Link checker blocks PRs that add content and link to where they would be

[heaths]

Recently, when trying to merge PR #14023, the link checker blocked me because my sample README.md contained a fully-qualified link (as guidelines state to use) to a sample file I was adding. Of course the file won't exist yet in master, but seems the link checker should maybe take into account the target of a PR and do some "magic" to verify the link will exist (e.g. get the relative link, and check that against the PR branch if targeting master).

Instead, I had to use a relative link and open #14060 to change it after, meaning I need two PRs to add content, which isn't very productive.

[weshaggard]

No matter what strategy we take you will require 2 changes to accomplish this scenario. In our recent guidance update at https://github.com/Azure/azure-sdk/pull/1587/files I suggest we provide no link the first time and then follow-up with the link in another PR once the change is in master. That way we don't end up with broken links between those 2 changes.

[weshaggard]

After reading your feedback a little closer it might also be possible to do some magic of ignoring links to files that are included in the PR that is targeting master but that isn't super straight forward and will also be difficult if you are targeting branches outside of master. It something we can consider but for now I still like starting with a commented out link and then following up to uncomment the link in a secondary PR.

[weshaggard]

@sima-zhu as part of the link work is is definitely worth exploring options to see if for PR's that are targeting master we could consider replacing master with the sha for the PR to see if the links resolve correctly and if they do then we can consider the link valid.

[sima-zhu]

Thanks @heaths to report this!

We can do one more check in Verify-Link.ps1 script.
When we read github link, we replace the master repository with development repository link and check whether it is broken. If it is broken, we failed the pipeline. If not, then we make it pass.
@weshaggard I will try on this approach first. Let me know if you have other suggestions.

[sima-zhu]

The change has merged into master. 6dfeb26
Closed the issue. Feel free to reopen if issue still exists.