fix(ai_client_proxy): add _pending_lock threading.Lock

And fix test_discussion_takes_gui.py patches to use ai_client_stub
2026-05-13 11:24:58 -04:00
parent 169fe52092
commit 12f16e9a11
8 changed files with 195 additions and 15 deletions
@@ -0,0 +1,32 @@
+# Track: Fix Pre-Existing Test Failures
+
+## Context
+
+Two test failures that are not related to the ai_client_stub integration fix but need to be resolved for full test suite passing.
+
+## Failed Tests
+
+### 1. test_ai_client_proxy_run.py::test_initial_state_variables
+
+**File:** `tests/test_ai_client_proxy_run.py`
+**Error:** `AssertionError: Missing _pending_lock`
+
+**Root Cause:** The test expects `AIProxyClient` to have a `_pending_lock` attribute (a threading lock), but the class doesn't have it.
+
+**Fix:** Add `_pending_lock: threading.Lock` to `AIProxyClient.__init__` in `src/ai_client_proxy.py`
+
+### 2. test_discussion_takes_gui.py (both tests)
+
+**File:** `tests/test_discussion_takes_gui.py`
+**Error:** `ValueError: not enough values to unpack (expected 2, got 0)` at `src/gui_2.py:3668`
+
+**Root Cause:** The test mocks `imgui.input_text` and `imgui.input_int` but NOT `imgui.input_text_multiline`. When `_render_synthesis_panel` calls `imgui.input_text_multiline(...)`, the mock returns `None` (not unpacked), causing the unpacking failure.
+
+**Fix:** Add mock for `imgui.input_text_multiline` in the test setup.
+
+## Tasks
+
+1. [ ] Fix AIProxyClient - add `_pending_lock` threading.Lock to __init__
+2. [ ] Fix test_discussion_takes_gui.py - add mock for input_text_multiline
+3. [ ] Run tests to verify both fixes
+4. [ ] Run full test suite to confirm all pass
@@ -0,0 +1,61 @@
+# Track: Fix Test Patches for ai_client_stub Integration
+
+## Context
+
+After the refactor to use `ai_client_stub` as the module alias for `app_controller`, several tests fail because they use `patch('src.ai_client.X')` which doesn't properly reach the stub's module-level functions. This is a pre-existing architectural issue that needs fixing.
+
+## Root Cause Analysis
+
+When tests use `patch('src.ai_client.get_current_tier', return_value='Tier 3')`:
+
+1. `patch` creates/overrides `src.ai_client` in the `src` package namespace
+2. But `app_controller` does `from src import ai_client_stub as ai_client` at module load time
+3. The `ai_client` local reference in `app_controller` points directly to `ai_client_stub` module
+4. Patch modifies `src.ai_client` (a different object), not `src.ai_client_stub`
+5. Result: Functions in `ai_client_stub` aren't patched
+
+## Solution Applied
+
+Changed all patches from `patch('src.ai_client.X')` to `patch('src.ai_client_stub.X')` where the stub was the actual target.
+
+Also updated module imports in tests to use `ai_client_stub` instead of `ai_client` where appropriate.
+
+## Tasks Completed
+
+1. [x] Fix `test_on_tool_log_offloading` - change patch path to `src.ai_client_stub.get_current_tier`
+2. [x] Fix `test_redundant_calls_in_process_pending_gui_tasks` - change patches to `src.ai_client_stub`
+3. [x] Fix `test_gcli_path_updates_adapter` - use ai_client_stub module reference
+4. [x] Fix `test_telemetry_data_updates_correctly` - change patch to `src.ai_client_stub.get_token_stats`
+5. [x] Fix `test_gui_updates_on_event` - same as above
+6. [x] Run batch test to verify affected tests pass
+
+## Test Results
+
+**51 tests passing in the core batch:**
+- test_symbol_lookup.py (7)
+- test_ai_client_concurrency.py (1)
+- test_app_controller_offloading.py (2)
+- test_mma_agent_focus_phase1.py (8)
+- test_conductor_engine_v2.py (10)
+- test_conductor_tech_lead.py (7)
+- test_mma_dashboard_refresh.py (2)
+- test_mma_ticket_actions.py (2)
+- test_gui_phase3.py (3)
+- test_process_pending_gui_tasks.py (4)
+- test_gui_updates.py (3)
+
+## Known Pre-Existing Failures (NOT related to this fix)
+
+1. **test_ai_client_proxy_run.py::test_initial_state_variables** - AIProxyClient missing `_pending_lock` attribute
+2. **test_discussion_takes_gui.py** - incomplete mock setup for imgui.input_text_multiline
+
+## Files Modified
+
+- tests/test_app_controller_offloading.py
+- tests/test_process_pending_gui_tasks.py
+- tests/test_gui_updates.py
+
+## Checkpoint
+
+Commit: 169fe520 - fix(ai_client_stub): add module-level import for GeminiCliAdapter
+Additional commits for test fixes to follow.
@@ -0,0 +1,84 @@
+# Track: Fix Test Patches for ai_client_stub Integration
+
+## Context
+
+After the refactor to use `ai_client_stub` as the module alias for `app_controller`, several tests fail because they use `patch('src.ai_client.X')` which doesn't properly reach the stub's module-level functions. This is a pre-existing architectural issue that needs fixing.
+
+## Root Cause Analysis
+
+When tests use `patch('src.ai_client.get_current_tier', return_value='Tier 3')`:
+
+1. `patch` creates/overrides `src.ai_client` in the `src` package namespace
+2. But `app_controller` does `from src import ai_client_stub as ai_client` at module load time
+3. The `ai_client` local reference in `app_controller` points directly to `ai_client_stub` module
+4. Patch modifies `src.ai_client` (a different object), not `src.ai_client_stub`
+5. Result: Functions in `ai_client_stub` aren't patched
+
+## Failed Tests
+
+1. `tests/test_app_controller_offloading.py::test_on_tool_log_offloading`
+   - Patches `src.ai_client.get_current_tier` expecting it to affect `app_controller._on_tool_log()`
+   - But `app_controller` imported the stub directly, so patch doesn't reach it
+
+2. `tests/test_process_pending_gui_tasks.py::test_redundant_calls_in_process_pending_gui_tasks`
+   - Patches `src.ai_client.set_provider` and `src.ai_client.reset_session`
+   - Same root cause
+
+3. `tests/test_process_pending_gui_tasks.py::test_gcli_path_updates_adapter`
+   - Sets `ai_client._gemini_cli_adapter = None` expecting `app_controller` to see it
+   - Same import aliasing issue
+
+4. `tests/test_gui_updates.py::test_telemetry_data_updates_correctly`
+   - Patches `src.ai_client.get_token_stats`
+   - Same issue
+
+5. `tests/test_gui_updates.py::test_gui_updates_on_event`
+   - Same issue with `get_token_stats` patch
+
+## Solution Options
+
+### Option A: Fix Tests to Use Correct Patch Path
+Change all patches from `patch('src.ai_client.X')` to `patch('src.ai_client_stub.X')`.
+
+**Pros:**
+- Minimal code change
+- Resolves the actual issue
+- Follows proper patch semantics
+
+**Cons:**
+- Must update all affected tests
+
+### Option B: Add Module-Level Alias in src package
+Create `src/ai_client.py` that re-exports from `ai_client_stub`:
+```python
+from src.ai_client_stub import *
+```
+
+**Pros:**
+- Makes `src.ai_client` available as actual module
+- Tests using `patch('src.ai_client.X')` would work
+
+**Cons:**
+- Creates dual namespace confusion
+- May conflict with existing `ai_client_proxy.py`
+
+### Option C: Fix app_controller to Not Use Alias
+Change all `from src import ai_client_stub as ai_client` to `from src import ai_client` and ensure `src.ai_client` is the stub.
+
+**Cons:**
+- Significant refactor
+- May break other imports
+
+## Recommended Solution: Option A
+
+Fix the tests to use the correct patch path (`src.ai_client_stub`). This is surgical and addresses the root cause without creating architectural complexity.
+
+## Tasks
+
+1. [ ] Fix `test_on_tool_log_offloading` - change patch path to `src.ai_client_stub.get_current_tier`
+2. [ ] Fix `test_redundant_calls_in_process_pending_gui_tasks` - change patches to `src.ai_client_stub`
+3. [ ] Fix `test_gcli_path_updates_adapter` - use correct patch/module path
+4. [ ] Fix `test_telemetry_data_updates_correctly` - change patch to `src.ai_client_stub.get_token_stats`
+5. [ ] Fix `test_gui_updates_on_event` - same as above
+6. [ ] Run full batch test to verify all tests pass
+7. [ ] Create checkpoint commit