fix(task-14): seed AutoCAD's default scale list before adding an annotation scale headlessly

An ezdxf-authored drawing has no ACAD_SCALELIST entries, and
`_ensure_scale_entry` appended only the requested scale. AutoCAD repopulates
its default list only when the dictionary is *empty*, so the saved file
opened with a one-entry scale list and `SetVariable("CANNOSCALE", "1:1")`
failed with `Error setting system variable` -- the annotation scale could
no longer be changed in AutoCAD, while the tool reported `ok: True` and the
test pinned the degraded state (`len(keys) == 1`).

An empty list is now seeded with AutoCAD's own default metric list before
the requested scale is considered: the 17 entries a fresh AutoCAD 2026
drawing carries (`Documents.Add()` saved as DXF, MEASUREMENT 1), in that
order and under AutoCAD's own keys A0 ... A9, B0 ... B6. New entries take
the next key in that sequence (`_scale_list_key`) instead of being keyed by
their name; presence is still decided by group 300, so a non-empty
AutoCAD-authored list is never re-seeded or duplicated. Verified live: the
headless file now opens in AutoCAD 2026 with the 17 keys, CANNOSCALE 1:50
current, and switching to 1:1 and 1:100 works.

Tests: the fresh-drawing test asserts the seeded names and keys, that
setting 1:50 appends nothing (it is a default), that 1:25 lands under B7
once, and that the list survives save + reopen with 1:1 present; a new test
pins that *reading* the annotation scale creates nothing (the read side
stays pure, AutoCAD fills an empty list itself); the AutoCAD-keyed list test
now expects the A3 key. The `drawing_settings` docstring says what each
engine does and that AutoCAD refuses the live write on a paper-space layout
with no active viewport (measured).

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01FKFStjEFzpGyLq5bMpWxTd

backends/ezdxf_backend.py

@@ -5954,39 +5954,52 @@ class EzdxfBackend(AutoCADBackend):
                         return str(tag.value)
         return None
 
-    @classmethod
+    # AutoCAD's default metric ACAD_SCALELIST, in its own order (measured:
-    def _ensure_scale_entry(cls, doc, name: str, paper: float, drawing: float) -> bool:
+    # AutoCAD 2026, MEASUREMENT 1, `Documents.Add()` saved as DXF — 17 SCALE
-        """Add `name` to ACAD_SCALELIST unless present; True when added.
+    # objects keyed A0 … A9, B0 … B6). An imperial seat (MEASUREMENT 0) adds the
+    # architectural fractions on top; those are not seeded — every ratio here
+    # is valid in either unit system, and AutoCAD's SCALELISTEDIT can add more.
+    _DEFAULT_SCALE_LIST: tuple[str, ...] = (
+        "1:1",
+        "1:2",
+        "1:4",
+        "1:5",
+        "1:8",
+        "1:10",
+        "1:16",
+        "1:20",
+        "1:30",
+        "1:40",
+        "1:50",
+        "1:100",
+        "2:1",
+        "4:1",
+        "8:1",
+        "10:1",
+        "100:1",
+    )
 
-        Presence is decided by the SCALE objects' group-300 names, never by
+    @staticmethod
-        the dictionary key — an AutoCAD-authored drawing keys the list `A0`,
+    def _scale_list_key(index: int) -> str:
-        `A1`, … (its default metric list already carries 1:1 … 1:100, 2:1),
+        """The dictionary key AutoCAD gives the ``index``-th SCALE: A0 … A9, B0 …"""
-        and matching on the key appended a duplicate `1:50` to every one of
+        return f"{chr(ord('A') + index // 10)}{index % 10}"
-        them. A new entry is keyed by its name unless that key is taken, in
+
-        which case it takes the next free `A<n>` slot like AutoCAD does.
+    @classmethod
+    def _add_scale_object(cls, doc, scales, name: str, paper: float, drawing: float) -> None:
+        """Author one SCALE object and file it under the next free AutoCAD-style key.
 
         ezdxf has no SCALE entity class, so the object is authored as raw tags
         and loaded through the factory (a `DXFTagStorage` that exports
         verbatim). Measured: it round-trips, ezdxf registers the CLASS, and
-        `doc.audit()` is clean after reopen. AutoCAD only honours a CANNOSCALE
+        `doc.audit()` is clean after reopen.
-        that names an entry of this list.
         """
         from ezdxf.entities import factory
         from ezdxf.lldxf.extendedtags import ExtendedTags
 
-        scales = doc.rootdict.get("ACAD_SCALELIST")
-        if scales is None:
-            scales = doc.rootdict.add_new_dict("ACAD_SCALELIST")
         existing_keys = set(scales.keys())
-        for _key, entry in scales.items():
-            if cls._scale_entry_name(entry) == name:
-                return False
-        key = name
-        if key in existing_keys:
         index = 0
-            while f"A{index}" in existing_keys:
+        while cls._scale_list_key(index) in existing_keys:
             index += 1
-            key = f"A{index}"
         handle = doc.entitydb.next_handle()
         unit = 1 if paper == drawing else 0
         text = (
@@ -5996,7 +6009,38 @@ class EzdxfBackend(AutoCADBackend):
         entry = factory.load(ExtendedTags.from_text(text), doc)
         doc.entitydb.add(entry)
         doc.objects.add_object(entry)
-        scales.add(key, entry)
+        scales.add(cls._scale_list_key(index), entry)
+
+    @classmethod
+    def _ensure_scale_entry(cls, doc, name: str, paper: float, drawing: float) -> bool:
+        """Add `name` to ACAD_SCALELIST unless present; True when added.
+
+        Presence is decided by the SCALE objects' group-300 names, never by
+        the dictionary key — an AutoCAD-authored drawing keys the list `A0`,
+        `A1`, … and only group 300 says `1:50`; matching on the key appended a
+        duplicate `1:50` to every one of them.
+
+        An ezdxf-authored drawing has **no** ACAD_SCALELIST entries at all.
+        AutoCAD repopulates its default list only when the dictionary is
+        empty, so appending the one requested scale to an empty list shipped a
+        file whose scale list *was* that one entry — measured (AutoCAD 2026):
+        `SetVariable("CANNOSCALE", "1:1")` then fails with `Error setting
+        system variable` and the annotation scale cannot be changed in the
+        UI. An empty list is therefore seeded with AutoCAD's own default list
+        first, so the saved file offers what a fresh AutoCAD drawing offers.
+        AutoCAD only honours a CANNOSCALE that names an entry of this list.
+        """
+        scales = doc.rootdict.get("ACAD_SCALELIST")
+        if scales is None:
+            scales = doc.rootdict.add_new_dict("ACAD_SCALELIST")
+        if len(scales) == 0:
+            for default in cls._DEFAULT_SCALE_LIST:
+                _, default_paper, default_drawing = parse_scale(default)
+                cls._add_scale_object(doc, scales, default, default_paper, default_drawing)
+        for _key, entry in scales.items():
+            if cls._scale_entry_name(entry) == name:
+                return False
+        cls._add_scale_object(doc, scales, name, paper, drawing)
         return True
 
     def _set_annotation_scale(self, doc, value: Any) -> str:

server.py

@@ -5082,7 +5082,9 @@ async def drawing_settings(
     headless engine (`capability: registry_sysvar` — AutoCAD keeps them in the
     registry, a file cannot); `dimstyle` / `textstyle` when the backend has no
     styles contract. On the live engine `annotation_scale` must name a scale in
-    the drawing's scale list (SCALELISTEDIT); the headless engine adds it.
+    the drawing's scale list (SCALELISTEDIT) and AutoCAD refuses the write on a
+    paper-space layout with no active viewport; the headless engine adds the
+    scale, seeding AutoCAD's default list first when the drawing has none.
     Headlessly, grid/snap are stored on the active VPORT and the annotation
     scale in the variable dictionary — the places AutoCAD reads them from.
     """

tests/test_settings_facade.py

@@ -17,6 +17,7 @@ from __future__ import annotations
 
 import json
 import math
+from types import SimpleNamespace
 
 import pytest
 
@@ -156,13 +157,30 @@ async def test_annotation_scale_writes_the_variable_dictionary_and_scale_list(ba
     vardict = backend._doc.rootdict.get("AcDbVariableDictionary")
     assert vardict["CANNOSCALE"].dxf.value == "1:50"
     scales = backend._doc.rootdict.get("ACAD_SCALELIST")
-    assert scales.get("1:50") is not None, "AutoCAD only honours a scale that is in the list"
+    assert "1:50" in _scale_names(scales), "AutoCAD only honours a scale that is in the list"
     assert await backend.system_get_variable("CANNOSCALEVALUE") == pytest.approx(0.02)
 
+    # Regression: the list used to be *only* the requested scale. AutoCAD
+    # repopulates its default list only when the dictionary is empty, so the
+    # saved file opened with a one-entry scale list and `CANNOSCALE` could no
+    # longer be changed there (measured, AutoCAD 2026). An empty list is seeded
+    # with AutoCAD's default metric list, keyed and ordered the way AutoCAD
+    # writes it, so 1:50 was already there and nothing was appended.
+    assert _scale_names(scales) == list(_AUTOCAD_DEFAULT_SCALES)
+    assert list(scales.keys()) == _AUTOCAD_DEFAULT_KEYS
+
     # re-setting the same scale is not a change and adds no second list entry
     again = await backend.drawing_settings({"annotation_scale": "1:50"})
     assert again["changed"] == {}
-    assert len(list(scales.keys())) == 1
+    assert len(list(scales.keys())) == 17
+
+    # a scale outside the default list is appended once, under the next key
+    # AutoCAD itself would use, and the defaults are not seeded twice
+    res = await backend.drawing_settings({"annotation_scale": "1:25"})
+    assert res["ok"] is True, res.get("errors")
+    assert _scale_names(scales) == [*_AUTOCAD_DEFAULT_SCALES, "1:25"]
+    assert list(scales.keys()) == [*_AUTOCAD_DEFAULT_KEYS, "B7"]
+    await backend.drawing_settings({"annotation_scale": "1:50"})
 
     path = str(tmp_path / "scale.dxf")
     await backend.drawing_save_as(path)
@@ -171,9 +189,32 @@ async def test_annotation_scale_writes_the_variable_dictionary_and_scale_list(ba
         "name": "1:50",
         "value": 0.02,
     }
+    reopened = backend._doc.rootdict.get("ACAD_SCALELIST")
+    assert _scale_names(reopened) == [*_AUTOCAD_DEFAULT_SCALES, "1:25"]
+    assert "1:1" in _scale_names(reopened), "AutoCAD refuses a list without its unit scale"
     assert backend._doc.audit().has_errors is False
 
 
+async def test_reading_the_annotation_scale_never_creates_the_scale_list(backend):
+    """The read side stays pure: a fresh headless drawing keeps its empty
+    ACAD_SCALELIST (AutoCAD fills it on open) until a scale is written."""
+    assert (await backend.drawing_settings())["settings"]["annotation_scale"] == {
+        "name": "1:1",
+        "value": 1.0,
+    }
+    scales = backend._doc.rootdict.get("ACAD_SCALELIST")
+    assert scales is None or len(list(scales.keys())) == 0
+
+
+# AutoCAD 2026's default metric list, in its own order and under its own keys
+# (`Documents.Add()` saved as DXF, MEASUREMENT 1): 17 SCALE objects, A0 … B6.
+_AUTOCAD_DEFAULT_SCALES = (
+    "1:1", "1:2", "1:4", "1:5", "1:8", "1:10", "1:16", "1:20", "1:30", "1:40",
+    "1:50", "1:100", "2:1", "4:1", "8:1", "10:1", "100:1",
+)  # fmt: skip
+_AUTOCAD_DEFAULT_KEYS = [f"A{i}" for i in range(10)] + [f"B{i}" for i in range(7)]
+
+
 def _seed_autocad_style_scale_list(doc, names):
     """ACAD_SCALELIST the way AutoCAD saves it: keyed A0, A1, ... with the scale
     name only in the SCALE object's group 300 (the default metric list already
@@ -222,11 +263,13 @@ async def test_annotation_scale_matches_an_autocad_keyed_scale_list_by_name(back
     assert _scale_names(scales).count("1:50") == 1
     assert backend._doc.rootdict.get("AcDbVariableDictionary")["CANNOSCALE"].dxf.value == "1:50"
 
-    # a scale the list does not carry is still appended, exactly once, under a
+    # a scale the list does not carry is still appended, exactly once, under
-    # key that does not collide with AutoCAD's A<n> keys
+    # the next key in AutoCAD's own A0 … A9, B0 … sequence; a non-empty list
+    # is never re-seeded with the defaults
     res = await backend.drawing_settings({"annotation_scale": "1:25"})
     assert res["ok"] is True, res.get("errors")
     assert _scale_names(scales).count("1:25") == 1
+    assert list(scales.keys()) == ["A0", "A1", "A2", "A3"]
     assert len(list(scales.keys())) == 4
     await backend.drawing_settings({"annotation_scale": "1:25"})
     assert len(list(scales.keys())) == 4
@@ -383,6 +426,24 @@ async def test_server_tool_carries_the_new_keys(backend):
 
 
 class _FakeApp:
+    """The ActiveX *Application*: it has an ActiveDocument and no sysvar
+    members at all. `GetVariable` / `SetVariable` live on `AcadDocument`;
+    a backend that calls them on the Application raises `AttributeError:
+    AutoCAD.Application.GetVariable` on a live seat before anything reaches
+    AutoCAD (measured, AutoCAD 2026) -- so this fake refuses them the same way
+    instead of quietly answering."""
+
+    def __init__(self):
+        self.ActiveDocument = _FakeDoc()
+        self.Documents = SimpleNamespace(Count=1)
+
+    def __getattr__(self, name):
+        if name in ("GetVariable", "SetVariable"):
+            raise AttributeError(f"AutoCAD.Application.{name}")
+        raise AttributeError(name)
+
+
+class _FakeDoc:
     """Records GetVariable / SetVariable; stores what AutoCAD would store."""
 
     def __init__(self):
@@ -425,6 +486,8 @@ def com_backend(monkeypatch):
     from backends import com_backend as module
 
     app = _FakeApp()
+    # Only the Application is faked: the real `_acad_doc()` runs and resolves
+    # `ActiveDocument`, so the test exercises the true call path.
     monkeypatch.setattr(module, "_acad_app", lambda: app)
     backend = module.ComBackend()
 
@@ -432,11 +495,11 @@ def com_backend(monkeypatch):
         return func(*args, **kwargs)
 
     monkeypatch.setattr(backend, "_run", _run_inline)
-    return backend, app
+    return backend, app.ActiveDocument
 
 
 async def test_com_facade_sends_autocad_shaped_values(com_backend):
-    backend, app = com_backend
+    backend, doc = com_backend
     res = await backend.drawing_settings(
         {
             "limits": [[0, 0], [500, 350]],
@@ -450,7 +513,7 @@ async def test_com_facade_sends_autocad_shaped_values(com_backend):
         }
     )
     assert res["ok"] is True, res.get("errors")
-    sets = [call[1:] for call in app.calls if call[0] == "SetVariable"]
+    sets = [call[1:] for call in doc.calls if call[0] == "SetVariable"]
     assert sets == [
         ("LIMMIN", (0.0, 0.0)),
         ("LIMMAX", (500.0, 350.0)),
@@ -469,15 +532,15 @@ async def test_com_facade_sends_autocad_shaped_values(com_backend):
 async def test_com_point_variables_travel_as_variant_double_arrays(com_backend):
     import pythoncom
 
-    backend, app = com_backend
+    backend, doc = com_backend
     seen: dict = {}
-    real_set = app.SetVariable
+    real_set = doc.SetVariable
 
     def _capture(name, value):
         seen[name] = value
         real_set(name, value)
 
-    app.SetVariable = _capture
+    doc.SetVariable = _capture
     res = await backend.system_set_variable("LIMMAX", (300, 200))
     variant = seen["LIMMAX"]
     assert variant.varianttype == pythoncom.VT_ARRAY | pythoncom.VT_R8
@@ -498,7 +561,7 @@ async def test_com_point_variable_reply_survives_the_mcp_output_schema(com_backe
 
     import server
 
-    backend, app = com_backend
+    backend, doc = com_backend
 
     async def _fake_make_backend():
         return backend
@@ -507,11 +570,49 @@ async def test_com_point_variable_reply_survives_the_mcp_output_schema(com_backe
     async with Client(server.mcp) as client:
         res = await client.call_tool("system_set_variable", {"name": "LIMMAX", "value": [300, 200]})
     assert res.data == {"ok": True, "variable": "LIMMAX", "value": [300.0, 200.0]}
-    assert app.store["LIMMAX"] == (300.0, 200.0)
+    assert doc.store["LIMMAX"] == (300.0, 200.0)
 
 
 async def test_com_refuses_before_touching_activex(com_backend):
-    backend, app = com_backend
+    backend, doc = com_backend
     res = await backend.drawing_settings({"limits": [[0, 0], [0, 0]], "annotation_scale": "x"})
     assert res["ok"] is False
-    assert not [call for call in app.calls if call[0] == "SetVariable"]
+    assert not [call for call in doc.calls if call[0] == "SetVariable"]
+
+
+async def test_com_sysvars_are_read_and_written_on_the_document(com_backend):
+    """Regression: both methods called `GetVariable` / `SetVariable` on the
+    object `_acad_app()` returns, which has neither, so every
+    `drawing_settings` read and write on the live engine died with
+    `AttributeError: AutoCAD.Application.GetVariable`."""
+    backend, doc = com_backend
+    assert await backend.system_get_variable("LIMMAX") == (420.0, 297.0)
+    res = await backend.system_set_variable("GRIDMODE", "1")
+    assert res == {"ok": True, "variable": "GRIDMODE", "value": 1}
+    assert doc.store["GRIDMODE"] == 1
+    assert doc.calls == [
+        ("GetVariable", "LIMMAX"),
+        ("GetVariable", "GRIDMODE"),  # the type probe
+        ("SetVariable", "GRIDMODE", 1),
+    ]
+
+
+async def test_com_sysvars_need_an_open_drawing(monkeypatch):
+    """With no document open there is nothing to read the variable from:
+    `_acad_doc()`'s clear refusal is what surfaces, not an ActiveX error."""
+    pytest.importorskip("win32com.client", reason="pywin32 not installed")
+    from backends import com_backend as module
+
+    app = _FakeApp()
+    app.Documents = SimpleNamespace(Count=0)
+    monkeypatch.setattr(module, "_acad_app", lambda: app)
+    backend = module.ComBackend()
+
+    async def _run_inline(func, *args, **kwargs):
+        return func(*args, **kwargs)
+
+    monkeypatch.setattr(backend, "_run", _run_inline)
+    with pytest.raises(RuntimeError, match="No drawing is open"):
+        await backend.system_get_variable("LIMMAX")
+    with pytest.raises(RuntimeError, match="No drawing is open"):
+        await backend.system_set_variable("GRIDMODE", 1)