Re: master 128ed5c9f17: Add one more mouse-set-point call to functions xref-find-*-at-mouse

[Dmitry Gutov]

On 31/08/2023 04:10, Po Lu wrote:
> > On 31/08/2023 03:43, Po Lu wrote:
> > > My experience with the entire touch screen affair says that many such
> > > errors would never have been written had `xref-find-definitions' taken
> > > an EVENT argument and used its posn-point if present.  Something to
> > > consider, I suppose.
> > Sorry, I don't quite understand which change you have in mind.
> >
> > The issue here was that we needed to move to the click's location both
> > when detecting the identifier (at that point), and before using the
> > identifier to find a set of locations.
> >
> > If we used a POSN value, it would also have to be used in two places,
> > I guess. One of which we could just as easily missed as we had here.
> In my opinion, commands or functions that respond to the location of a
> click should always take an EVENT argument, and use:
>
>    (posn-point (event-start event))
>
> if EVENT is present and incorporates a mouse position list.  Rather than
> searching for an identifier at point, it should search for an identifier
> at the point of the mouse click.  Adhering to such a convention is
> almost guaranteed to eliminate every bug where Emacs consults something
> under point instead of the mouse pointer when the user meant the latter.

Sorry, I'm still not seeing your point. Especially the last paragraph in 
my last email seems to remain unaddressed.

And both commands take EVENT as an argument already.

The problem is not where is searched for an identifier, but *from where* 
it searched for the identifier's locations. And that position is 
important only for particular Xref backends (both LSP-based ones, I'm 
assuming).