fitzee commented on PR #44034: URL: https://github.com/apache/superset/pull/44034#issuecomment-5600169174
@rebenitez1802 Thanks — I went through the remaining low nits as well. Implemented in `0c283b88d8`: - aligned tiled repaint handling with the standard path by catching Playwright errors while explicitly preserving Celery `SoftTimeLimitExceeded` propagation; - switched the frozen execution-context test rebuild to `dataclasses.replace`; - made combined-image blankness advisory after individually validated contentful tiles; - added structural-edge calibration and high-resolution light/thin chart and sparse-table coverage. Intentionally left unchanged: - standard content-probe failure remains fail-open because rejecting when DOM state itself cannot be read would create an unbounded false-positive class; the probe failure is logged; - multi-tone dark-theme blank detection is outside this incident’s evidence. Exactly uniform dark captures are already caught by the dominant-color rule; broader dark heuristics should be calibrated against a real failure raster; - `getcolors(maxcolors=pixel_count)` operates only on bounded 256/1024 samples and is needed for the diagnostic dominant ratio, so the old 256-color fast-bail is not applicable; - the standard/tiled retry loops stay separate because tiled capture has clip positioning, per-tile holder state, thumbnail retention, and combination semantics that the standard path does not share; - `is_screenshot_nearly_uniform` remains as a compatibility wrapper; production decisions use `get_screenshot_blankness_metrics`, whose name reflects the broader behavior. Validation: 129 affected utility tests pass; Ruff and mypy pass. The actual incident raster remains classified blank with `structural_edge_ratio=0.00000`. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
