No dependency resolution for can_macroexpand + external macro

[damourChris]

A function definition that use the @htl macro and has a docstrings attached to it cannot be called before the definition as the cell isn't ordered properly.

You can see in MWE that the cell with the call to the bad function is located
before the function definition and does not work. When placed after, it does.

Note that this only seems to be when combining @htl and docstring.

Minimum Working Example

Tested in Pluto v0.19.38

[fonsp]

Thanks for writing a MWE!

[geoextra]

I was facing the same issue and I think go introduced in a recent version. In my git history I can see that my notebook file got "corrupted" without having docstrings added.

[fonsp]

@Pangoraw Do you see a way to fix this? Otherwise let's just revert #2797

[Pangoraw]

@Pangoraw Do you see a way to fix this? Otherwise let's just revert #2797

It is interesting because it should carry on with the macro resolution but it doesn't. Maybe we can revert for now?

I think the problem is that we don't mark failed ExEx.try_macroexpand as unresolved so they are never resolved. See #1388 (comment). Another example with @bind + @u_str:

Screencast.from.01-03-2024.14.04.50.webm