bug#23906: 25.0.95; Undo boundary after process output is not consistent

[Markus Triska <triska@metalevel.at>]

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

>> about it, but what I think would work in this particular case is to
>> record it as "Emacs, to undo this, you need to remove all text between X
>> and Y", and Y simply increases as more text is inserted into the buffer.
>> This would keep the length of the undo list constant throughout the
>> whole interaction with the subprocess, at least in this concrete
>> case. Can I do this in ediprolog myself? If so, then I will do it.
>
>
> And what happens if the prolog process spams and never stops? Y, gets
> longer and longer. Eventually, Emacs signals an error with a "this is
> probably a bug" message.

The way I imagine it, Y does not get "longer and longer", but is simply
an integer that gets higher and higher when more output arrives. That
would be one way to merge subsequent insertions in the undo list.

> I think you are doing two things -- one is telling the user that
> something has happened and one is storing the result. I think that this
> is the problem.
>
> My suggestion is add the %@ into the buffer when the result comes
> through. If you want %@ to *appear* before this, then insert an overlay
> with a display string.

Please let us consider the more general issue of making *all*
insertions, the first _and also subsequent ones_ that happen during the
interaction with the Prolog process (and which can stem either from
process output *or* from user input) *atomic* in the sense that they are
all undone as a single unit. Please try out ediprolog to see what I
mean. In my view, the transaction concept that Stefan has proposed is a
very good solution for this and would benefit also other parts of Emacs.
It is this general issue that I would like to address and solve with a
suitable (likely: new) Emacs mechanism, not only the first %@.

>> In the typical interaction mode, point cannot move between 1) and 4):
>> Each key press is sent to the Prolog process, so that you can interact
>> with Prolog like in a terminal.
>
> Emacs blocks?

It is an interaction loop that can be left with C-g at any time. Please
try out ediprolog to see what I mean, for example on the query:

?- length(Ls, L).

> My own feeling here is, still, that while ediprolog's current system is
> straight-forward, it's probably not correct!

Please let me clearify: All that ediprolog does is inserting text into
the buffer. The text can either stem from process output, user input, or
(only in the very special case of the *first* %@ that is inserted when a
new interaction starts) even by ediprolog directly, as hardcoded string.

This seems completely normal and correct to me. Undo boundaries are
currently inserted without my control over them. I understand that this
is for efficiency reasons, to allow Emacs to trim the undo list.  I
would like to have more control over this, for example, by removing
these undo boundaries when the whole interaction is over. At that point,
we know that Emacs has not raised an error, so it should be very
safe. Several other solutions would also fit this particular case, such
as merging subsequent insertions in the undo list to keep its length
constant. These solutions should not depend on the timing of insertions.

We have found one case where I could prevent the undo boundary with
special-case code (for the initial %@), but a more general solution
would be much more welcome and work for all parts of the interaction.

All the best,
Markus