Re: The surprising value of mouse-1 modifiers

[Lennart Borgman (gmail)]

Glenn Morris wrote:
> Lennart Borgman wrote (on Fri, 18 Jan 2008 at 01:44 +0100):
>
> > Why does
> >
> >    M-: (event-modifiers (list 'mouse-1))
> >
> > gives such a strange value while
> >
> >    M-: (event-modifiers (list 'M-mouse-1))
> >
> > is rather reasonable?
>
> Give a complete description, rather than leaving us to guess what you
> class as "surprising" and "strange".
>
> (click) and (meta click) seem equally reasonable to me.

This is a bit more problematic than my other problem reports yesterday 
(which where just plain mistakes by me I hope). I am not sure what is 
going on but I got for example

 (event-modifiers '(mouse-1)) => (click mouse-1 mouse-1 mouse-1 mouse-1 
mouse-1)

or just (click mouse-1) sometimes. I tried over and over again and got 
the first version so I assumed that was what everyone else saw too ;-)

I took a look at the C sources and to my surprise I saw that 
event-modifiers did a bit more than looked at the symbol parameter.

For the moment I am doing this before calling event-modifiers:

 (put k 'event-symbol-element-mask nil)

where k could be for example 'mouse-1.

It seems to work (as a workaround), but something is wrong I believe. 
Unfortunately I can not give a simple reciepe to reproduce this, but the 
ingredients might be:

  - using read-event to read mouse events
  - do not call read-event until there is a timeout (using 
double-click-time)
  - use edebug

What makes it even more difficult is that I have done some patches to 
w32 C sources where I am not sure that the latest merge did exactly what 
I wanted. Most of those patches are straight forward but some of them 
are in the difficult code for handling w32 messages.