Fix ubuntu-24.04-arm wheel test failure (phreeqc engine)

[rkingsbury]

Summary

The v1.6.0 release wheel job failed only on ubuntu-24.04-arm while ubuntu-latest (x86_64) and macos-latest (arm64) built and tested cleanly.

Root cause: test_repeated_equilibrate (newly added since v1.5.0) lives in tests/test_engine_phreeqc2026.py — the file run inside the built wheel by cibuildwheel — but was the only test there using engine="phreeqc", the legacy engine backed by the third-party phreeqpython package. phreeqpython ships native libraries for x86_64 Linux (viphreeqc.so), Windows (.dll), and macOS universal (.dylib, incl. arm64), but no aarch64-Linux library. So on ubuntu-24.04-arm, PhreeqPython() raises OSError, and PhreeqcEOS was left in a broken state that crashed with a confusing ValueError: There is a problem with your input ... 'str' object has no attribute '_phreeqc'.

This is why the other platforms pass: x86_64 Linux loads its .so, and macOS arm loads the universal .dylib.

Changes

1. test(phreeqc) — test_repeated_equilibrate now uses engine="phreeqc2026" (the in-tree compiled engine), matching every other test in this file. This is the fix that turns the arm wheel job green; the "phreeqc" was a copy-paste slip.

2. fix(PhreeqcEOS) — make the engine degrade cleanly when the phreeqpython backend can't load, honoring the documented contract ("pyEQL will work, but equilibrate() will have no effect"):

• set self.pp = None in the OSError handler so state is well-defined (it had been left pointing at the wrong-typed in-tree wrapper inherited from super().__init__());
• equilibrate() no-ops when self.pp is None;
• _setup_ppsol() raises a clear ValueError when self.pp is None — already caught by the activity-coefficient fallback, so property reads keep working;
• generalize the previously Apple-only diagnostic message to also mention aarch64 Linux.

The PhreeqcEOS OSError bug is pre-existing (identical at v1.5.0); the new test was simply the first to exercise it in the wheel build. Fix #2 does not change CI outcome — it just prevents the crash if anyone hits that path at runtime.

Verification

• tests/test_engine_phreeqc2026.py → 20 passed (the wheel-tested file)
• tests/test_engine_phreeqc.py → 19 passed (real phreeqpython path, unaffected)
• Simulated the arm case (engine.pp = None): equilibrate() no-ops, get_activity_coefficient degrades to 1, _setup_ppsol raises the legible ValueError

The new guards are inert for engine="phreeqc2026" (its self.pp is never None), so that engine is untouched.

🤖 Generated with Claude Code

[rkingsbury]

@vineetbansal this is the problem I emailed you about - it turned out to be a result of a copy/paste error in a newly added test. All good now!

[codecov]

Codecov Report

❌ Patch coverage is 0% with 6 lines in your changes missing coverage. Please review.
✅ Project coverage is 87.54%. Comparing base (1e877c3) to head (dd1be5f).
⚠️ Report is 6 commits behind head on main.

Files with missing lines	Patch %	Lines
src/pyEQL/engines.py	0.00%	4 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #455      +/-   ##
==========================================
- Coverage   87.81%   87.54%   -0.28%     
==========================================
  Files          14       14              
  Lines        1945     1951       +6     
  Branches      338      340       +2     
==========================================
  Hits         1708     1708              
- Misses        188      192       +4     
- Partials       49       51       +2     

☔ 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.