[KristofferC]

I saw a number of invalidations stemming from this definition in a ReverseDiff workload.

hash(t::TrackedReal) = hash(value(t)) gives the same result as Base's generic hash(x) = hash(x, HASH_SEED) combined with the two-argument hash(::TrackedReal, ::UInt) that ReverseDiff already defines. Being more specific, it invalidates hash(::Any) callers (e.g. Dict rehashing) when the package loads.

On a small gradient workload (using ReverseDiff; ReverseDiff.gradient(x -> sum(abs2, x), rand(10))) on Julia master (with the Base invalidation fixes from the JuliaLang/julia kc/ttfx-invalidations branch) this takes the invalidated method instances from 932 to 763, and the ones that actually get re-inferred from 91 to 37.

🤖 Generated with Claude Code

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 89.37%. Comparing base (b796032) to head (b93d6b3).

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #322      +/-   ##
==========================================
- Coverage   89.38%   89.37%   -0.01%     
==========================================
  Files          19       19              
  Lines        1978     1977       -1     
==========================================
- Hits         1768     1767       -1     
  Misses        210      210              

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.