It takes our "build docs and audit them" action a long time to run, and a large chunk of this is because we run make linkcheck on all of our docs for every PR. This is probably a bit excessive and we certainly don't want to block somebody's PR because an old link has gone stale.
A pattern that has worked well with 2i2c is to use this re-usable workflow to check the links and open an issue with any broken ones. This will re-run each week and update a pre-existing issue if one is already open.
That way we can remove stale link checks and speed up the PR process. It means we should be careful that somebody doesn't introduce a broken link in a PR, but if that happens then the linkchecker will catch it relatively quickly anyway.
It takes our "build docs and audit them" action a long time to run, and a large chunk of this is because we run
make linkcheck
on all of our docs for every PR. This is probably a bit excessive and we certainly don't want to block somebody's PR because an old link has gone stale.A pattern that has worked well with 2i2c is to use this re-usable workflow to check the links and open an issue with any broken ones. This will re-run each week and update a pre-existing issue if one is already open.
That way we can remove stale link checks and speed up the PR process. It means we should be careful that somebody doesn't introduce a broken link in a PR, but if that happens then the linkchecker will catch it relatively quickly anyway.