Thank you @mac-adder! I will have a look and think about it. I see why generating a new unique might be necessary but I have to think about the consequences.

So basically:

• If you find a default method you give it a new, unique name
• you store documentation for this new name
• when rendering and we encounter a default method, we uniquify its name again to see if there is some documentation.

This seems to be the plan, right?

Two other points:

• Is this also working for reexports?
• And could you add a test for default methods?

Sorry for the delay ! (I needed some time to set up an environment where I could actually test the changes on the latest versions)

So, unique names are necessary to ensure that comments applied to default signatures will be preserved (unless we change how comments are stored), and using setNameUnique / deriveUnique gives us "transparent" unique names, which greatly limits the number of changes to be propagated compared to other potential solutions (but I agree that it may not feel completely reliable, so I'm looking forward to receiving feedback about it).

Regarding reexports, it was actually not fully working, and should now be fixed, thanks !

Also, I've just added the new test for default signatures, tell me if it looks fine

@alexbiehl, I think it would be a good thing to have this in Haddock. Is there something more that needs to be done here?

Did I miss something here? Why was this closed?

Oh Boy I have no idea

Aah. I just deleted the master branch and this pr targeted it. Maybe?

Gotcha. I'm definitely interested in getting this into Haddock, so maybe I'll try to rebase this onto the ghc-8.4 branch when I get a chance.

This PR was closed accidentally, see this comment for how to reopen.