Travis TODO

[peternewman]

• Check Open Graph via html-proofer Support Open Graph gjtorikian/html-proofer#111
• Check tags when html-proofer supports HTML 5 HTML5 parsing and error checking gjtorikian/html-proofer#362
• Fail on broken external URLs Ignore the existing broken URLs #200
• Ensure people use the correct filename/folder format Fix some file/folder names #196
• Confirm the link validation will catch people giving their org folder the wrong name
• Switch to checking the owner/org in markdown - revert Use the original orgpath URL #198
• Turn Jekyll build warnings into failures https://travis-ci.org/pidcodes/pidcodes.github.com/jobs/195217329#L342
• Validate layout options e.g. https://travis-ci.org/pidcodes/pidcodes.github.com/jobs/195217329#L342 and astrojr1@91d3de5

[Arachnid]

FYI, #186 is passing despite the owner field being incorrect.

[peternewman]

It did actually catch it, but as an external link, which we currently ignore, unless you manually review them.

Before:
https://travis-ci.org/pidcodes/pidcodes.github.com/jobs/167023698#L1651

After:
https://travis-ci.org/pidcodes/pidcodes.github.com/jobs/167102370#L1572

[peternewman]

Still not right, see the last two commits here:
#195

[peternewman]

I've realised, it's not catching the broken org links, as the site is already coded to defend against them.

orgpath contains the actual, potentially incorrect URL https://github.com/pidcodes/pidcodes.github.com/blob/master/_layouts/pid.html#L2
But the link used on the page is org.url, which comes from https://github.com/pidcodes/pidcodes.github.com/blob/master/_layouts/pid.html#L3 and only matches if the two are the same already, otherwise you'll get nothing for org.url or org.title, but there will be nothing for the proofer to find either.

[peternewman]

Proved here:
https://travis-ci.org/peternewman/pidcodes.github.com/jobs/179990067

Are you happy for me to make the change @Arachnid ? It would only make a difference if something gets merged with a broken link.

[Arachnid]

Good catch - please do!

[peternewman]

Done, see #198 . To be undone once we get some better suited method of checking the fields, like the hex markdown checks will require.

[peternewman]

Confirmed the link validation will catch people giving their org folder the wrong name, sorted in #198

[peternewman]

Some possible validation options:
https://github.com/cityoffortworth/jekyll-data_validation
https://github.com/18F/jekyll_frontmatter_tests

[Arachnid]

I'm reasonably happy with verifying the generated HTML; is there a big disadvantage to doing that, or stuff we can't catch?

[peternewman]

For starters, 000a and 000A are different folder names, so you could get two people grabbing the same PID for example (perhaps even force a specific case to fix the sorting).

I'm sure once it's in place, there will be other benefits that become easy wins, like ensuring people have filled in the required fields. Perhaps even validating for a license from the various types.

[Arachnid]

Yup, case sensitivity and accidental collisions because of that is something I've been worried about for a while. Good point that it's fixable with more sophisticated validation.

[peternewman]

Likewise https://travis-ci.org/pidcodes/pidcodes.github.com/jobs/195217329#L342 , we can validate people are using valid layouts.