bug#25482: 26.0.50; Allow setting `query-replace-from-to-separator` to nil

[Juri Linkov <juri@linkov.net>]

>> >> So that design decision and the resulting code we have now had some
>> >> justification, and any replacement will have to support the same
>> >> features, which I think a simple string will not do.
>> >
>> > Ok, thanks for explanation, perhaps you can close this bug.
>>
>> Please don't close this bug.  I think your idea for customization was good.
>> We just need to find the right implementation that will keep the current
>> default intact, at the same time allowing you to customize the string
>> or set it to nil.
>
> I intended to move the char-displayable-p call from the defcustom, and
> remove the re-evaluate call.  But if you want to handle this part, it
> would be even better.  It's just that you've been very silent on this
> issue until now.

I've been silent because I agreed to your replies to Thierry :)
At one point in the beginning of this thread I believe Thierry sent
a patch closest to what would be good to have (before “code simplification”
where development went astray).  We could return to that patch and test
if it does these things right: instead of re-evaluating checks
char-displayable-p, but when query-replace-from-to-separator is nil
falls back to pre-24.5 that adds strings to the history without ^@.