bug#60464: 29.0.60; Regression - pcomplete-arg fails with argument 'last

[Daniel Mendler]

On 1/4/23 15:36, Gregory Heytings wrote:
> 
>>> IOW, instead of unconditionally returning arg, we now sometimes (when 
>>> arg is not already a string and when index is not last) return the 
>>> textual representation of that argument (what the user typed in) with 
>>> the value of arg attached to it.
>>
>> I don't understand the `last` condition.
>>
>> Could you explain it (e.g. with a concrete example that breaks if you 
>> don't include it)?
>>
> 
> Yes: when index is 'last' returning a non-string (more precisely: a list 
> of strings) is allowed (and in fact expected, at least by Eshell).  A 
> recipe, in the Emacs repository:
> 
> M-x eshell
> cd lisp
> ls *.el TAB
> 
> (Note that there is no SPC before TAB.)
> 
> Without the 'last' condition, you don't get what you're supposed to get at 
> that point (and what you got with Emacs 27 or 28), namely the list of .el 
> files in *Completions*.  Instead you see "Complete, but not unique", and 
> you have to press TAB a second time to see the completions.

Don't you think that this is by accident? Shouldn't Eshell fixed instead
to not expect a list in this case? I believe it would be better if
pcomplete-arg behaves uniformly for all arguments.

Daniel