Duplicate Results? This detects platform specific "-expected.txt" files which are identical to the non-platform specific "-expected.txt" results.
Created attachment 84254 [details] [SCRIPT] Find Platform Expected Results Matching Regular Results Script I used to find the issues.
Created attachment 84255 [details] [OUTPUT] Unnecessary Platform Specific Results Files Summary: 180 platform specific expected results files matching regular expected results. I'm guessing this is the direction we want to move in for the obvious reason of less duplication but also because by tracking the normal expected results if the regular expected results change and these tests continue to match normal expected results then they won't be listed as failures because their dubious platform specific results were outdated. Does that make sense?
(In reply to comment #2) > Created an attachment (id=84255) [details] > [OUTPUT] Unnecessary Platform Specific Results Files > > Summary: > > 180 platform specific expected results files matching regular expected results. > > I'm guessing this is the direction we want to move in for the obvious reason > of less duplication but also because by tracking the normal expected results > if the regular expected results change and these tests continue to match > normal expected results then they won't be listed as failures because their > dubious platform specific results were outdated. Does that make sense? Yes. It also makes for fewer files to check out. Please put the script up for review and then remove the duplicate results!
It looks like `Tools/Scripts/deduplicate-tests` does this already, and correctly handles platform chains. I'll see if I can match its results with my script before I advance. With -v you can get output like: > * svg/text/select-textLength-spacingAndGlyphs-stretch-1-expected.txt > redundantly on mac-leopard and mac > consider deleting LayoutTests/platform/mac-leopard/svg/text/select-textLength-spacingAndGlyphs-stretch-1-expected.txt The script is poorly named. Its really deduplicating expected results, not tests. Also it checks png + checksums which I could do.
I don't understand the following case deduplicate-tests found: > * fast/canvas/canvas-text-alignment-expected.txt > redundantly on chromium-gpu-linux and chromium-gpu-win > consider deleting LayoutTests/platform/chromium-gpu-linux/fast/canvas/canvas-text-alignment-expected.txt I guess I should read run-webkit-tests to see the exact fallback behavior, but I don't see how chromium-gpu-linux would fall back to chromium-gpu-win. Updating my script to take care of fallback scripts it finds 497. I should verify my fallback behavior is what layout test actually does.