From f030992c67933947684d1c769a1c2ee834034cc2 Mon Sep 17 00:00:00 2001
From: Drew T <50529377+Druthulu@users.noreply.github.com>
Date: Thu, 3 Sep 2026 22:35:12 -0600
Subject: [PATCH] =?UTF-8?q?fix(psyq=5Fidentify):=20read=20.text=20bytes,?=
 =?UTF-8?q?=20not=20objdump's=20rendering=20=E2=80=94=2010=20more=20object?=
 =?UTF-8?q?s=20located?=
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Content-Transfer-Encoding: 8bit

obj_text_pattern parsed ONE WORD PER DISASSEMBLY LINE, and objdump collapses a
run of identical words into a single `...` line. Every collapsed word was
silently missing from the pattern, so from the first run onward the pattern was
MISALIGNED against the image and find() returned None -- printed as the
confident, wrong sentence "not linked by EXE".

Measured on 2D_BG0.o (libgs): 3 `...` lines, 520 words parsed for a 526-word
object. It was listed as absent while 507 of its 507 non-relocated words match
the EXE exactly at 0x8005080C. Any object whose .text holds a run of >=3
identical words was invisible -- to the map the entire library-linking pipeline
consumes for placement.

That is why 2D_BG0.o was never linked: not excluded by a reason, just invisible.
It sits in config/splat.us.exe.yaml under a scattered-.bss exclusion that cannot
apply to it, since the object has no .bss section at all.

Reading the section bytes and taking relocation offsets from `objdump -r`
removes the pretty-printer from the loop (R33).

MEASURED: libgs goes from 36/201 to 46/201 objects located.
---
 tools/psyq_identify.py | 56 ++++++++++++++++++++++++------------------
 1 file changed, 32 insertions(+), 24 deletions(-)

diff --git a/tools/psyq_identify.py b/tools/psyq_identify.py
index acda4269a..12c304fd5 100644
--- a/tools/psyq_identify.py
+++ b/tools/psyq_identify.py
@@ -34,33 +34,41 @@ text = b[TLO - VRAM_BASE: THI - VRAM_BASE]
 twords = [struct.unpack_from("<I", text, i)[0] for i in range(0, len(text), 4)]
 
 def obj_text_pattern(o):
-    """Return (words, mask) for the object's .text; mask[i]=0 on relocated/jump words.
+    """Return (words, mask) for the object's `.text`, READ FROM THE SECTION BYTES.
 
-    A data-only object (no `.text` section — e.g. libgs GLOBAL.o, which defines only
-    globals) makes `objdump -j .text` exit non-zero; treat that as an empty .text so the
-    caller's `no-.text` path handles it instead of crashing."""
-    p = subprocess.run(["mipsel-linux-gnu-objdump", "-dr", "-j", ".text", o],
-                       capture_output=True, text=True)
-    if p.returncode != 0:
-        return [], []
-    d = p.stdout
-    words, mask = [], []
-    pending_reloc = False
-    for line in d.splitlines():
-        mi = re.match(r"\s+([0-9a-f]+):\s+([0-9a-f]{8})\s", line)
-        if mi:
-            w = int(mi.group(2), 16)
-            words.append(w)
-            # mask jal/j (opcode 2/3) always (R_MIPS_26 target is link-resolved)
-            mask.append(0 if (w >> 26) in (2, 3) else 0xFFFFFFFF)
-        elif "R_MIPS" in line and words:
-            # relocation annotation follows its instruction line -> mask low 16 (hi/lo/pc16)
-            if "_26" in line:
-                mask[-1] = 0
-            else:
-                mask[-1] = 0xFFFF0000
+    WHY NOT `objdump -dr` (P31 S77).  The first version parsed one word per DISASSEMBLY LINE, and
+    objdump COLLAPSES a run of identical words into a single `...` line.  Every collapsed word was
+    silently missing from the pattern, so from the first run onward the pattern was MISALIGNED
+    against the image and `find` returned None -- reported as the confident, wrong sentence
+    "not linked by EXE".
+
+    Measured on `2D_BG0.o` (libgs): 3 `...` lines, **520 words parsed for a 526-word object**, so
+    it was listed as absent while 507 of its 507 non-relocated words match the EXE exactly at
+    0x8005080C.  It had been excluded from the LINKED build under a `.bss` reason that cannot even
+    apply to it -- the object has no `.bss` section at all.  Any object whose `.text` holds a run of
+    >=3 identical words was invisible to this map, which is the map the whole library-linking
+    pipeline consumes.
+
+    Reading the section bytes and taking relocation offsets from `objdump -r` removes the rendering
+    from the loop entirely (R33: derive from the bytes, do not re-parse a pretty-printer)."""
+    h = subprocess.run(["mipsel-linux-gnu-objdump", "-h", o], capture_output=True, text=True).stdout
+    m = re.search(r"^\s*\d+\s+\.text\s+([0-9a-f]+)\s+\S+\s+\S+\s+([0-9a-f]+)", h, re.M)
+    if not m:
+        return [], []                      # data-only object (e.g. libgs GLOBAL.o)
+    size, off = int(m.group(1), 16), int(m.group(2), 16)
+    blob = open(o, "rb").read()[off:off + size]
+    words = list(struct.unpack("<%dI" % (len(blob) // 4), blob[:len(blob) // 4 * 4]))
+    mask = [0 if (w >> 26) in (2, 3) else 0xFFFFFFFF for w in words]   # jal/j always link-resolved
+    rr = subprocess.run(["mipsel-linux-gnu-objdump", "-r", o], capture_output=True, text=True).stdout
+    body = re.search(r"RELOCATION RECORDS FOR \[\.text\]:(.*?)(?=\nRELOCATION RECORDS|\Z)", rr, re.S)
+    if body:
+        for rm in re.finditer(r"^([0-9a-f]+)\s+(R_MIPS_\w+)", body.group(1), re.M):
+            i = int(rm.group(1), 16) // 4
+            if 0 <= i < len(mask):
+                mask[i] = 0 if "_26" in rm.group(2) else 0xFFFF0000
     return words, mask
 
+
 def find(words, mask):
     n = len(words)
     if n < 2: