* chore: add gitignore entries for local eval repos, e2e fixtures, and env files
* fix: restore clean bubble_sort_method.py test fixture
The call-site ID commit re-contaminated this file with instrumentation
decorators, causing tests to fail with missing CODEFLASH_LOOP_INDEX.
* fix: resolve ruff and mypy errors in codeflash-python
- Add import-not-found ignores for optional torch/jax imports
- Extract magic column index to _STDOUT_COLUMN_INDEX constant
- Fix unused variable in _instrument_sync.py
- Cast cpu_time_ns to int for mypy arg-type
* fix: add skip markers for optional deps and apply ruff formatting to tests
Skip torch/jax/tensorflow tests when those packages are not installed.
Move has_module helper to conftest.py for reuse across test files.
Apply ruff format to all test files that drifted.
* fix: resolve remaining ruff format and mypy errors
- Add missing blank line in conftest.py (ruff format)
- Remove unused import-untyped ignore on jax import (mypy unused-ignore)
- Add type: ignore comments for object-typed SQLite row values
* chore: bump codeflash-python to 0.1.1.dev0
Replace the old AST-injected codeflash_wrap/InjectPerfOnly sync path
with decorator-based instrumentation matching the async path:
- Add codeflash_performance_sync and codeflash_behavior_sync decorators
with GPU device sync (torch CUDA/MPS, JAX, TensorFlow) via find_spec
- Add sync_devices_before/sync_devices_after with lazy cached detection
- Clean _instrumentation.py to a thin sync/async dispatcher (~47 lines)
- Remove dead code from _instrument_core.py (create_wrapper_function,
create_device_sync_statements, get_call_arguments, etc.)
- Fix all production imports to point at source modules directly
- Drop underscore prefixes on internal helpers (connections, get_async_db,
close_all_connections, detect_device_sync, etc.)
- Rewrite all test files for the new sync path assertions
- Add real-framework GPU device sync tests (torch, jax, tensorflow)
Measure both wall-clock time (perf_counter_ns) and CPU thread time
(thread_time_ns) in instrumented test code. cpu_runtime is now a required
int field on FunctionTestInvocation, stored in the SQLite test_results
table as a 10th column.
Also fixes the sleeptime.py bug (10e9 → 1e9 divisor) and removes the
binary pickle parser (parse_test_return_values_bin) since no writer
exists in the current codebase — SQLite is the sole data capture path.
The busy-wait sleep function can overshoot by 90%+ under CPU contention
(observed 190ms for a 100ms target). The test verifies that
instrumentation produces runtimes in the right order of magnitude,
not that sleep timing is precise. Widen rel_tol from 0.05 to 1.0.
