Turn on Sphinx Link checker on CI?

[jenshnielsen]

The linkchecker that we use at the moment does not seem to check external links hence #5379 and others

I tried turning on the sphinx linkchecker python make.py linkcheck which seems to work but generate a fair number of redirect warnings and errors. Perhaps it is worth turning this on pr default?

See https://gist.github.com/jenshnielsen/b3dceef10df743154942 for the present output.

The linkcheck command is broken on python 3.5 but that should be fixed in the next Sphinx point release

[mdboom]

Thanks. I didn't realize we weren't checking external links, so this is great. Also solves the problem that linkchecker doesn't work on Python 3.x.

[tacaswell]

Huh, nifty that sphinx has that functionality built in....

[timhoffm]

How it works: Essentially, from docs run

make linkcheck

Open issues:

• When I ran this, github temporarily blocked me because of suspected abuse due to many page requests. Possible workaround add linkcheck_ignore = [r'https://github.com/.*'] to conf.py.
  Is too much traffic a general concern if we run this with with every CI build?
• There's a lot of output and the link checks must be filtered out. Either there has to be a way to silence sphinx mostly or one would have to grep the results.

[github-actions]

This issue has been marked "inactive" because it has been 365 days since the last comment. If this issue is still present in recent Matplotlib releases, or the feature request is still wanted, please leave a comment and this label will be removed. If there are no updates in another 30 days, this issue will be automatically closed, but you are free to re-open or create a new issue if needed. We value issue reports, and this procedure is meant to help us resurface and prioritize issues that have not been addressed yet, not make them disappear. Thanks for your help!

[story645]

Um is this still relevant?

[rcomer]

We have this in another project I'm on. The vast majority of the failures it generates are because some external documentation has moved, or because there is a temporary connection problem. So it creates CI failures in PRs that have nothing to do with the problem link. If a PR does introduce or modify a link, I would have thought the reviewers would check it manually (to verify the content is appropriate as much as whether the link works).

So I would suggest that if we want this we should have it on a schedule (similar to the nightly dependency tests) rather than within PRs.

[tacaswell]

I think it also takes ~hours to check the links in our docs (a lot of that is I think throttling to avoid getting flagged as a DDOS attack) so it is not obvious to me that we can even run our link checker on CI without getting timed out!