Aggregates broke when SET became optional (upstream CI regression)

[ako]

Fixes the TestMxCheck_DoctypeScripts/02-microflow-examples.mdl failure on the upstream sync PR — six errors on both engines, from a regression in e0744b9e ("make SET optional"):

[error] [CE0109] "Undefined variable 'ProductList.Price'." at Aggregate list activity 'Take sum ProductList.Price'
[error] [CE0015] "Aggregate function must specify a valid attribute." at Aggregate list activity 'Take sum ProductList'

What broke

Making SET optional turned setStatement into $X = <expr>, which overlaps every VARIABLE EQUALS <function-call> statement in the grammar. It sat at alternative 5 of ~50, and ANTLR's ALL(*) picks the lowest-numbered alternative that matches — so $Sum = sum($List.Price) stopped reaching aggregateListStatement and fell into the generic SET conversion, which joined the list and the attribute into one name and dropped the per-item expression.

Two fixes, because there were two defects

1. setStatement moved last in microflowStatement, so it only claims what no dedicated rule parses. Restores the previous routing for aggregates, list operations and RANGE alike.
2. The SET conversion now agrees with the canonical one (buildSetAggregate), and buildListAggregateAsFunction appends the per-item expression argument it never appended. This half was wrong before the bare form ever reached it — set $X = sum($List.Price) produced the same broken aggregate on main today.

Verification

Mendix 11.12.1 locally (CI used 11.12.2).

Check	Result
TestMxCheck_DoctypeScripts — all scripts, both engines, real mx check	ok, 629s
go test ./mdl/... ./cmd/...	pass
Control sweep, all mdl-examples/**/*.mdl, new vs main binary	no differences
New bug-test script → blank project → mx check	0 errors (control binary: 8 errors, the reported CE0109s)
Both fixes reverted, new tests re-run	fail with the reported symptom

Note on why this shipped

The original SET? change was swept with a control binary over every example — but with mxcli check, which parses and validates and never serializes. This defect lives between the AST and the BSON, where only exec + mx check can see it. Recorded in the symptom table: for a grammar change, the control sweep has to run the integration gate, not check.

Files

• mdl/grammar/domains/MDLMicroflow.g4 — setStatement last, with a comment saying why
• mdl/visitor/visitor_microflow_statements.go — buildSetAggregate replaces extractVariableAndAttribute
• mdl/visitor/visitor_microflow_expression.go — buildListAggregateAsFunction appends the expression argument
• mdl/visitor/visitor_microflow_aggregate_test.go — new
• mdl-examples/bug-tests/aggregate-after-optional-set.mdl — new
• .claude/skills/fix-issue.md — symptom row

🤖 Generated with Claude Code

https://claude.ai/code/session_01JXnEgoc2NQP1Y2TWMCMXC4

---

Generated by Claude Code