mirror of
https://github.com/compiler-explorer/compiler-explorer.git
synced 2026-09-10 16:57:56 -04:00
406b628a11
The Cypress harness starts the dev server with `--language c++`, so the browser only ever sees C++ and CMake. Any spec that needs to exercise behaviour *across* two languages therefore cannot run at all. This blocks the chained-compiler tests in #8678, whose entire purpose is compiling one pane's output as a different language. Its `beforeEach` selects C for the downstream pane, and on a C++-only server it times out looking for an option that does not exist. The suite does not merely fail, it is **skipped wholesale**, because the failure happens in the hook rather than in a test. That includes the one test written specifically to answer a question asked during review, so the review has been stuck waiting on an answer the harness could never produce. `--language` is variadic (`--language <languages...>`), so adding C is a one word change. ### Why nobody has noticed `test-frontend.yml` is `on: [push]` and guarded by `if: github.repository_owner == 'compiler-explorer'`. For a pull request from a fork the push event fires in the fork, where the owner check fails, and the workflow does not run for pull requests in this repo at all. So Cypress has never run against #8678. The failure would first appear on main, after merge. That guard is worth a separate look, but it is not what this PR changes. ### Verification Ran the full suite locally against this branch with the new flag, using the same command the workflow uses: ``` 14 specs, 91 tests, 91 passing, 0 failing (8m23s) ``` And against #8678's branch, the chained-compiler spec goes from 2 passing / 1 failing / 5 skipped to **8 passing** once C is available. 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-authored-by: mattgodbolt-molty <mattgodbolt-molty@users.noreply.github.com> Co-authored-by: Claude Opus 5 <noreply@anthropic.com>