GIMP now has UI tests

GIMP has introduced automated UI tests for its desktop image editor, prompting reactions that range from praise for better stability to surprise that such a mature project is only now adding end-to-end interface testing. Much of the debate centers on GIMP’s explicit “no AI” contribution policy and whether refusing AI-generated code is prudent quality control or self‑sabotage in an era when competitors like ArtCraft’s tools are rapidly iterating with heavy AI assistance. Commenters also revisit long‑standing critiques of GIMP’s user interface and question its future relevance amid newer, more polished open-source and commercial alternatives.

Nature of the New UI Tests

  • Thread clarifies these are not basic unit tests; GIMP already had those.
  • New tests are near end-to-end UI tests: they open real windows, interact with dialogs/menus, and exercise the app like a user.
  • Several comments note that building a reusable UI test harness for a desktop app is the hard, valuable part; 13 tests likely reflect an early but important infrastructure step.
  • Desktop UI testing is seen as harder than web (no Cypress-equivalent), so this is viewed as non-trivial progress.

Debate over AI Use and GIMP’s No-AI Policy

  • Commenters discover a project policy forbidding AI use for contributors without a certain role, interpreted by some as a strict “no AI” stance for newcomers.
  • Some call this outdated and argue GIMP should “embrace the future,” while others see it as a pragmatic way to avoid floods of low-quality, AI-generated PRs and potential backdoors.
  • There’s speculation whether this policy also applies to blog posts, with some wondering if the announcement itself would violate the rule.

Quality and Purpose of AI-Generated Tests

  • Pro-AI side: recent models can quickly generate comprehensive UI tests, and some report thousands of tests (including UI) written with AI in short time.
  • Critical side: AI tends to mirror existing (possibly broken) behavior rather than infer correct intent; it produces redundant, low-value tests, asserts too many things per test, and misses important invariants.
  • Several emphasize that the hard part is designing orthogonal, meaningful test cases and making legacy code testable; AI mainly reduces typing, not conceptual work.
  • Some describe elaborate prompting / reflection workflows to improve AI output; others find “prodding” models tedious and prefer learning tools directly.

Perceptions of GIMP’s UI and Alternatives

  • Many criticize GIMP’s UI as unintuitive, “dinosaur,” or long-neglected, especially compared to Photoshop-like tools and newer AI-assisted apps (PhotoCraft, VectorCraft, Affinity with forthcoming Linux support).
  • Others defend GIMP as a long-lived, solid FOSS project; users who invested years in it report being more lost in commercial tools now.
  • Some feel GIMP has been irrelevant for their workflows for decades; others rely on it daily and expect to continue.

Community Dynamics and AI Evangelism

  • Multiple comments complain that any announcement (like UI tests) gets derailed by “why didn’t you use AI?” arguments.
  • One side characterizes strong AI promotion as quasi-religious and bullying, similar to past crypto hype; the other sees criticizing or debating AI in a testing thread as legitimate process discussion.
  • A few note ethical, environmental, or personal-learning reasons for avoiding AI, preferring human-built software and direct understanding over acting as a “meat proxy” for tools.