CompatHelper: bump compat for OrderedCollections to 2, (keep existing compat) - #511

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 15.31%. Comparing base (4351faf) to head (f50c8f4).

@@           Coverage Diff           @@
##             main     #511   +/-   ##
=======================================
  Coverage   15.31%   15.31%           
=======================================
  Files          22       22           
  Lines        1593     1593           
=======================================
  Hits          244      244           
  Misses       1349     1349           

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

+1 on this — it's the last OrderedCollections cap in our dependency graph.

The test matrix here is green on all five jobs, including min and 1.11:

test (1, ubuntu-latest)      pass      test (1, macOS-latest)     pass
test (1, windows-latest)     pass      test (1.11, ubuntu-latest) pass
test (min, ubuntu-latest)    pass

The two red checks are docs and format, neither of which exercises OrderedCollections, and both pass on main. Their annotations only give a generic Process completed with exit code 1 plus a Failed to delete old caches warning (both on Julia 1.13.0), so I can't confirm the cause from outside — but this run is from 22 September and main has had a green docs build since, so a re-run or rebase may well clear them.

For downstream context on why this matters: StanSample's MCMCChainsExt currently cannot reach OrderedCollections 2 at all, because every MCMCChains release including 7.7.0 caps it to 1.x. We're widening the StanJulia side (StanJulia/StanSample.jl#87, StanJulia/StanBase.jl#53, StanJulia/StanOptimize.jl#39) and AbstractPPL (TuringLang/AbstractPPL.jl#190), and this PR is what unblocks the MCMCChains path.

If it would help, I'm happy to run the suite locally with OrderedCollections pinned to 2.0.1 and report back — that's what we did for AbstractPPL (1500 passed, 0 failed), and it would confirm the compat bump directly rather than relying on whatever the CI happened to resolve.