Repository navigation
Sync ako/mxcli: check catches CE1406/CE0114/CE6585 and version gates, revoke and OData fixes, run --listen-addr - #1508
Merged
Merged
Conversation
…etwork
`run --local` binds the app to 127.0.0.1 and there is no way to widen it:
LocalRuntimeOptions.ListenAddr is the only field that reaches the runtime's
`runtime_listen_addresses`, and nothing sets it but the default at
applyDefaults. So the app cannot be opened from another machine on the LAN —
the use case being a phone or a second laptop pointing at a dev box, which is
not exotic enough to leave unsupported.
It cannot be fixed after boot either. appContainerParams is sent once, at
startup (StartLocalRuntime); re-sending update_appcontainer_configuration on a
running app fails, because the runtime already holds 127.0.0.1:<port> and
Linux rejects a second bind of the same port on 0.0.0.0:
Runtime port '8095' is already in use
No permission to open socket or another server is already running on
Some(0.0.0.0):8095
So the address has to be right at startup, which means a flag.
The M2EE admin API is a privileged surface — it boots the app, rewrites
configuration and runs actions — so it must not be widened onto the network as
a side effect of widening the app bind. m2eeAdminListenAddr is a constant at
127.0.0.1, used for both the runtime's M2EE_ADMIN_LISTEN_ADDRESSES and mxcli's
own admin client. The client part is not tidiness: 0.0.0.0 is a bind address,
not a destination, so a client dialling it is not portable, and when the admin
bind was simply mirrored the readiness poll timed out with "runtime admin API
did not come up" on a runtime that was up and serving.
The default is unchanged at 127.0.0.1, so nothing behaves differently unless
asked.
Note for callers: the runtime derives ApplicationRootUrl from the bind address
when none is set, and 0.0.0.0 is not a host a browser can ask for. Pair this
with a reachable root URL — an ApplicationRootUrl naming a real host in the
model, or --runtime-setting ApplicationRootUrl=http://<host>:<port>/. The flag
help says so.
Closes #1248
…labs#1505) Studio Pro stores a multi-line note with CRLF line breaks and describe prints them raw. A script saved by an editor with LF no longer matched the stored free annotations, so `create or modify` under mdl 1 refused: "the free annotations change; the splice edits activities only". A note on an activity re-spliced that activity for the same reason. declaredMatches now compares note text with every line break spelled LF. A matched note is left as stored, so its CRLF bytes are kept and the run writes nothing. A real change to a free note is still refused, and the refusal now names the note, quoted with escapes so an invisible-character difference shows as one. The bug-test holds raw CR bytes and is marked -text in .gitattributes, since eol=lf on *.mdl would strip the case on commit. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012mdoMLzgHy7h8qbvVNkb92
…ir rules (mendixlabs#1501) `revoke write (NewAttr) on entity M.E from M.RoleA, M.RoleB, M.RoleC` printed "nothing to revoke", exit 0, and every rule kept NewAttr ReadWrite. The rules were granted one role each; RevokeEntityMemberAccess only touched a rule whose role set equalled the statement's, so none matched. It now narrows every rule all of whose roles are named (as a full revoke already did per role) and refuses, before writing, a rule shared with a role the statement does not name. Backend tests reproduce the 0-rules-modified symptom before the fix; the single-role revoke on the same project was the control that ruled out the reported default-member-access hypothesis. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VKmbj53BC2vuUWYmgWgfQ5
…ute (mendixlabs#1440) Reported (mxcli v0.26.0): alter snippet M.S { replace <widget> with { combobox x (Attribute: $Asset.Asset_Customer, CaptionAttribute: CustomerManagement.Customer.CustomerName) } }; check: "Check passed!" exec: "property attributeEnumeration binds Asset_Customer, which is an association ... not an attribute" exec is the correct side. With no DataSource the combo box is in enumeration mode, and attributeEnumeration cannot hold a reference (#830). The refusal existed only in the widget engine's Attribute mapping, and no check pass called it. check now asks the same question through the same function, and resolves the name against the same entity exec uses: - CREATE PAGE/SNIPPET: in the pluggable attribute-scope walk (checkWidget). `$Param.Name` resolves against the parameter's entity; a bare name resolves against the #647 binding scope. - ALTER INSERT/REPLACE: in the insertion-point pass, against the stored document's ParamScope and alterEntityContext. A combo box gets its own remedy, which names the association mode (`Attribute: $Asset.M.Asset_Customer, DataSource: ..., CaptionAttribute:`). On mx 11.6.6 that mode gives 0 errors, and describe -> exec reports "Unchanged snippet". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W3x8Fyvir8kLyPgJs1BDmF
) Reported: "File Uploader: an `allowedfileformat` in `configMode: 'simple'` passes check and exec, then fails `mx check` with CE0463 ... hidden TextTemplates inside object-list items are never nulled". The File Uploader editorConfig hides an allowedFileFormats item's `typeFormatDescription` (with mimeType/extensions) per item while that item's configMode is "simple". The extractor already lifted the nested rule, but ApplyVisibilityRules only walked top-level properties, so the required item TextTemplate kept the populated ClientTemplate the item builder gives it. `mx update-widgets` on a copy changes exactly that slot to null. ApplyVisibilityRules now applies nested rules per list item, resolving each condition in its own scope (item vs widget), and skips nested rules in the top-level pass so a colliding top-level key is never touched. Measured on Mendix 11.6.6 with File Uploader 2.5.1: CE0463 before, 0 errors after; describe -> exec -> describe is identical. The same fix clears the CE0463 on doctype-tests/32 Accordion `acc2` (custom-header groups hide the required headerText); fixtures 30 and 31 are unchanged. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W3x8Fyvir8kLyPgJs1BDmF
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NDLRXjMy1B18Cj3nX5RuCX
fix: a note's line endings are not a change to a spliced flow (mendixlabs#1505)
…r-access fix(backend): a partial revoke from several roles narrows each of their rules (mendixlabs#1501)
…le KEY (mendixlabs#1426) Reported: "Published OData with no KEY, or a KEY on an attribute without a unique rule, passes check" -- `mxcli check` said "Check passed!", exec wrote it, and mx check reported [CE6585] "Published entity 'ModA.EntA' must have a key defined." [CE6624] "Add a unique validation rule to attribute 'Name' of entity 'ModA.EntA' to be able to use it as the key." Measured on mx 11.6.6: no KEY is CE6585 for every entity kind (also `expose (*)` and no expose clause). A single KEY on a persistable entity needs a unique validation rule, on OData3 and OData4 alike; not null or an index does not satisfy it. A composite key, or a key on a view or non-persistent entity, needs no rule. New check rules MDL-ODATA08 (CE6585) and MDL-ODATA09 (CE6624), both errors, in ValidateProgram so exec's preflight refuses too. MDL-ODATA09 resolves the entity from the script, then the project, and stays silent when it cannot (unknown entity, inherited attribute, entity altered by the script). Two bug-test scripts that were themselves CE6585 are keyed. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W3x8Fyvir8kLyPgJs1BDmF
Review follow-ups to mendixlabs#1249: - AppURL is what mxcli dials (health check, screenshots, "App is running at"), but it was built from the bind address: 0.0.0.0 is not a destination and `::` produced the unparseable http://:::8080/. A wildcard now dials loopback of the same family, and IPv6 hosts are bracketed via net.JoinHostPort. - --listen-addr must be an IP address, refused before --detach and the build instead of as a bind failure in the runtime log. - A wildcard bind with no application root URL (model, hub or --runtime-setting) warns at boot, since the runtime derives one from the bind address. The network URLs the app is reachable at are printed after boot. - Flag rows in the run-local skill and docs-site page, CHANGELOG entry. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UBgg5k5sdMTsURKiigb19C
mendixlabs#1420) Reported (mxcli v0.26.0). A param-bound legacy reference selector in a snippet was replaced with combobox x (Association: M.A, DataSource: database M.E, CaptionAttribute: Attr) -- no data container check and exec both passed, and the build then failed: [CE7007] "Selected value is not valid for entity 'Asset_Customer/Customer'. Please select the value again." Reproduced on mx 11.6.6. The writer stores the association with no SourceVariable, which means "the enclosing object", and outside every data container there is no such object. Mendix can express the construct. Studio Pro stores this binding with SourceVariable = the snippet or page parameter. MDL already writes that from `Attribute: $Asset.M.Asset_Customer`, which gives 0 errors on 11.6.6 and which describe -> exec reports as "Unchanged snippet". The bare spelling is now refused, and the message names CE7007 and the spelling that builds. mxcli does not guess which parameter was meant. One decision (comboBoxAssociationNeedingObject) is used in three places: - exec: in the widget engine's Association mapping, when Build started with no entity in scope; - check without a project: new rule MDL-WIDGET46 at a document root; - check with -p: for an ALTER INSERT/REPLACE at an insertion point with no entity (unscopedInsertion). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W3x8Fyvir8kLyPgJs1BDmF
Port of mendixlabs#1229 (closes mendixlabs#1228) onto current main. `alter page P { set (RenderMode: H2) on <dynamictext> }` was refused as "not a property of this built-in widget", although CREATE PAGE and REPLACE accept RenderMode on a dynamictext and the MCP backend's mutator already sets it. setRawWidgetPropertyMut had no case for it, so it fell through to the pluggable-property setter. Add a RenderMode case dispatched on the stored $Type (Forms$DynamicText): the value is validated case-insensitively against Text, Paragraph and H1-H6 and written in its canonical spelling; anything else is refused with the accepted list, leaving the document untouched. Other widgets keep the previous path. Changes against the upstream PR: - docs, syntax help and the bug-test script use `set (Key: value)` instead of the deprecated `set Key = value` (MDL-DEPR101); the bug-test script is upgraded to the `mdl 1;` form so the conformance gate accepts it - the finding is its own file under findings/mdl-backend/ instead of an append to the merged-back mdl-backend.jsonl - the quick reference no longer claims every other widget gets an error naming the allowed set; only a bad value on a dynamic text does Tested: the new pagemutator tests fail on unmodified main and pass with the change. Bug-test script against a copy of testdata/pedapp (Mendix 11.13.0): exec applies it and describe shows H2, Paragraph, H4; the unfixed binary refuses the first alter. H7 is refused by check and by --no-check exec, with nothing written. make test, make lint and make check-findings pass. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01NDLRXjMy1B18Cj3nX5RuCX
…1613) Publishing a specialization and exposing a member its generalization declares passed check and exec, then mx check reported [CE1613] "The selected attribute 'OdInh.Sub.Code' no longer exists." for an inherited attribute and an inherited association alike. lookupEntityMembers read only the published entity's own attributes and the associations it is directly a side of, so an inherited member fell to the bare-name attribute fallback and the writer prefixed the published entity. Inherited attributes are now qualified by the entity that declares them (any depth, keeping their EDM type), and inherited associations publish as association ends with the specialization standing in for the generalization's side. Found by the mendixlabs#1426 fix, which could not measure an inherited KEY while this error masked the others. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W3x8Fyvir8kLyPgJs1BDmF
…ixlabs#1424) Reported: "Mxcli-written Data grid 2 fails CE0463 on a 10.x target until `fix widgets`" — a blank 10.24.0 app plus one `create page` with a DATAGRID gives CE0463 on the grid; `mxcli fix widgets` clears it. 10.24.0 ships Data Widgets 2.30.3 and the embedded template is 3.4-shaped, so AugmentTemplate rebuilds the definition from the .mpk. Against `mx update-widgets` output the Object was identical and the Type differed in three places, all augmentation defects: - PropertyDefs were indexed by bare key across nesting levels; 2.30.3 declares `filter` twice (column: widgets, filterList item: attribute), so the column's filter got the item's definition. All reconcile passes now address properties by scoped key ("columns/filter"). - Exemplar indices were taken before stale properties were removed, so the new `filterList` object list was cloned from an Integer property and had no ObjectType. Indices are now taken after removal (top level and nested). - <associationTypes> and selectableObjects were never parsed or reconciled (column filterAssociation / filterAssociationOptionLabelAttr). Also drop template Translations where the package ships none (drop-down filter emptyOptionCaption), which surfaced once the grids stopped failing. Measured: the report's script on 10.24.0 1 -> 0 CE0463, describe -> exec -> describe identical; doctype-tests/31 on 10.24.0 7 -> 0, on 10.24.24 2 -> 0, on 11.6.6 0 -> 0; doctype-tests/30 on 11.6.6 0 -> 0; doctype-tests/32 on 11.6.6 3 -> 2. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W3x8Fyvir8kLyPgJs1BDmF
fix: refuse a published OData entity with no KEY or a non-unique single KEY (mendixlabs#1426)
…tion-check fix: check refuses a combo box bound to an association through Attribute (mendixlabs#1440)
…et-param fix: refuse an association-mode combo box with no object to start from (mendixlabs#1420)
…bute fix: publish inherited OData members under their declaring entity (CE1613)
fix: null hidden TextTemplates inside object-list items (mendixlabs#1198)
fix: rebuild an older Data Widgets definition correctly on 10.x (mendixlabs#1424)
…5e8da feat(run): add --listen-addr so a local app can be reached from the network
Measured on 11.12.1 with a blank app (HSQLDB) and an after-startup microflow logging Core.getConfiguration().getApplicationRootUrl(): default (127.0.0.1) -> null --listen-addr 0.0.0.0 -> null 0.0.0.0 + --runtime-setting ApplicationRootUrl=http://192.0.2.2:8080/ -> http://192.0.2.2:8080/ So the claim, in #1249's help text and in two older comments, that the runtime derives its root URL from the listen address is false: unset stays unset. The wildcard warning, flag help, docs rows and CHANGELOG now say that instead. Behaviour is unchanged. Also verified live: under 0.0.0.0 the app answers 200 on the LAN address and loopback, the admin API only on loopback; under a single interface address mxcli dials that address (loopback refuses) and the admin API stays on loopback. `::` could not be measured: this container has no IPv6 (AF_INET6 unsupported), and Mendix reports the failed bind as "Runtime port '8080' is already in use". Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01UBgg5k5sdMTsURKiigb19C
…improvements-b5e8da
fix(pages): ALTER PAGE/SNIPPET SET RenderMode on a dynamic text (port of mendixlabs#1229)
…5e8da docs(run): the runtime does not derive a root URL from the bind address
bun audit fix, lockfile only: brace-expansion 5.0.5 -> 5.0.12, esbuild 0.28.0 -> 0.28.1, qs 6.14.1 -> 6.16.0, underscore 1.13.7 -> 1.13.8. bun audit: 12 vulnerabilities -> 0. Extension still compiles. Addresses Dependabot alert #23 (alert body not readable from the session; this clears every advisory bun audit reports for vscode-mdl). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XNquFqC4yEhzhY1LcCP5KE
npm audit fix --package-lock-only: 51 -> 49 advisories. The remaining roots (braces, sprintf-js, postcss-selector-parser) arrive via @mendix/pluggable-widgets-tools 11.12.1; npm's only offered fix is a downgrade to 8.18.4, so they are left for an upstream tools release. A placeholder-filled copy of the template still builds an .mpk. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XNquFqC4yEhzhY1LcCP5KE
sprintf-js <= 1.1.3 has no patched release (unbounded precision specifiers -> RangeError). Its only path was @istanbuljs/load-nyc-config -> js-yaml 3 -> argparse 1 -> sprintf-js, and js-yaml loads argparse only from its CLI, so it was never reachable; but with no fix upstream the alert can only close by leaving the tree. Override js-yaml to ^4.1.0 for load-nyc-config alone: v4 uses argparse 2 (no sprintf-js), and load() keeps its signature. Verified a .nycrc.yaml still loads through it and a placeholder-filled copy still builds an .mpk. npm audit: 49 -> 29. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XNquFqC4yEhzhY1LcCP5KE
…ot (mendixlabs#1433) `mxcli new --version 10.24.0` created a project whose first build failed with 4x CE6083 "Design property ... is not supported by your theme", then printed "Project created" and exited 0. The scaffold layout hardcoded Atlas 4's flex design properties on the topbar container. Atlas is versioned with the project: 10.24's template ships Atlas Core 3.17, which declares none of them. The topbar properties are now written only where the project's themesource declares them (executor.ThemeDeclaresDesignProperty). On Atlas 3 none are, and none are needed: its .topbar-content rule is already a centred flex row. A first build that mxbuild refuses for model errors is now a typed docker.ModelErrorsError, and `new` finishes its steps and exits 1 with the errors listed. Other first-build failures (no JDK, no mxbuild) stay warnings, since they say nothing about the project. Measured with mx check: 10.24.0 4 errors before, 0 after; 11.6.6 0 both, flex properties kept. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W3x8Fyvir8kLyPgJs1BDmF
Dependabot alert #21 (GHSA-rj75-hqrm-r3gf): flat selectors such as .a.a.a... parse in O(n^2). The widget lock held 7.1.5 (postcss-modules) and 6.1.4 (cssnano 5 plugins); no 6.x release is patched and `npm audit fix --force` proposes downgrading pluggable-widgets-tools to 8.18.4, so pin the patched version with an npm override. 7.0.0's only breaking change is safe insertion during iteration. The built widget is byte-identical to the baseline, and cssnano output on a selector corpus is identical under 6.1.4 and 7.1.6. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_012EqHXncQcCt3C8WJHHN2xJ
`mxcli check -p` passed scripts that `mxcli exec` then refused for the project's Mendix version (e.g. `create view entity` on a 9.24 project), after running the statements before the refused one. Every version gate lived inline in its handler as a checkFeature call, which neither check nor exec's pre-flight ever reaches. The gates now come from one statement-to-feature table, statementVersionGates (mdl/executor/version_gates.go). The handlers refuse through checkVersionGates(ctx, stmt); Executor.CheckVersionGates reports the same refusals as MDL-VERSION01, run by `check -p` and by exec's pre-flight, so exec refuses before anything is written. A test fails on any new direct checkFeature call outside the table. Still exec-only: gates that depend on stored state (drop/alter of a message definition document that is not there) and the write-time property floor. The issue's own StableId/CaseValues refusal was the latter and is already cured by #1410; the flows now pass both check and exec on 9.24. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01W3x8Fyvir8kLyPgJs1BDmF
fix(new): scaffold layout builds on Mendix 10.24; fail when it does not (mendixlabs#1433)
fix: check -p applies exec's version gates (mendixlabs#1434)
… it, MDL-SEC22 (mendixlabs#1500) No check modelled CE0114 "A microflow that does not apply entity access can only call microflows that also do not apply entity access" - a consistency rule between two documents, not a reference - so `check -p --references` passed and mxbuild refused. CheckEntityAccessCalls simulates stored and scripted microflows/rules (carriedApplyEntityAccess for absent annotations), reports new off->on calls and off->rule-on decisions as errors, stays silent on unresolved callees. Measured on mxbuild 11.13.0/11.14.0: reverse, on->on, off->off, nanoflow and excluded callers build clean, at Off and Prototype. Two variants on a fresh 11.14 project: issue script -> MDL-SEC22 exit 1, exec -> 1x CE0114; caller fixed -> check passes, 0 errors. Revert-proof: gate stubbed -> 6 tests fail "got 0"; callee test dropped -> off->off control fails. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VKmbj53BC2vuUWYmgWgfQ5
The index on main is stale (make check-skill-index fails on every branch); regenerated here so this PR's CI is green. A no-op once main carries it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VKmbj53BC2vuUWYmgWgfQ5
…ixlabs#1498) Nothing read Projects$ModuleSettings.ExportLevel and the entity read dropped ExportLevel, so check passed what mxbuild refuses as CE1406 (a Hidden element referenced from navigation / another module) and CE5400 (an API microflow exposing a Hidden entity). New validateProtectedModuleRefs in the --references pass; describe module/entity and show modules now say Protected / the level. Rules measured on mxbuild 11.13.0 and 11.14.0 with controls; tests fail with the validator stubbed and with the entity read reverted. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VKmbj53BC2vuUWYmgWgfQ5
The index on main is stale (make check-skill-index fails on every branch); regenerated here so this PR's CI is green. A no-op once main carries it. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01VKmbj53BC2vuUWYmgWgfQ5
…parser build(deps): force postcss-selector-parser 7.1.6 in vega-charts widget
fix(check): report a microflow without entity access calling one with it, MDL-SEC22 (mendixlabs#1500)
…vels fix(executor): check references into Protected (add-on) modules (mendixlabs#1498)
Conflict in the vega-charts widget: main added a postcss-selector-parser override. Kept both overrides and regenerated package-lock.json with npm (main's lockfile + npm install + in-range npm audit fix), not by hand. sprintf-js stays out of the tree; npm audit 19 advisories, all braces (no patched release). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01XNquFqC4yEhzhY1LcCP5KE
fix(deps): clear vscode-mdl npm advisories within existing ranges
AI Code ReviewReview SummaryThis PR syncs 23 commits from ako/mxcli:main, bringing several important improvements to mxcli's checking and validation capabilities, write correctness, and runtime features. What Looks Good
Moderate Issues Requiring Verification
Minor Observations
RecommendationApprove this PR. It delivers valuable improvements that close significant gaps between mxcli check and mxbuild behavior, enhances correctness of writes, and adds useful functionality. The changes follow architectural patterns, include proper testing, and maintain the project's evidence-based approach to fixes. Verification needed before merge: Confirm LSP wiring for new MDL features and user documentation updates for Automated review via OpenRouter (Nemotron Super 120B) — workflow source |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Syncs
ako/mxcli:maininto upstream: 23 commits from fork PRs ako#1373–ako#1389. Two of them port upstream contributor PRs onto currentmain: #1249 (--listen-addr, by Wengao Liu) and #1229 (dynamic-text RenderMode, by MendixMau). If this merges, those two PRs can be closed as superseded.check catches more of what mxbuild refuses
checkpasses builds that fail with CE1406 and CE5400 #1498). Check now catches what mxbuild refuses as CE1406 and CE5400.describe module/describe entityandshow modulesshow the export level.check -p --referencespasses a flow without entity access that calls a flow with entity access; mxbuild refuses it (CE0114) #1500): a microflow without entity access calling one with it (CE0114).Attribute:is refused at check, not first at exec (Combo box bound to an association through the attribute key passes check, exec refuses #1440).Correct writes
revoke write (<attribute>)says "nothing to revoke" on rules whose default member access is ReadWrite; the attribute stays writable #1501). It used to print "nothing to revoke" and leave every rule unchanged.fix widgets#1424). A Data grid 2 on 10.24 no longer needsfix widgetsto clear CE0463.allowedfileformatinconfigMode: 'simple'passes check and exec, then failsmx checkwith CE0463 — hidden TextTemplates inside object-list items are never nulled #1198). This fixes CE0463 on a File Uploader in simple config mode.create or modifyrefuse an LF script.set (Key: value)syntax.mxcli new --version 10.24.0scaffolds a layout that builds (mxcli new --version 10.24.0ships a layout that cannot build and still exits 0 #1433). It writes Atlas 4 design properties only where the theme declares them, and fails instead of printing "Project created" when the first build fails.run
run --local --listen-addrlets other machines on the network reach a local app. This ports feat(run): add --listen-addr so a local app can be reached from the network #1249. The M2EE admin API stays on loopback.Dependencies
These address the fork's Dependabot alerts.
vscode-mdl: in-range
bun audit fix(brace-expansion, esbuild, qs, underscore), lockfile only. The audit goes from 12 advisories to 0.vega-charts widget template:
npm audit fixsprintf-jsis dropped by overridingjs-yamlto v4 forload-nyc-configpostcss-selector-parseris pinned to 7.1.6 (GHSA-rj75-hqrm-r3gf), and the built widget is byte-identicalThe remaining advisories come in through
@mendix/pluggable-widgets-toolsand wait on an upstream tools release.Docs
/mxcli-dev:reviewgets a recurring finding about upstream PRs cut from a stale base.