck SUSPECT — downloading, MUST post-verify content")
+            time.sleep(3)
         download(f_id, out_dir, name)
+    elif cmd == "precheck":
+        f_id, e_id = sys.argv[2], sys.argv[3]
+        pages = int(sys.argv[4]) if len(sys.argv) > 4 else None
+        sys.exit(precheck(f_id, e_id, pages))
     elif cmd == "md5":
         print(file_info(sys.argv[2])["md5"])
     else: