html.idl has become almost empty

[foolip]

In w3c/webref@7a622d9 the whatwg/idl/html.idl file became almost empty:
https://github.com/tidoust/reffy-reports/blob/7a622d9dbc3a8da6c61f76f95943eb90a6271007/whatwg/idl/html.idl

Looking at https://github.com/whatwg/html/commits/master and https://github.com/whatwg/html-build/commits/master and remains in html.idl now, I suspect some connection to whatwg/html#3677.

@tidoust, are there particulars about how IDL is extracted from HTML that might be causing this?

[tidoust]

The IDL extractor was a bit conservative and assumed that a given spec would follow one and only one convention to define IDL. This had worked fine until now.

whatwg/html#3677 introduces IDL definitions with a <pre class="idl"> convention in the HTML spec, different from the <pre><code class="idl"> convention used in the rest of the spec. That explains why the IDL extractor got confused.

I made the IDL extractor more flexible. It now tries all patterns. This fixes the issue for the HTML spec and does not change the outcome for all other specs. The html.idl file will be updated next time the crawler runs.

[foolip]

Thanks for the quick fix, @tidoust! I've also sent whatwg/html#3990 to make the style consistent, but making the scraper more robust is still a good idea I think.

[tidoust]

IDL extract now back to normal via w3c/webref@18044a1

Closing the issue as a result.