ROB: Tolerate streams without length field (#1717)

The field /Length is normally required, but Acrobat Reader and other readers are tolerant

Closes #1715

pypdf/_utils.py

@@ -178,10 +178,10 @@ def read_until_regex(stream: StreamType, regex: Pattern[bytes]) -> bytes:
         tok = stream.read(16)
         if not tok:
             return name
-        m = regex.search(tok)
+        m = regex.search(name + tok)
         if m is not None:
-            name += tok[: m.start()]
-            stream.seek(m.start() - len(tok), 1)
+            stream.seek(m.start() - (len(name) + len(tok)), 1)
+            name = (name + tok)[: m.start()]
             break
         name += tok
     return name

pypdf/generic/_data_structures.py

@@ -447,15 +447,26 @@ def read_unsized_from_steam(
                 stream.seek(-1, 1)
             # this is a stream object, not a dictionary
             if SA.LENGTH not in data:
-                raise PdfStreamError("Stream length not defined")
+                if pdf is not None and pdf.strict:
+                    raise PdfStreamError("Stream length not defined")
+                else:
+                    logger_warning(
+                        f"Stream length not defined @pos={stream.tell()}", __name__
+                    )
+                data[NameObject(SA.LENGTH)] = NumberObject(-1)
             length = data[SA.LENGTH]
             if isinstance(length, IndirectObject):
                 t = stream.tell()
                 assert pdf is not None  # hint for mypy
                 length = pdf.get_object(length)
                 stream.seek(t, 0)
             pstart = stream.tell()
-            data["__streamdata__"] = stream.read(length)
+            if length > 0:
+                data["__streamdata__"] = stream.read(length)
+            else:
+                data["__streamdata__"] = read_until_regex(
+                    stream, re.compile(b"endstream")
+                )
             e = read_non_whitespace(stream)
             ndstream = stream.read(8)
             if (e + ndstream) != b"endstream":

tests/test_generic.py

@@ -380,17 +380,22 @@ def test_dictionaryobject_read_from_stream_stream_no_newline():
 
 
 @pytest.mark.parametrize(("strict"), [(True), (False)])
-def test_dictionaryobject_read_from_stream_stream_no_stream_length(strict):
-    stream = BytesIO(b"<< /S /GoTo >>stream\n")
+def test_dictionaryobject_read_from_stream_stream_no_stream_length(strict, caplog):
+    stream = BytesIO(b"<< /S /GoTo >>stream\n123456789endstream abcd")
 
     class Tst:  # to replace pdf
         strict = False
 
     pdf = Tst()
     pdf.strict = strict
-    with pytest.raises(PdfReadError) as exc:
-        DictionaryObject.read_from_stream(stream, pdf)
-    assert exc.value.args[0] == "Stream length not defined"
+    if strict:
+        with pytest.raises(PdfReadError) as exc:
+            DictionaryObject.read_from_stream(stream, pdf)
+        assert exc.value.args[0] == "Stream length not defined"
+    else:
+        o = DictionaryObject.read_from_stream(stream, pdf)
+        assert "Stream length not defined" in caplog.text
+        assert o.get_data() == b"123456789"
 
 
 @pytest.mark.parametrize(