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

[Daniel Mendler <mail@daniel-mendler.de>]

On 1/4/23 16:39, Gregory Heytings wrote:
>> Your implementation with the 'last condition doesn't make sense.
> 
> Can you please try to make it fail?

That was not my point. The way the function behaves now is problematic
on theoretical grounds.

>> It is not a good idea to treat 'last specially such that we return a
>> different type in this case.
>>
> 
> I'm not the one who wrote that code.  According to my analysis of the 
> code, the only place where pcomplete-arg is called with a 'last' argument 
> is pcomplete-parse-arguments, where you will see the following:

I am not blaming you for it. If your analysis is right (and there are
not many or even only a single caller), what about deprecating support
for 'last and 'first then, if these lead to problems? Stefan said before
that Pcomplete needs some refactoring and fixing so one could do it in
small steps. We could then fix the few call sites which are there. But
such changes may be too late for Emacs 29.

Daniel