Re: [Nano-devel] [PATCH] Implement incremental search v5

[Benno Schulenberg]

Op 01-02-18 om 21:27 schreef Marco Diego Aurélio Mesquita:
> The attached patch implements incremental search in nano. It addresses 
> complaints from the previous version.

Hrm!  I said you should get rid of the do loop in do_search.  But what
do you do?  You add another one!  --  No!  Make your patch smaller, not
bigger.  First make it work flawlessly for a single thing, and only then
maybe start thinking about extending it.

> I had to change the behavior of do_replace so that is is now consistent with
> the behavior of do_search.

No.  I don't want to have incremental search when replacing something.
It doesn't make sense to me.  And keep incremental search out of the
findinhelp too for now -- look how ugly it makes the help lines look.

> Also, I didn't use the changes suggested by Benno on the previous revision
> because I was still unhappy with the way do_right was used While fixing it,
> some problems were found and then fixed.

Yes, using do_right() is not nice, but adding an unneeded loop is
far worse.

> Please, give it a spin and review or commit it.

Commit it!?  You are going way too fast!

A short spin:  'src/nano +1 NEWS', then type:
^W M-I ju.*
It beeps two times, correctly, because I forgot the regexp toggle.
So type:
M-R

Hm. It doesn't change the highlight to fit the changed situation.
Continue typing:
t\>

After the "t" it does update the highlight, after the "\" it beeps,
and after the ">" nothing changes, correctly.  But now, type:
<Backspace>

The ">" is removed, but it takes a full second before the cursor
reappears and the beep occurs.  Weird.  Try it again: > <Backspace>

Same thing.  Do it again: > <Backspace>

It again takes a second for the cursor to reappear, but now it
*doesn't* beep.  Stranger and stranger...  Do it again: > <Backspace>

Now the cursor reappears instantly and there is no beep.
What's going on there?  Somewhere the error message from
regcomp ("Trailing backslash") gets eaten or suppressed
or overwritten.  Apparently a failing regexp_init() needs
some different handling.

Benno