Re: disabling undo boundaries

[Phillip Lord]

Stefan Monnier <address@hidden> writes:
>> My fault for not being clear.  AFAICT, undo-boundary does not set
>> last_undo_buffer at all.
>
> Oh, indeed, it sets last_boundary_buffer but not last_undo_buffer!
> Duh!
>
>> I'm still trying to understand the reason behind the logic in the first
>> place; unfortunately, the code seems to predate the earliest VC records.
>
> IIUC the reason is for modifications in another buffer that take place
> either within a command or between commands (e.g. process output), so
> they don't accumulate arbitrarily within a single undo-boundary.
>
> E.g. if you have a command "add timestamp to log buffer" which you
> typically use from another buffer then, without this logic, running this
> command 100 times would result in all entries in the buffer being lumped
> together into a single undo-unit, so you couldn't undo them separately.


In the other buffer? So, you can undo the changes to the log buffer one
after the other? I agree that makes sense, but it does seem to me that
this could be the responsibility of "add timestamp to log buffer".

I'm checking at the moment how often this undo-boundary gets called. So
far, it's about 1.5% of the time that an undo event gets added, so it's
not very often (obviously this is without an a-c-f entry or it would be
much higher!).

Phil