Re: Native line numbers landed on master

[Ergus]

On Wed, Oct 02, 2019 at 01:51:17AM +0200, Juanma Barranquero wrote:
> On Wed, Oct 2, 2019 at 12:53 AM Ergus <address@hidden> wrote:
>
> > 1) I don't have any strong feeling about using one or two values. But in
> > any case they should be configurable. This will add more use cases that
> > could benefit with this and I think it will be simple to implement (just
> > need to add 2 int variables).
>
> Adding the variables is trivial, yes. The problem is dealing with errors,
> like the user setting the variables to non-numbers, etc. I'm sure we don't
> want to throw signals there, do we?
>
> Eli, opinions?
We deal with similar issues very often. For example in
display_fill_column_indicator_column it uses a general case where it can
be t or a positive number. You in the simpler case you can use also
DEFVAR_INT to bypass part of the needed tests. But in general you only
need to avoid divisions by zero, value = NILP,or any non numbers. if so,
the code does nothing.

You only need this:

if (RANGED_FIXNUMP (1, my_Lisp_Object, INT_MAX))
  int my_C_object = XFIXNUM (my_Lisp_Object);



> > 2) Is it possible to call merge_faces more lazily? I mean inside the
> > `if` `else if` so only one of them will be called when needed. AFAIK
> > maybe_produce_line_number is called for every line and the code needs
> > only one of those merges every 5/10/x lines.
>
> You're right. Thanks. Improved patch attached.