empty line after jldoctest invocation causes errors

[ajinkya-k]

I noticed in a recent PR review that an empty line after ```jldoctest causes issues.

```jldoctest

julia> do something here 
```

will cause an error but removing the newline fixes it:

```jldoctest
julia> do something here 
```

Here are two live examples:
https://github.com/JuliaStats/GLM.jl/actions/runs/14439279742/job/40485771112?pr=487#step:5:318

https://github.com/JuliaStats/GLM.jl/actions/runs/14439279742/job/40485771112?pr=487#step:5:391

I could reproduce these errors on my local machine, and can confirm that removing the new line indeed fixes this issue.

[fingolfin]

Thank you for this report.

As I see it, there are the following options to deal with this:

1. consider this as "work as intended", and close it
2. like 1 but also clarify this in the documentation
3. like 1 but also ensure a helpful error message is produced to allow quickly diagnosing and resolving the issue
4. changing things to accept this kind of input

Of course options 2 and 3 can (and probably should) be combined.

For option 4, I think this needs some thinking to decide what the specific semantics should be. Just: "Ignore all empty leading lines"? But what about leading lines which are not empty but only consist of whitespace? What about leading lines that contain comments?

[ajinkya-k]

Thank you for this report.

As I see it, there are the following options to deal with this:

1. consider this as "work as intended", and close it
2. like 1 but also clarify this in the documentation
3. like 1 but also ensure a helpful error message is produced to allow quickly diagnosing and resolving the issue
4. changing things to accept this kind of input

Of course options 2 and 3 can (and probably should) be combined.

For option 4, I think this needs some thinking to decide what the specific semantics should be. Just: "Ignore all empty leading lines"? But what about leading lines which are not empty but only consist of whitespace? What about leading lines that contain comments?

Thanks for the thoughtful response! While 4 would be ideal, I'd be happy with 2 and 3, I.e. a better error message. Right now what happens is that documenter associates the expected output for one command to a completely different command, which is obviously the wrong behavior.