Special case normalization of help--symbol-completion-table

[clemera]

See #320

[minad]

@clemera Thanks for the quick fix!

[clemera]

I had a bad feeling to generally switch the buffer, as you said it doesn't match with the general assumption so errors are likely to happen

[minad]

I somehow had a bad feeling to generally switch the buffer, as you said it doesn't match with the general assumption so errors are likely to happen

Yes, I was worried too, also since you merged this very quickly. We are more safe now, I guess :)

[clemera]

Yes, I also think this is less likely to get us into trouble

[clemera]

This won't work for recursive session where the last buffer can't be obtained by minibuffer-selected-window we would need the buffer from the previous minibuffer session. Doing the computation before entering the next session for the descrbe symbol table would work but then this would produce slowdowns for embark with describe-variable + selectrum.

[minad]

@clemera please explain. Ist ist necessary to revert this?

[clemera]

No revert but can be improved. describe-variable is still slow for recursive sessions. We would need to obtain the last minibuffer session for recursive sessions or compute the candidates before the next session is entered (with the downside that this would slowdown embark interaction for describe-variable as described above).

[minad]

Shall we open a new issue to keep the improvement in mind?

[clemera]

I wanted to experiement if/how I can get the last session and fix it with some next PR but if I fail to do that I will open one.