qunitjs / qunit / 10126824269

90%

push

github

Core: Convert "No tests were run." from fake test to error

* Remove hacky re-entrance from
  ProcessingQueue.done() -> test() + advance() -> done(),
  existed only for this purpose.

  This also removes the need for the 99aee51a8a workaround, which
  avoided a crash by infinite loop.

* Remove unused internal `test` injection to ProcessingQueue,
  existed only for this purpose.

* Remove "omit stack trace" logic in test.js,
  existed only for this purpose. To keep output for the "No tests"
  error similarly clean and distraction-free, the TAP reporter
  treats error stack traces with a similar cleaner since
  https://github.com/qunitjs/qunit/pull/1789.

* Remove unused internal `validTest` mechanism
  existed only for this purpose.

  This was originally impossible to trigger externally because it
  required setting `validTest` to a private symbol. In QUnit 1.16
  this was simplified as part of commit 3f08a1aa1e, to take any
  boolean true value to ease some implementation details, however it
  remained internal in purpose. A search today for `/validTest:/`
  and `/validTest = /` over public GitHub-hosted repositories, shows
  that (fortunatley) nobody has started relying on this. I found only
  copies of QUnit itself.

As a nice side-effect, fixtures like async-module-error.tap.txt
now no longer display a useless "No tests" after an uncaught error
that already bailed the test run. This happened previously because
errors are not tests, and so technically there was no test. No that
"No tests" is itself considered a bail out, TAP absorbs this after
the first error, just like it alreayd does for other cascading
errors.

Closes https://github.com/qunitjs/qunit/pull/1790.

7 of 7 new or added lines in 3 files covered. (100.0%)

1848 of 2049 relevant lines covered (90.19%)

3463.82 hits per line