Dataflow: Simplify the call-edge join in reverse through-flow.

[aschackmull]

Alternative to #18355.

For a summary edge arg->par->ret->out found in reverse flow with access paths ap at arg/par and innerReturnAp at ret/out it seems superfluous to include innerReturnAp in the call-edge arg-par join. The added constraint is that a path exists between par and ret with those access paths, but when we join flowThroughIntoCall in reverse flow, then that exactly what we just established, so I don't think it adds much (if anything). Dropping the column avoids the need to materialize a combination of 3 access paths (ap, innerReturnAp, and returnAp) that can vary independently in degenerate cases.

[Copilot]

Copilot wasn't able to review any files in this pull request.

Files not reviewed (1)

• shared/dataflow/codeql/dataflow/internal/DataFlowImpl.qll: Language not supported

Tip: Copilot only keeps its highest confidence comments to reduce noise and keep you focused. Learn more

[aschackmull]

Dca is mostly uneventful, except for some clear performance improvements for a couple of JS projects.

[smowton]

Looks good! I've started https://github.com/github/codeql-dca-main/issues/26112 to check if this fixes pathological case yaklang/yaklang as well as #18355, which reduced its analysis time by ~20 minutes (previous DCA for comparison: https://github.com/github/codeql-dca-main/tree/data/smowton/read-fields-bef__yaklang__nightly__1/reports)

[aschackmull]

Looks like yaklang/yaklang is fixed by this, right?

[smowton]

To a very similar degree, yes

So, 👍