slalombuild
Enable security teams to reproduce security findings
Constantly re-running linters on all source code, the default mode of operation for most linters, is very time-consuming and gets frustrating fast when you know that you've only modified a few functions in these files, but the linters are re-linting all these other hundreds of files that you haven't modified. We need to cache results so that you don't have to wait for linter work that's already been done, which is a surprisingly difficult thing to do:
We need to guarantee that results are reproducible, this could be achieved by sandboxing linter runs, similar to how Google's Bazel project sandboxes compiler actions.
We need to intelligently invalidate/ignore cache entries: if fileA is modified and check results for fileB depend on the contents of fileA, then check results for fileB from prior to fileA's modification cannot be reused. We could solve this by keying cache entries on the linter configuration, the actual linter target file, and all dependencies of said target file.
This is a really cool, heavy lift nice to have. Deprioritized below other features, but still valid for future state. To be reassessed later as an epic
