RE: dabbrev-completion doc string

[Drew Adams]

       The last sentence of the doc string should be removed, IMO:

         "With no prefix argument, it reuses an old completion list
          if there is a suitable one already."

    It would seem very odd to explain in detail what happens if you use
    various prefix arguments and not say anything whatsoever about what is
    going to happen if you use no prefix argument.

I agree. I should have said "replaced", not "removed".

         "With no prefix argument, it reuses an old completion list
          if there is a suitable one already."

       This is a user command, and (IIUC) the last sentence means
       nothing to users (it just documents the implementation).

       If I misunderstand the importance of the last sentence, then
       please consider rephrasing it in user terms.

    Unless I am the one misunderstanding the last sentence, the sentence
    you quoted and want to delete seems completely self-explanatory.  How
    can it be phrased more clearly?  Without a prefix argument, if there
    is already a completion list, `dabbrev-completion' just uses it
    without trying to update it (meaning it may not be up to date).

Again, it's quite possible I misunderstand the sentence.

I would expect it to say something about the set of buffers that are
searched to come with the candidate completions, which is what the two
non-nil ARG cases explain.

IIUC, a nil ARG means to search only the current buffer. If that is the
case, then the ARG might be better called BUFFERS, as it indicates which
buffer(s) to search.

I don't see why a user would care whether or not an existing completion list
(whatever that might mean) is reused. Does that mean that it uses the same
set of completions that the user saw last time? What if the user used a
prefix arg last time?

Perhaps this all hinges on the meaning of "if there is a suitable one
already" - just what does that mean? I suspect that explaining what
"suitable" means here, in user terms, would mean explaining that only the
current buffer is to be searched. But I could be wrong.