bug#34525: replace-regexp missing some matches

[Alan Mackenzie]

Hello, Eli.

On Wed, Feb 27, 2019 at 18:39:31 +0200, Eli Zaretskii wrote:
> > Date: Wed, 27 Feb 2019 14:22:51 +0000
> > Cc: Eli Zaretskii <eliz@gnu.org>, daniel.lopez999@gmail.com,
> >   34525@debbugs.gnu.org
> > From: Alan Mackenzie <acm@muc.de>

> >   if (! NULL_RIGHT_CHILD (i))
> >     {
> >       i = i->right;
> >       while (! NULL_LEFT_CHILD (i))
> >         i = i->left;                  <===============

> >       i->position = next_position;
> >       return i;
> >     }

> > Here, in seeking the next interval, we go down a chain of `left's.  We
> > do not set the ->position field of these intervals, except for the last
> > one, which we return.

> The position field is just a cache, isn't it?

It's a cache, yes.  But it's used in update_interval, for example.  It
would appear bad things were happening because it wasn't being kept up to
date.

> > So the returned interval doesn't satisfy the condition that all its
> > parents have their ->position's set correctly.  Thus if we use this
> > interval as an argument to update_interval, we will likely fail.  I
> > think this can happen in update_syntax_table.

> next_interval and previous_interval are used extensively, so I'm
> having hard time believing that they have such a blatant bug.

Yes, how come we haven't seen many more consequences?

Maybe syntax-table text properties aren't as widely used as all that.

Also, the effect would have to be seen within the time between successive
calls to SETUP_SYNTAX_TABLE and friends, since each such call
reinitialises this cache, in an important sense.

And we only saw the effect when the replacement text "SharedBitmap" was
exactly twice the length of the original word "Bitmap".

Anyhow, fixing this item appears to fix the bug (see the tentative patch
in my post from 15:08:49 +0000 to Stefan).

I'm guessing this fix will be deemed too unsafe to make it into the
emacs-26 release branch.  ;-(

-- 
Alan Mackenzie (Nuremberg, Germany).