Opened 3 years ago

Closed 3 years ago

#19363 closed task (wontfix)

We should check for dead links in the website

Reported by: irl Owned by: irl
Priority: Medium Milestone:
Component: Webpages/Website Version:
Severity: Normal Keywords:
Cc: Sebastian Actual Points:
Parent ID: Points:
Reviewer: Sponsor:

Description

Following on from #19359, I suspect that similar problems are present in the website.

W3C have a link checker utility that could be run from a Jenkins job.

http://search.cpan.org/dist/W3C-LinkChecker/

I have a local Jenkins instance to experiment with this. If it proves to be effective/useful we can consider moving this to jenkins.tpo.

Sebastian noted in IRC that some failures may be temporary and each failure would require individual investigation. In any documentation we should note that we should never automate fixes, even where there are permanent redirects in place (a redirect may occur simply because the content doesn't exist anymore and it just takes you to a home page for instance).

Child Tickets

Change History (4)

comment:1 Changed 3 years ago by Sebastian

Is this in the right component? What's the status of the jenkins instance?

comment:2 Changed 3 years ago by irl

For now this is the correct component. Once I've got the Jenkins I'll do a child ticket for the jenkins component. This ticket will then have to document what is going on somewhere (wiki page update probably).

The current jenkins job is at:

https://jenkins.erg.abdn.ac.uk/job/tor-website-linkcheck/

But I need to see where there are false positives.

comment:3 Changed 3 years ago by irl

Status: newaccepted

comment:4 Changed 3 years ago by irl

Resolution: wontfix
Status: acceptedclosed

This didn't go anywhere, and given current time constraints I'm unlikely to take it anywhere in the near future. This could be reconsidered after the redo of the website, but for now I'm closing this ticket.

The process I was using for this was a webwml checkout, followed by a build and then running a local web server and running a link checker over it. This threw up all sorts of false positives where things were not built from webwml and also without content negotiation configured, as is expected I believe by the former translation infrastructure, there were false positives where URLs pointed at languaged versions of pages which again were not served.

When there is a better way of doing this, I believe it would warrant a new ticket.

Note: See TracTickets for help on using tickets.