PrefectHQ
prefect
Blog
Docs
Changelog
Blog
Docs
Changelog
Overview
Branches
Benchmarks
Runs
Performance History
Latest Results
Preserve a statement cache size of 0 AsyncPostgresConfiguration resolves both cache-size arguments with `or`: self.statement_cache_size = statement_cache_size or <setting> 0 is falsy, so it is replaced by the setting, which defaults to None. The key is then never added to connect_args and asyncpg keeps its cache. 0 is not an absent value for either of these, and the settings that back them say so. statement_cache_size: "Setting this to 0 is required when using PgBouncer in transaction mode." prepared_statement_cache_size: "When set to 0, statement caching is disabled." So `or` discards exactly the value the documentation gives as the reason to set them. Resolve with `is None` instead. Passing 0 through the settings already worked, because the argument is None there and the setting is consulted; only callers constructing the configuration directly were affected. Non-zero values were already preserved and still are; the tests cover both so the fix cannot be read as special-casing zero.
ckarnell:fix/statement-cache-size-zero
2 hours ago
Don't let a deleted working directory break all imports load_module reads the cwd with a bare os.getcwd() to put it on sys.path for relative imports. When the working directory has been removed out from under the process that raises FileNotFoundError, and nothing can be imported at all — including the machinery that would set a valid working directory again. Reachable from Runner._run_on_crashed_hooks: it calls load_flow_from_flow_run, which re-runs the deployment's pull steps, and running a step imports the step function via import_object -> load_module. If the run's directory is already gone, importing prefect.deployments.steps.set_working_directory fails before that step can correct the cwd, so on_crashed hooks silently never run. Reuses the existing guard rather than adding a second one: the private _safe_current_working_directory in deployments/steps/core.py moves to utilities/importtools.py as safe_current_working_directory, and core.py imports it from there.
fix-getcwd-in-load-module
8 hours ago
fix(ui-v2): render sanitized inline HTML in markdown content (#22763) Co-authored-by: Devin AI <158243242+devin-ai-integration[bot]@users.noreply.github.com> Co-authored-by: Alexander Streed <alex.s@prefect.io> Co-authored-by: alex.s <ajstreed1@gmail.com>
main
15 hours ago
Apply ruff format to the new tests
vedjaw:fix/19064-cap-event-tag-related-resources
16 hours ago
style: add blank line before class definition
kingsley-123:fix/validate-negative-timeout-in-pause-flow-run
21 hours ago
style: fix ruff formatting on pytest.raises
kingsley-123:fix/validate-negative-timeout-in-pause-flow-run
21 hours ago
fix: raise ValueError for negative timeout in pause_flow_run Negative timeout values were silently passed to time.sleep() which raised a confusing 'sleep length must be non-negative' error with no indication that the timeout parameter was invalid. Added input validation to raise ValueError immediately when a negative timeout is passed to pause_flow_run().
kingsley-123:fix/validate-negative-timeout-in-pause-flow-run
22 hours ago
Merge branch 'main' into worker-start-background
SreeramaYeshwanthGowd:worker-start-background
2 days ago
Latest Branches
CodSpeed Performance Gauge
0%
Preserve a statement cache size of 0
#22767
2 hours ago
b1d870e
ckarnell:fix/statement-cache-size-zero
CodSpeed Performance Gauge
+1%
Don't let a deleted working directory break all imports
#22765
9 hours ago
aac3fff
fix-getcwd-in-load-module
CodSpeed Performance Gauge
0%
fix: cap tag related resources to event maximum
#22694
17 hours ago
1963382
vedjaw:fix/19064-cap-event-tag-related-resources
© 2026 CodSpeed Technology
Home
Terms
Privacy
Docs