mirror of
https://github.com/comfyanonymous/ComfyUI.git
synced 2026-05-28 01:47:32 +08:00
dc6190e8ba
Some checks failed
Python Linting / Run Ruff (push) Has been cancelled
Python Linting / Run Pylint (push) Has been cancelled
Build package / Build Test (3.10) (push) Has been cancelled
Build package / Build Test (3.11) (push) Has been cancelled
Build package / Build Test (3.12) (push) Has been cancelled
Build package / Build Test (3.13) (push) Has been cancelled
Build package / Build Test (3.14) (push) Has been cancelled
The per-tag microsecond stagger preserves intra-batch order, but two back-to-back write batches on the same reference (e.g. set_reference_tags for path tags, then add_tags_to_reference for user tags) call get_utc_now() independently. On Windows the system clock can return the same datetime for both calls if no OS tick elapsed between the commits — both batches end up sharing microseconds and ORDER BY added_at, tag_name falls back to the alphabetic tiebreaker, sorting user tags ahead of path tags they were meant to follow. Add _next_added_at_base(reference_id) that reads max(existing added_at) and returns max(existing + 1us, get_utc_now()), guaranteeing the new batch sorts strictly after anything previously written for that reference. Used by set_reference_tags and add_tags_to_reference; batch_insert_seed_assets stays on raw get_utc_now() since seed inserts are always the first writes for a new reference. The accompanying regression test pins get_utc_now() to a frozen value so the previously-Windows-only race becomes a platform-independent failure mode under test. |
||
|---|---|---|
| .. | ||
| assets | ||
| database | ||
| __init__.py | ||
| app_settings.py | ||
| custom_node_manager.py | ||
| frontend_management.py | ||
| logger.py | ||
| model_manager.py | ||
| node_replace_manager.py | ||
| subgraph_manager.py | ||
| user_manager.py | ||