Skip to content

RX1: don't flag Atlas_Core's stock theme-folder import as a FILES-route risk - #12

Merged
MendixMau merged 1 commit into
mainfrom
feat/scss-boilerplate-guard
Nov 17, 2026
Merged

MendixMau merged 1 commit into
mainfrom
feat/scss-boilerplate-guard

Conversation

@MendixMau

Copy link
Copy Markdown
Owner

Summary

  • A held-modules audit on a real 10.24 to 11.12 run flagged Atlas_Core's own @import "../../../theme/web/custom-variables" / exclusion-variables lines as a hand-edited vendor file at FILES-route risk. The same two lines are Mendix's own module-template boilerplate, present identically in 46 of the app's 52 themesource modules, including modules already confirmed to carry no other edits — an Atlas_Core update regenerates them every time.
  • Adds bin/scss_boilerplate_check.py: given a themesource module, classifies each outward theme/web/ import as BOILERPLATE (the two stock names, no route needed) or RISK (anything else: a real app edit, still routed FILES as before). Catch + pass case wired into controls/run-controls.sh as scss-boilerplate-pass / scss-boilerplate-app-edit-risk-catch.
  • Second lesson from the same audit (RX1-2, doc only): it read HTMLSnippet as still on 3.9.9 from an archived run's decision log; the real delivered .mpk had already migrated every instance to Events widget + a bridge nanoflow / native HTML element. Added a caution to playbooks/widgets.md's HTMLSnippet section: judge widget-version claims from the shipped model, not an archived decision trail.
  • FEEDBACK.md: new RX1 glossary entry, rows RX1-1/RX1-2, and the running counts table updated (686 -> 688 rows).

Test plan

  • bin/scss_boilerplate_check.py run directly against a synthetic pass-case module (only the two stock imports) and a catch-case module (stock imports + one app-specific outward import): exit 0 / exit 1 respectively, correct BOILERPLATE/RISK lines.
  • controls/scss_boilerplate_control.sh pass and ... catch run standalone, output matches the ctl lines added to run-controls.sh.
  • Full run-controls.sh suite not run end-to-end in this PR (it has slow/interactive fixtures unrelated to this change); the two new ctl lines were verified by running the control script directly and checking its output against the exact regexes used.
  • customer_names_check.sh-style manual grep and the secret-pattern scan run against every changed/new file: clean.

No app code, customer name, or real app name appears anywhere in this PR (code name RX1 used throughout, per CONTRIBUTING.md).

🤖 Generated with Claude Code

…te risk

An audit on a real 10.24 to 11.12 run read Atlas_Core's own
"@import ../../../theme/web/custom-variables" as a hand-edited vendor file at
FILES-route risk. The same line is Mendix's own module-template boilerplate,
present identically in 46 of the app's 52 themesource modules, including
modules the audit had already confirmed carried no other edits. An Atlas_Core
update regenerates it every time; it's not what an update wipes.

Add bin/scss_boilerplate_check.py: classifies an outward theme/web/ import as
BOILERPLATE (custom-variables / exclusion-variables: stock, no route needed)
or RISK (anything else: a real app edit, route it FILES same as before).
Catch and pass case wired into controls/run-controls.sh.

Second lesson from the same audit (RX1-2): it read HTMLSnippet as still on
3.9.9 from an archived run's decision log, but the real delivered .mpk had
already migrated every instance to Events + a bridge nanoflow / native HTML
element. Note added to playbooks/widgets.md: judge widget-version claims from
the shipped model, not an archived decision trail.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
@MendixMau
MendixMau merged commit 485728e into main Nov 17, 2026
1 check failed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant