On 2024-03-10 03:39, Eli Zaretskii wrote:
> OK, so let's wait for Paul to chime in.

The problem was that I mistakenly believed the documentation when it 
said that a symbol with position behaves like its bare symbol when 
symbols-with-position-enabled is t. Unfortunately it appears that this 
part of the doc wasn't intended to apply to bare-symbol, so when I fixed 
something else involving bare-symbol I got the semantics wrong.

As penance I installed the attached, which makes a simple code change 
along the lines that you suggested and adds a regression test to help 
prevent this bug from happening again.

The hardest part of writing this patch was adjusting the documentation 
to match what I think was the intent of the behavior. Alan, if you find 
mistakes in that please let me know.

A couple of other things.

Currently (position-symbol 'x -1) creates a symbol with position where 
the position is negative; is that intended? The documentation says 
positions are nonnegative.

Also, more test cases of the symbols with position primitives would not 
go amiss. I'm not a good person to write them, though, as I easily get 
confused by symbols with position.