Sync ako/mxcli: test --attach on 11.14.0 explains its refusal; brain list

.claude/skills/fix-issue/findings/cmd-mxcli/2026-11-04-check-summary-omits-reference-errors.json

@@ -0,0 +1 @@
+{"area":"cmd/mxcli","date":"2026-11-04","symptom":"`mxcli check --references` printed `✗ 1 reference error(s) found` and then `0 issues: 0 errors, 0 warnings, 0 info` — the reporter asked why a reference error does not count as an error (mendixlabs/mxcli#1726)","cause":"the closing summary sums the `printed` list, which only the `report` helper appends to; the reference tier prints its errors and warnings itself, so they never reached the total","file":"cmd/mxcli/cmd_check.go (reference tier appends to printed); test cmd/mxcli/check_reference_summary_test.go","insight":"When one summary is computed over a list that several tiers feed, a tier with its own print path drops out of the total silently. Only the run in the report had an earlier tier's violation, so the summary line printed at all; a script with nothing but a reference error printed no summary, which is why the test asserts `1 errors` rather than the absence of `0 errors`. Control: without the append, the test fails with no summary in the output.","refs":["ako/mxcli#1928","mendixlabs/mxcli#1726"]}

.claude/skills/fix-issue/findings/cmd-mxcli/2026-11-04-nightly-11-14-only-test-attach-refused-cold-restart.json

@@ -0,0 +1 @@
+{"area":"cmd/mxcli","date":"2026-11-04","symptom":"The nightly goes red on the Mendix 11.14.0 leg only, in `TestTestAttach_KeepsTheHostingAppsPageBundles` (added the day before by ako/mxcli#1917): `mxcli test --attach` exits 1 with `this change needs a runtime restart (an entity or association changed), which --attach cannot do`, although the attach only adds a test microflow. 11.12.x, 11.13, 11.15 and 11.16 are green","cause":"On 11.14.0 every later build in an mxbuild --serve process fails, even with the model unchanged (`Cannot find module …/web/rollup.config.mjs`), and the failed /build response carries no restartRequired key at all. #1917 routes the attach's build through serveRecovery, whose legacy-hint path restarts onto a private serve and builds cold; a cold build reports restartRequired=true whatever changed, and runner_attach read that as a structural change. Before #1917 the same attach failed with the build failure itself, so attach never worked on 11.14.0; the test exposed it because its default 11.12.6 falls back to whatever mxbuild the nightly leg caches","file":"`cmd/mxcli/docker/serve_recovery.go` (rebuildFresh marks the cold answer `BuildResult.ColdRestart`; the incomplete-bundle path clears it when it takes the incremental answer), `cmd/mxcli/testrunner/runner_attach.go` (`attachRestartRefusal`), `cmd/mxcli/cmd_test_attach_bundle_integration_test.go` (11.14 branch)","insight":"A recovery that substitutes a fresh process's answer must say so: a cold build's restartRequired is \"unknown\", and only a caller that can restart may treat it as true. Settled by driving /build directly (a throwaway test with StartServe + Build, no watch loop): 11.14.0 build 2 with no change = Failure with no restartRequired; 11.15.0 = Success, false for no change and for a microflow, true for an entity, 0 pages (the 11.12.5 shape #1917 already recovers). That table is what puts the limit on 11.14.0 alone, so the test pins `>= 11.14 && < 11.15` and fails when an 11.14 mxbuild builds incrementally again. Reproduce in about 4 minutes: `mxcli setup mxbuild --version 11.14.0`, JAVA_HOME at a JDK 25, `MXCLI_RUNLIFE_VERSION=11.14.0 go test -tags integration -run TestTestAttach_KeepsTheHostingAppsPageBundles ./cmd/mxcli/`. Controls: the old test fails with the nightly's exact message; the new test passes on 11.14.0 and 11.15.0 and fails on 11.14.0 against a binary without the fix (it names the entity); stubbing the ColdRestart mark fails the two unit tests that pin it","refs":["mendixlabs/mxcli nightly run 41801646309","ako/mxcli nightly run 41801123003","ako/mxcli#1917","mendixlabs/mxcli#1723"],"test":"cmd/mxcli/cmd_test_attach_bundle_integration_test.go; cmd/mxcli/docker/serve_attached_test.go; cmd/mxcli/docker/serve_recovery_test.go; cmd/mxcli/testrunner/runner_attach_restart_test.go"}

.claude/skills/fix-issue/findings/mdl-executor/2026-11-04-check-resolves-every-statement-target.json

@@ -0,0 +1 @@
+{"area":"mdl/executor","date":"2026-11-04","symptom":"`check --references` passed drop/rename/move of a document that does not exist, and alter enumeration / alter microflow on one, while exec refused each with `<kind> not found` — 9 of the 11 statement forms that name an existing document (only alter entity and alter page were resolved)","cause":"validateAlterTarget was written for ALTER and extended one statement at a time; the drop/rename/move statements resolve against the same listings but were never routed to it","file":"mdl/executor/validate_alter_target.go (validateAlterTarget, targetListing, checkTarget, moveTargetKind), mdl/executor/validate.go (collectSingle: indirectDocs, renamedModules, opaqueModules, opaqueScript); test cmd/mxcli/check_statement_target_test.go","insight":"The risk in widening a not-found check is the false refusal, not the miss: a document can reach a name without a create statement naming it — an earlier rename or move, a rename module, an execute script, a generated SQL connector, a business event service, a bulk external-entities import. Record each in the pre-pass and stay silent for that name or module; the check must stay quieter than exec. Matching is exact-case because exec's lookups are. A PedApp test per form, each with an existing-target control, plus if-exists / created / renamed / moved controls. Control: without the new cases all 13 forms pass check.","refs":["ako/mxcli#1821","ako/mxcli#555"]}

.claude/skills/fix-issue/findings/mdl-executor/2026-11-04-check-resolves-widget-attributes.json

@@ -0,0 +1 @@
+{"area":"mdl/executor","date":"2026-11-04","symptom":"`textbox (Attribute: PlannedOn)` in a data view, or `ContentParams: ({1} = PlannedOn)` / `CaptionParams`, on an entity without that attribute: check --references passed, exec wrote it, mxbuild CE1613 \"The selected attribute 'M.E.PlannedOn' no longer exists\" — measured 3 of 3 on a fresh 11.12.6 app, control page clean","cause":"the input-widget scope walk (checkInputBinding) looked the attribute up and returned silently when it was not found unless the name was an association; template parameters were never looked up","file":"mdl/executor/validate_widget_attribute_type.go (reportMissingAttribute, checkTemplateParams), mdl/executor/validate_widget_attribute_scope.go (walk); test cmd/mxcli/check_widget_attribute_exists_test.go","insight":"The context walk, the member index with inheritance and the association hops were already there for the type rule (MDL-WIDGET39) — the not-found case was a silent `return` in it. Before writing a new resolver, grep for the lookup that already returns found/known and see what it does on a miss. Stay quiet where the index is not sure: an unknown chain (known=false), an unresolved context, an altered entity, an expression or $var path, and the audit members (createdDate/changedDate/owner/changedBy), which bind by name but are not in the attribute list. Control: without the two calls the textbox and content-parameter subtests pass check.","refs":["ako/mxcli#555"]}

.claude/skills/fix-issue/findings/mdl-executor/2026-11-04-show-page-ignores-alter-page-parameters.json

@@ -0,0 +1 @@
+{"area":"mdl/executor","date":"2026-11-04","symptom":"`alter page M.P { add parameters $X: … }` then a show page passing X, in one script: check --references and exec refused it with `page M.P has no parameter \"X\"`; the same statements as two scripts ran (mendixlabs/mxcli#1736)","cause":"the script pre-pass recorded page parameter names only from create page, so for an altered page the show page argument check fell through to the stored parameter list","file":"mdl/executor/validate.go (collectDefinitions AlterPageStmt case, pageParamEdits), mdl/executor/validate_show_page_params.go (recordPageParamEdits, applyPageParamEdits); test validate_show_page_params_alter_test.go","insight":"A script-context index that only listens to create statements reads every alter as a no-op — the same class as the #1355 script-set fix, one level down. Record the alter's edits rather than a computed list, because the pre-pass has no project to read the stored list from; apply them over whichever list wins at lookup time. End-to-end on PedApp 11.13.0: the base build refuses with the reported message, the fix passes check and exec writes both. Control: without the pre-pass case, the add/drop/create-then-alter subtests fail.","refs":["ako/mxcli#1940","mendixlabs/mxcli#1736"]}

.claude/skills/fix-issue/findings/mdl-executor/2026-11-04-validator-indexes-hold-excluded-twin.json

@@ -0,0 +1 @@
+{"area":"mdl/executor","date":"2026-11-04","symptom":"with a live and an excluded microflow of one name, `check --references` refused `call microflow M.Test(param1 = …)` with `has no parameter \"param1\" (declared parameters: param2)` — the excluded twin's parameters (mendixlabs/mxcli#1726)","cause":"the validators index stored documents by name in maps (`result[name] = …` over ListMicroflows/ListPages), so the twin listed LAST won: buildFlowSignatures, buildMicroflowReturnTypes, buildPageParamNames, buildPageParamEntities, the workflow call-microflow parameter map and microflowsByQualifiedName","file":"mdl/executor/validate_datasource_args.go, helpers.go, validate_show_page_params.go, validate_workflow_task_signature.go, validate_workflow_refs.go, cmd_odata_actions.go; test validate_excluded_twin_signature_test.go","insight":"The #1469 sweep fixed first-match LOOPS (`for … { if name matches { act; return } }`), where the excluded twin wins by being listed FIRST; a name-keyed MAP is the mirror image — the twin wins by being listed LAST — so #1469's excluded-first fixtures could not see it. Grep `GetQualifiedName(.*ContainerID, .*Name)\\] = ` for value maps (boolean name sets are safe), iterate liveFirst and keep the first entry. Test both list orders; the excluded-last case is the reported one. Control: without the change, the excluded-last subtests fail with the reporter's exact message.","refs":["ako/mxcli#1928","mendixlabs/mxcli#1726","mendixlabs/mxcli#1469"]}

.claude/skills/fix-issue/findings/other/2026-11-04-fixture-copy-races-unit-cache-journal.json

@@ -0,0 +1 @@
+{"area":"other","date":"2026-11-04","symptom":"A unit test fails at random on one nightly leg, in setup, not in what it tests: `TestCreateMicroflow_NotesRefusedOnFullConstructor` (mdl/backend/mcp) with `copy fixture: open .mxcli/cache/unit-contents.db-journal: no such file or directory`. The same commit passes on every other leg and on re-run","cause":"A cross-package race on the checked-in fixtures, not a product defect. Since the unit-contents store (ako/mxcli#1804) any reader that opens a project writes `.mxcli/cache/unit-contents.db` next to it, and tests that open testdata/expr-checker, testdata/pedapp or modelsdk/mpr/testdata/v1-project in place write it into the fixture; its SQLite journal appears and disappears with each write. go test runs packages in parallel, so another package's whole-directory `os.CopyFS` (25 tests) or hand-written walk listed the journal and then failed to open it. Nothing under a fixture's .mxcli is tracked","file":"`internal/testutil/fixture.go` (`CopyFixture`: os.CopyFS over a view that hides every .mxcli), used by every whole-fixture copy in tests, and by cmd/mxcli's copyTree and copyFixtureTree","insight":"A cache that writes next to the project turns every checked-in fixture into shared mutable state between test packages; any whole-directory copy of one is then a race whose odds grow with the suite. Copy what the fixture commits, never its .mxcli, which also makes the copy what a fresh checkout holds whichever tests ran first (exec_atomic's hashTree already ignored .mxcli for the same reason). After a full `make test`, `find . -name .mxcli -path '*testdata*'` lists the fixtures being written in place. The race is reproduced deterministically by an fs.FS whose listing names a journal that Open reports missing: plain os.CopyFS fails with the nightly's error, CopyFixture copies; with its filter stubbed it fails the same way","refs":["ako/mxcli nightly run 41801123003 (test (Mendix 11.12.2), Unit tests)","ako/mxcli#1804"],"test":"internal/testutil/fixture_test.go"}

.claude/skills/mendix/test-microflows/SKILL.md

@@ -16,14 +16,14 @@ description: "Write and run MDL-based microflow tests with `mxcli test` — anno
 | 185-227 | Test File Formats |
 | 228-243 | string Operations |
 | 244-267 | Running Tests |
-| 268-461 | How It Works |
-| 462-500 | Check a test file before you run it |
-| 501-622 | Writing Good Tests |
-| 623-637 | Test File Organization |
-| 638-650 | Interpreting Failures |
-| 651-672 | CI Integration |
-| 673-679 | Related Skills |
-| 680-682 | Warm test loop (`mxcli test --local [--watch]`, `--attach`, `run --local --test-endpoint`) |
+| 268-465 | How It Works |
+| 466-504 | Check a test file before you run it |
+| 505-626 | Writing Good Tests |
+| 627-641 | Test File Organization |
+| 642-654 | Interpreting Failures |
+| 655-676 | CI Integration |
+| 677-683 | Related Skills |
+| 684-686 | Warm test loop (`mxcli test --local [--watch]`, `--attach`, `run --local --test-endpoint`) |
 
 <!-- /skill-index -->
 # Test Microflows Skill
@@ -389,6 +389,10 @@ Three things to know before reaching for it:
 - **A change needing a runtime restart is refused** — a new entity or
   association. That runtime belongs to the other process. Restart it, or drop
   `--attach`.
+- **On Mendix 11.14.0 every attach is refused**, whatever the tests change: its
+  `mxbuild --serve` fails every build after its first, so the attach falls back
+  to a full build, and a full build asks for a runtime restart whatever changed.
+  Use `--local` there. (11.12.x, 11.13 and 11.15+ build incrementally.)
 
 | | Boot | Database | Owns the runtime |
 |---|---|---|---|

CHANGELOG.md

@@ -13,6 +13,7 @@ The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/).
 
 ### Fixed
 
+- **`test --attach` on Mendix 11.14.0 says why it is refused instead of blaming an entity** — that version's `mxbuild --serve` fails every build after its first, so since ako/mxcli#1917 the attach builds on a fresh serve process, whose first build asks for a runtime restart whatever changed. The refusal claimed "an entity or association changed" and suggested restarting the dev loop, which does not help; it now says the build could not be done incrementally and points to `--local`. The hosting app's pages survive either way. 11.15.0 builds incrementally and attaches as before.
 - **`describe microflow` no longer leaves out a note linked to the start event or to another note** (ako/mxcli#1933, mendixlabs/mxcli#1731) — such a note counted as linked, so it was not printed as a free note, and the element it is linked to prints no `@annotation`, so it was not printed at all; re-executing the description deleted it. It is now described as a free note, after a `-- WARNING` line saying the link itself has no MDL form and a rewrite drops it.
 - **Rebuilding a described decision no longer adds a merge** (ako/mxcli#1929, mendixlabs/mxcli#1727) — a `case`, `split type` or `if … elsif` one of whose branches returns is described with its own merge as a label (`end case; join shared1; merge shared1;`). Executing that as a fresh create built the decision's merge and a second one for the label, wired merge to merge, and every further describe → drop + create added another (`join junctionN; merge junctionN;`). The label's merge is now the decision's merge when nothing else reaches it, so the round trip keeps one merge.
 - **`describe microflow` prints a blank `call external action` argument as `= nothing`** (ako/mxcli#1915) — Studio Pro 11.16 stores an optional OData action parameter left blank as an empty argument, and `describe` printed it as `additional = )`, which does not parse, so every `mdl 1;` `create or modify` of such a microflow was refused (MDL-V1-REBUILD). `call external action` now takes `Param = nothing`, the blank-argument spelling `call microflow` and `call nanoflow` already use; `Param = empty` (what 11.14 stored) stays a different value, and each round-trips as itself.

cmd/mxcli/brain/list.go

@@ -0,0 +1,87 @@
+// SPDX-License-Identifier: Apache-2.0
+
+// list.go - the committed store as data.
+//
+// `brief --json` returns each shard's rendered markdown, which is right for a
+// prompt and wrong for a UI: rendering a tree from it means parsing a format
+// mxcli owns, a second parser of the same thing that drifts the first time the
+// format changes (ako/mxcli#1952). List hands out the parsed entries instead.
+package brain
+
+import (
+	"fmt"
+	"slices"
+)
+
+// ListFilter narrows a listing. The zero value lists everything.
+type ListFilter struct {
+	// Slice keeps one slice's entries: its requirements and the questions
+	// filed against it.
+	Slice string
+	// Kind keeps decisions or requirements. An open question is a decision
+	// unless it was filed against a slice.
+	Kind Kind
+	// Open, when set, keeps only open questions (true) or only settled
+	// entries (false).
+	Open *bool
+}
+
+// ListedEntry is one committed entry with the shard it is in — the same shape
+// `staged --json` gives a queued entry, so a consumer reads both sides alike.
+type ListedEntry struct {
+	Entry
+	Shard string `json:"shard"`
+}
+
+// ListResult is a listing plus the blocks that could not be read. Those are
+// reported rather than dropped: a UI that silently shows fewer entries than
+// the file holds is the failure a structured read exists to prevent.
+type ListResult struct {
+	Entries   []ListedEntry `json:"entries"`
+	Malformed []string      `json:"malformed"`
+}
+
+// List returns the committed entries matching f, in store order: project.md,
+// then modules, then slices, and file order within each.
+func (s *Store) List(f ListFilter) (ListResult, error) {
+	out := ListResult{Entries: []ListedEntry{}, Malformed: []string{}}
+	switch f.Kind {
+	case "", KindDecision, KindRequirement:
+	default:
+		return out, fmt.Errorf("kind %q: use %q or %q", f.Kind, KindDecision, KindRequirement)
+	}
+	if f.Slice != "" && f.Kind == KindDecision {
+		return out, fmt.Errorf("a decision has no slice, so --slice with --kind %s matches nothing; drop one of them", KindDecision)
+	}
+
+	shards, err := s.ListShards()
+	if err != nil {
+		return out, err
+	}
+	if f.Slice != "" {
+		want := PlanShard(f.Slice)
+		shards = slices.DeleteFunc(shards, func(sh string) bool { return sh != want })
+	}
+	for _, sh := range shards {
+		entries, malformed, err := s.LoadShard(sh)
+		if err != nil {
+			return out, err
+		}
+		for _, m := range malformed {
+			out.Malformed = append(out.Malformed, sh+": "+m)
+		}
+		for _, e := range entries {
+			// Explicit, so a consumer never has to know that an empty
+			// kind means a decision.
+			e.Kind = e.EntryKind()
+			if f.Kind != "" && e.Kind != f.Kind {
+				continue
+			}
+			if f.Open != nil && e.Open != *f.Open {
+				continue
+			}
+			out.Entries = append(out.Entries, ListedEntry{Entry: e, Shard: sh})
+		}
+	}
+	return out, nil
+}