[Fix] lua_content: Move PDF ligature substitutions from string unescape to text handler

StandardEncoding/MacRomanEncoding ligature substitutions (e.g. byte 0xAD -> 'ffl')
were applied to all PDF strings including /URI annotation values. This corrupted
soft hyphens (U+00AD) in URLs, preventing the URL parser from detecting zero-width
space obfuscation and setting the ZW_SPACES flag.

Move ligature substitutions to text_op_handler where they belong, so they only
apply to rendered text content (Tj/TJ operators), not to dictionary string values.

diff --git a/lualib/lua_content/pdf.lua b/lualib/lua_content/pdf.lua
index 4c37b06385fd72109e1091c118cf75f967c1df82..def48892074474aad4024020f5ddbc4bab4b9598 100644 (file)
--- a/lualib/lua_content/pdf.lua
+++ b/lualib/lua_content/pdf.lua
@@ -160,6 +160,21 @@ local function compile_tries()
   end
 end
 
+-- StandardEncoding/MacRomanEncoding ligature substitutions
+-- Applied only to rendered text, NOT to dictionary strings (URI values etc.)
+-- to avoid corrupting soft hyphens (U+00AD = byte 0xAD = \173) in URLs
+local function apply_ligature_substitutions(s)
+  if not s then return s end
+  s = s:gsub('\171', 'ff')
+  s = s:gsub('\172', 'ffi')
+  s = s:gsub('\173', 'ffl')
+  s = s:gsub('\174', 'fi')
+  s = s:gsub('\175', 'fl')
+  s = s:gsub('\222', 'fi')
+  s = s:gsub('\223', 'fl')
+  return s
+end
+
 -- Returns a table with generic grammar elements for PDF
 local function generic_grammar_elts()
   local P = lpeg.P
@@ -183,17 +198,6 @@ local function generic_grammar_elts()
       res = lua_util.unhex(s:sub(1, #s - 1)) .. lua_util.unhex((s:sub(#s) .. '0'))
     end
 
-    if res then
-      -- StandardEncoding/MacRomanEncoding ligature substitutions
-      res = res:gsub('\171', 'ff')
-      res = res:gsub('\172', 'ffi')
-      res = res:gsub('\173', 'ffl')
-      res = res:gsub('\174', 'fi')
-      res = res:gsub('\175', 'fl')
-      res = res:gsub('\222', 'fi')
-      res = res:gsub('\223', 'fl')
-    end
-
     return res
   end
 
@@ -216,15 +220,6 @@ local function generic_grammar_elts()
     end
     s = s:gsub('\\%d%d?%d?', ue_octal)
 
-    -- StandardEncoding/MacRomanEncoding ligature substitutions
-    s = s:gsub('\171', 'ff')
-    s = s:gsub('\172', 'ffi')
-    s = s:gsub('\173', 'ffl')
-    s = s:gsub('\174', 'fi')
-    s = s:gsub('\175', 'fl')
-    s = s:gsub('\222', 'fi')
-    s = s:gsub('\223', 'fl')
-
     return s
   end
 
@@ -491,6 +486,11 @@ local function gen_text_grammar()
       res = table.concat(tres)
     end
 
+    -- Apply ligature substitutions for rendered text only (not dictionary strings like /URI)
+    if type(res) == 'string' then
+      res = apply_ligature_substitutions(res)
+    end
+
     res = sanitize_pdf_text(res)
 
     -- Apply text quality filtering to reject garbage chunks