Checking for broken links with TravisCI

[peterjc]

See #19 for fixing all the existing broken links.

Once that's done, we should add a link checker to TravisCI to keep appraised of any broken links (due to external site changes) or bad links in incoming pull request.

@JoaoRodrigues suggested following https://www.jacobtomlinson.co.uk/jekyll/2015/02/18/test-you-jekyll-blog-with-travis-ci/ which suggests this for the .travis.yml file:

language: ruby
rvm:
  - 2.1
script:
  - bundle exec jekyll build
  - bundle exec htmlproof ./_site --only-4xx --check-favicon --check-html
env:
  global:
    - NOKOGIRI_USE_SYSTEM_LIBRARIES=true # speeds up installation of html-proofer

[souravsingh]

@peterjc We can make use of linkchecker for checking broken links https://wummel.github.io/linkchecker/

[peterjc]

@souravsingh have you tried using linkchecker under TravisCI to check local files which GitHub Pages and Jekyll will generate?

This repository seems to use it, https://github.com/mingwpy/mingwpy.github.io/blob/source/.travis.yml

There's an open issue on the link checker repository about how best to do this (or indeed, if we should do this): https://github.com/wummel/linkchecker/issues/573

[souravsingh]

@peterjc Should we go with using linkchecker or using the solution @JoaoRodrigues suggested? I think the latter solution is the best.

[peterjc]

@souravsingh You can try either.

I suggest you try installing and running the link checker on your own computer, and if you find any broken links please submit fixes.

Then you'd need to test this works via TravisCI. If you fork Biopython on GitHub, you should be able to turn on TravisCI for your fork. Once you have something working please make a pull request to the main Biopython website repository (which will run TravisCI again via the Biopython account).

[souravsingh]

@peterjc I have run the linkchecker on the website. Here is the list of dead links-https://gist.github.com/souravsingh/cfa0a363bfdad416e89dfe807078d99a

[peterjc]

@souravsingh see #19 for dead links (and #7 for the releases).

Note there are some false positives in the linkchecker output, e.g. ConnectionError examples probably need a retry. Perhaps there are some linkchecker options you can use here, as otherwise this will not be useful for TravisCI testing.