frame visibility wrong after redisplay

frame visibility wrong after redisplay

[David Reitter]

[-- Attachment #1: Type: text/plain, Size: 783 bytes --]

I'm still trying to fix #3303, and the problem I am seeing now is that  
FRAME_VISIBLE_P returns the wrong value.

In the NS port, but not on GNU/Linux (or perhaps X), I get the  
following behavior when evaluating this expression:

(let ((f (selected-frame)))
   (make-frame)
   (make-frame-invisible f t)
   (setq	aa2v (frame-visible-p f))
   (redisplay)
   (setq	aa3v (frame-visible-p f)))

(list aa2v aa3v) is, surprisingly, '(nil t).

Why is the frame deemed visible after the redisplay when it is, in  
fact, invisible?

What should I look for here?
I think drawrect: is called for the view of that frame, which sets  
async_visible (unclear why that's done there).  Why?

This is what's causing much pain somewhere else where we need to  
decide how to correctly raise it.




[-- Attachment #2: smime.p7s --]
[-- Type: application/pkcs7-signature, Size: 2193 bytes --]

Re: frame visibility wrong after redisplay

[Stefan Monnier]

> I'm still trying to fix #3303, and the problem I am seeing now is that
> FRAME_VISIBLE_P returns the wrong value.

> In the NS port, but not on GNU/Linux (or perhaps X), I get the following
> behavior when evaluating this expression:

> (let ((f (selected-frame)))
>   (make-frame)
>   (make-frame-invisible f t)
>   (setq	aa2v (frame-visible-p f))
>   (redisplay)
>   (setq	aa3v (frame-visible-p f)))

> (list aa2v aa3v) is, surprisingly, '(nil t).

> Why is the frame deemed visible after the redisplay when it is, in
> fact, invisible?

> What should I look for here?
> I think drawrect: is called for the view of that frame, which sets
> async_visible (unclear why that's done there).  Why?

> This is what's causing much pain somewhere else where we need to decide how
> to correctly raise it.

If you run under GDB, you should be able to place a watchpoint on the
`visible' (and `async_visible') field of frame `f', so you'll be
immediately told when those fields change.  The backtraces at those
times should give us a good indication of when/where the visibility is
incorrectly set back to t.


        Stefan