Subst Model err in BEAUti

[walterxie]

Repeat #8 , but continue to choose HKY in Subst Model drop-down list:

The err dialog will pop out as attached, but no err message in console.

[alexeid]

This one I can't yet reproduce or pin down, so I don't want to guess at a fix.

Two things would help:

1. Please re-test on master after the ClassCastException in BEAUti #8 fix. ClassCastException in BEAUti #8 threw a ClassCastException during the Site Model panel's init, which leaves the panel half-built. Selecting HKY on top of that broken state could easily be what raised this dialog, in which case ClassCastException in BEAUti #8 may have taken it with it.

2. If it still happens, what does the dialog say? You noted there's nothing in the console, and I can't read the text out of the screenshot — the exact message (and any stack trace behind a "Details"/"Show Details" button, if there is one) is what I need to trace it.

One thing worth checking while you're there: does selecting HKY on a plain (non-bModelTest) partition from the standard template work? That tells us whether this is specific to switching away from the bModelTest site model or a more general spec-BEAUti subst-model issue.

[alexeid]

Fixed in ab76665 on master.

Root cause

The BEAUti template still declared two operators with legacy classes:

<operator spec='kernel.BactrianScaleOperator'         parameter='@BMT_mutationRate...'/>
<operator spec='kernel.BactrianDeltaExchangeOperator' parameter='@BMT_frequencies...'/>

kernel.* resolves through the namespace to beast.base.inference.operator.kernel.*. Those legacy operators expect a legacy RealParameter, but the migrated parameters are spec RealScalarParam, so parsing the "BEAST Model Test" subtemplate threw:

Input 102: type mismatch for input parameter
Provide type ...RealScalarParam, but expected ...RealParameter
  <operator id='BMT_mutationRateScaler.s:dna' spec='kernel.BactrianScaleOperator'>

That parse exception is caught inside BeautiSubTemplate.createSubNet, which aborts the rest of subtemplate setup — including applying suppressInputs. bModelTest's suppressInputs is meant to hide the Subst Model dropdown under BEAST Model Test; with it skipped, the dropdown stayed visible. Switching it to HKY then called removeSubNet on RevJump, which no template matches → RuntimeException: Cannot find template for removing RevJump.s:dna → the dialog you saw (with nothing in the console).

Fix

Both operators now use beast.base.spec.inference.operator.ScaleOperator / DeltaExchangeOperator — the same spec operators the example XML already uses, which accept spec parameters. The subtemplate parses cleanly, suppressInputs applies, and the Subst Model dropdown is correctly hidden, so the crash cannot occur. Verified headlessly (TestFX): after selecting BEAST Model Test, #substModelComboBox is absent.

FxTemplateTest now also lints the template for legacy operators, so this can't regress.

This was the same class of issue as #8 (legacy vs spec), just surfacing through the operators rather than the input editor. Could you re-test #7/#8/#9 together on master when you get a chance?