bug#16411: undo-only bugs

[Stefan Monnier <monnier@iro.umontreal.ca>]

>> undo-in-region does not implement undo-only, indeed.  I think the way
>> to implement undo-only is to change undo-make-selective-list so it
>> also skips redo entries (depending on undo-no-redo, obviously) using
>> the above while loop.
> I don't see how that's a correct solution in the face of arbitrary
> regional undos and prefix args. Would you have the redo records
> resulting from regional undos still map to t in the undo-equiv-table?

I was talking about simply making undo-in-region
properly skip the previous undos (presuming they don't themselves come
from an undo-in-region).

IIUC, you're talking of skipping (e.g. in a non-region undo) the undos
that took place during undo-in-region, right?  If so, I don't have an
answer for how we could do that, in general.

In your pseudo-code:

  While pending-undo-list non nil:
    Pop undo-i from pending-undo-list:
    If undo-i is a redo record (exists in the table):
      Remove undo-i's undone records from pending-undo-list (or
      otherwise arrange to skip it)
    Else:
      Undo undo-i
      If "undo undo-i" was done prefix-arg times:
        Break (finished)
  Reached end of undo history

I have no idea how to implement the part:

      Remove undo-i's undone records from pending-undo-list (or
      otherwise arrange to skip it)

I guess I do have some idea how to do it, but it looks like a lot of
work, since we have to adjust the positions in the rest of
pending-undo-list.

Other than that I don't understand what your redo-record-table does.
AFAICT the test "undo-i is a redo record" can be performed with
undo-equiv-table.

> How does your solution account for redo records that undid several
> because of prefix-arg?

As you have discovered the current code does not even try to account for
prefix args.

> undo-equiv-table only maps to the change group
> after the last undone record without information about the (1-
> prefix-arg) others.

Right: the loop that undoes N steps (either in undo-more or in undo if
we change undo to only call undo-more with a 1) needs not only to use
undo-equiv-table at each iteration to skip redo entries, but it also
needs to add an entry in undo-equiv-table at each iteration.


        Stefan