test: links: fix failures

[casperdcl]

• fix dead links detected by test: check links #958
  • see daily workflow builds on https://circleci.com/gh/iterative/dvc.org (e.g. https://circleci.com/gh/iterative/dvc.org/2264) for an up-to-date list
  • fix/replace https://en.wikipedia.org/w/skins/Vector/images/external-link-ltr-icon.svg
    • do we really need this link? It's in a code comment
    • from 59e399d by @iAdramelk
  • fix links in exclude-links.txt
  • sort exclude-links.txt
• should warnings (status codes 0xx, 3xx etc) be counted as errors instead?
• fixes fix broken links #974
• continues Mostly link-related (regular) updates #978

[jorgeorpinel]

What's the logic behind the sorting?

[casperdcl]

er just piped though sort. Makes it easy to see diffs in future changes (was hard to see real changes in https://github.com/iterative/dvc.org/pull/978/files#diff-01c133d99b04d1390925d3d06f3bbe77).

[casperdcl]

yay re-ran daily and succeeded: https://app.circleci.com/jobs/github/iterative/dvc.org/2292

[shcheklein]

I'm thinking if we should create an exclusion list on file level? (don't analyze tests?). To avoid adding a lot of links to the existing file? what do you think @casperdcl @jorgeorpinel ?

[casperdcl]

Well we could do comments both in js and markdown similar to # noqa in py/flake8

[casperdcl]

You mean this sort of thing?

[md]http://dead.link <!-- noqa -->

'http://dead/link' // noqa

[shcheklein]

more like specify somewhere that I would like to exclude tests/** ... .gitinore-like interface I would say?

[jorgeorpinel]

exclusion list on file level... specify somewhere that I would like to exclude tests/** ...

Yes, only certain js and the md files really matter here, so I would even make it a white list with a pattern to match (similar to what prettier accepts, maybe hard to implement though). The existing list could be scripts/.link-ignore and the path list in scripts/.link-config or even sent as arguments to link-check.sh from package.json

Would also be nice if these files can ignore # comments (and blank lines) to make them self-documenting.

Ideally if it could actually render all web pages and only look at actual <a href=" and <img src=" URLs, that would be great, but probably overkill to do a separate yarn build just for this.