Files
Dario Gabriel LipicarandClaude Opus 5 1f95a75f83 feat: own the view-plugin templates, and the check that proves they work
This repo already declared itself the owner of the ui_qml authoring flavour
("everything needed to build a Logos view module ... end to end"), but the
LogosView*.in templates that logos_module(REP_FILE ...) actually instantiates
were live in logos-plugin-qt, and the copies here were stale duplicates nothing
read. Adopt the maintained versions and publish them.

logos-plugin-qt's argument for holding them (cmake/README.md) was that
logos-module-builder depends on it one-way, so it was "the only place both
consumers can read from". That only weighed two candidates. This repo inputs
logos-nix alone, so every consumer can depend on it without a cycle — and
logos-plugin-qt is meant to handle exactly one thing: making a cdylib module
loadable by logos-module-loader-qt. View authoring is not that.

- cmake/LogosView*.in: replaced with the maintained versions, keeping the
  ABI-agreement comments (they document a contract CHECKED by
  logos-module-builder's view-interface-abi against the runtime's host headers).
- packages.<system>.logos-view-templates: same output name logos-module-builder
  looks up.
- cmake/README.md: rewritten for this home.
- tests/rep-file-plugin + the rep-file-plugin check: moved here, standing on
  this repo's own Qt inputs rather than logos-plugin-qt's common.nix, so no
  edge is added and this repo stays a leaf. It is the binary-side half of the
  view ABI guard, and it now has CI in the repo that owns it.

One correction against the source it came from: the emitter that puts
LogosViewPlugin into a module's Q_INTERFACES is logos-qt-sdk's
`logos-qt-generator --backend ui`, not this repo's logos-view-generator. The
two are near-duplicates, but modulePreConfigure.nix runs the former, and
view-generator currently has no consumer at all.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
2026-08-19 15:39:18 -03:00
..