#19599 Create a benchmark repository that reproduces performance issues Issue created by jriddy Problem I, as well, as many users reporting in Slack, struggle with poor performance of Pants when performing core features like dependency resolution or inference. However, the repos the reproduce these issues are usually private corporate repos whose contents cannot be shared outside their owning organizations. In order to solve these issues in Pants open source, we need to be able to reliably reproduce them. Problem We need a complicated, convoluted monorepo that replicates the performance issues we're seeing, so we can use this to run benchmarks with different versions, to keep us from stabbing in the dark with optimizations. Alternatives • Tailoring an open-source Python repo • Converting an open-source Bazel repo to Pants • Generating one • Anonymizing a corporate repo Criteria • Running

pants peek ::

on the repo should take on the order of ~60 sec with no memoization and a warm cache. • Running

pants dependencies ::

should consume approaching 4 GiB of memory pantsbuild/pants