Conversation
|
I have definitely thought the same. This should make the spec test system easier to scale. I'm wondering if this needs a separate type, or if we could get away with using There's actually already a new class of tests merged into branch-0.14, which would be the next release. Could you change your merge target to that and add the new class introduced in #1043? (We have been doing branching like this, where |
|
I started a thread about branching in the old Zulip: |
sure! I'll do that.
Yeah, I could use |
c016b49 to
6479cf9
Compare
|
@ollpu I have updated the generated tests to use |
|
Thanks, I will take a closer look soon. We clarified our branching strategy. |
|
Sorry, it's been a while. I have an alternative approach in #1122, which requires no changes to the test generator when adding new Options, and there are no global defaults. |
|
No worries! Feel free to close this PR if this is no longer needed. |
Problem
While exploring how to implement my own extension, I noticed that adding a new argument to
tests::test_markdown_htmlcauses 1000+ lines of changes, because every test cases requires an additionalfalseargument. For example, in PR #991, more than half of the diff is just adding an extrafalsetotest_markdown_html.So I think it is a good idea to refactor the test generation to reduce the diff noise.
Solution
Introduce a configuration struct (with
#[derive(Default)]) to hold all test options. This allows new options to be added without modifying every single test invocation, only the tests that actually depend on the new option need to be updated.