Add back vnu-jar for HTML validation

[XhmikosR]

After discussing this with @bardiharborow, I think this is the best middle solution; if Java is available HTML validation will run, otherwise it will be skipped.

Also, this is a lot faster than the html-validator solution and we don't introduce another point of external failure (like with Sauce Labs).

TODO:

• Fix ignores
• Fix multiple ignores
• Combine ignores to an array instead of manually doing this?
• Update jekyll-toc when they fix the HTML validation issues upstream
• Verify ignores

Fixes #23171

[XhmikosR]

I believe this is ready. Perhaps we'd like to wait until the jekyll-toc errors are fixed upstream before merging this... or not.

@patrickhlauke: which of the other errors should be ignored?

Note that I'm ignoring warnings on purpose. We might want to revisit this later.

[bardiharborow]

@XhmikosR I think you should add Element "ul" not allowed as child of element "ul" in this context. as an ignore for now. Without further investigation I'm unsure about the other errors.

[XhmikosR]

I think I'll wait a couple of days since the jekyll-toc issue seems to be WIP.

We still need @patrickhlauke's feedback about the remaining errors and which we should ignore.

[XhmikosR]

Does anybody have any idea why ignoring specific errors doesn't work?

EDIT:

All right, after discussing this with @sideshowbarker on IRC, it appears the messages are inconsistent thus leading to this issue. grunt-html filters the errors differently and doesn't pass filterpattern thus the issue isn't present there.

I'll wait a bit for this to be addressed and we should be ready to merge in a few days, hopefully.

[XhmikosR]

@bardiharborow @Johann-S @mdo: I changed this PR to not remove htmlhint and just add vnu-jar which will run if Java is available.

What do you think?

There's also the need to verify what we ignored in this branch. @patrickhlauke: can you have a quick look?

[Johann-S]

LGTM 👍 just a question, so we will use your script only if folks have Java ? Not on Travis ?

[Johann-S]

If you return directly after that console.error you should throw an error instead

[XhmikosR]

We don't want to fail here; if Java isn't installed we are just skipping vnu-jar.

[bardiharborow]

@XhmikosR seems fine to me.

• Attribute “autocomplete” is only allowed when the input type is “color”, “date”, “datetime-local”, “email”, “hidden”, “month”, “number”, “password”, “range”, “search”, “tel”, “text”, “time”, “url”, or “week”. is likely intentionally non-standard.
• Attribute “autocomplete” not allowed on element “button” at this point. is likely intentionally non-standard.
• Attribute “title” not allowed on element “circle” at this point. I'm not sure about, likely an SVG issue (does the validator handle inline SVG correctly?).
• Bad value “tablist” for attribute “role” on element “nav”. I'm not sure about, patrickhlauke will likely know.
• Element “img” is missing required attribute “src”. is a false positive, due to how we use Holder.js.
• Element “legend” not allowed as child of element “div” in this context.* I'm not sure about, I'd have to run tests locally to find the context.

I don't have an issue with merging now and then removing some ignores later.

[XhmikosR]

@Johann-S: Travis has Java so it will run there. See also this PR's Travis builds :)

[XhmikosR]

I'm gonna merge this but I still think we need to go through our ignores.

For example, the error about title on circle seems legit to me and it's the issue I was telling @Johann-S the other day on Slack; this is non standard.