bug#63988: 30.0.50; Recent header line format changes cause spin/seg fault with format-mode-line

[Aaron Jensen]

On Sat, Jun 10, 2023 at 7:22 AM Eli Zaretskii <eliz@gnu.org> wrote:
>
> > From: Aaron Jensen <aaronjensen@gmail.com>
> > Date: Sat, 10 Jun 2023 07:08:20 -0400
> > Cc: 63988@debbugs.gnu.org
> >
> > > But what is the purpose of such a strange (to use a civilized word)
> > > setting of header-line-format?  Why do you need :eval at all in this
> > > case?
> > >
> > > IOW, why not say "don't do that" and be done?
> >
> > It's a minimal repro for an issue I encountered with a package that does 
> > this:
> >
> > https://github.com/rougier/nano-modeline/blob/master/nano-modeline.el#L532
> >
> > It's how this modeline/header line adds line number/cursor position to
> > a more complicated line. As I understand it, it has to format it first
> > in order to use its width to properly right align it:
> >
> > https://github.com/rougier/nano-modeline/blob/master/nano-modeline.el#L278
> >
> > Is there a better way to do this?
>
> Use just 'format'?  I still don't understand why they use
> format-mode-line there.

Unless I'm mistaken, the closest equivalent to (format-modeline
"%l:%c) is something like:

(format "%d:%d" (line-number-at-pos) (current-column))

It seems like that behaves the same in narrowed buffers and not, so I
can probably use that instead.

> > > So maybe we should declare this feature a failed experiment and remove
> > > it?
> >
> > I'll admit I don't really understand the change. Is it actually
> > evaluating the cdr of the eval form up to two additional times to in
> > order to determine whether or not to display the headerline at all?
>
> Only two?

Well, 2 per render. I'm just guessing at reading the diff.

> > Wouldn't this have performance implications?
>
> Probably, although I wouldn't expect the performance to suffer too
> much.  Header-line is rarely used.

By some it's rarely used. By me and others it's used often as I use it
in place of modeline. I prefer the buffer name to be at the top.

> But yes, that's one more implication to consider when introducing such
> minor convenience features.

I'd personally prefer performance to being able to hide the header
line optionally, though I still don't really understand the change and
why the one time it's eval'd in order to get the actual text/format
can't be used to determine whether or not to show the header line.

Aaron