prevent adding methods to the functions > and >=

[nsajko]

As documented, the intended way to implement > is to add a method to <. Similarly with >=. A package should never add a method to either > or >=.

[odow]

Assuming the fallbacks work

[odow]

This is the only downside. We might talk about < when the user wrote >.

[odow]

Aside from recommended practice, what is the motivation for this?

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (4056c00) to head (b58a1c0).
⚠️ Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff            @@
##            master     #4065   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           43        43           
  Lines         6201      6201           
=========================================
  Hits          6201      6201           

☔ View full report in Codecov by Sentry.
📢 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.

[nsajko]

Aside from recommended practice, what is the motivation for this?

This PR, together with PR JuliaDiff/ForwardDiff.jl#771, greatly decreases the amount of sysimage invalidation on using JuMP:

• Status before both PRs:

  • unique invalidation count: 2570

  • data: without_open_PRs.json

• Status after both PRs:

  • unique invalidation count: 1468

  • data: with_open_PRs.json

EDIT: the above is with nightly (v1.13) Julia, JuliaLang/julia@865b8be

[nsajko]

Status with this PR, but without PR JuliaDiff/ForwardDiff.jl#771:

• unique invalidation count: 2262

• data: with_open_JuMP_PR.json

So an improvement, but not as much as with PR JuliaDiff/ForwardDiff.jl#771.

[nsajko]

NB: if you want to reproduce the above results, do using JuMP without loading anything (other than SnoopCompileCore) before. Not even the REPL or Pkg (both load StyledStrings which does piracy so causes invalidation, changing the results).

[odow]

Invalidation is a good reason