bug#23871: 25.1.50; Undo unexpectedly leads to blank buffer

[Markus Triska <triska@metalevel.at>]

Hi Phillip,

phillip.lord@russet.org.uk (Phillip Lord) writes:

> I believe the following patch addresses the issue.

This patch causes a regression with the original case I posted in this
report (#23871):

Without your patch, after undo, point is *either*

(1) at the beginning of line 15 OR
(2) at a position that is outside the buffer (point 323 of 322).

Case (1) is the expected behaviour, and case (2) is the issue I reported
and which, just FYI, also still arises with your patch applied.

However, your patch also leads to the new problem that even in cases
where the point is not outside the buffer, it is unexpectedly placed at
the *end* of line 15 instead of at the beginning, where ediprolog-dwim
was invoked before the insertion is undone.

In this sense, the behaviour is worse than before for #23871 with your
patch, because point is never restored to where the interaction started.

All the best!
Markus