bug#25875: 26.0.50; Hang logging out of MS-Windows

[Richard Copley <rcopley@gmail.com>]

On 27 February 2017 at 15:36, Eli Zaretskii <eliz@gnu.org> wrote:
>> From: Richard Copley <rcopley@gmail.com>
>> Date: Mon, 27 Feb 2017 08:14:22 +0000
>> Cc: Eli Zaretskii <eliz@gnu.org>, 25875@debbugs.gnu.org
>>
>> > Here's a quick and dirty attempt.  If I haven't made a mistake, it replaces
>> > every relevant call to SendMessage by a call to SendMessageTimeout with a
>> > 100ms timeout.
>> >
>> > --- a/src/w32term.c
>> > +++ b/src/w32term.c
>> > @@ -537,6 +537,15 @@ x_update_begin (struct frame *f)
>> >  }
>> >
>> >
>> > +#undef SendMessage
>> > +#define SendMessage DebugSendMessage
>> > +
>> > +static LRESULT WINAPI
>> > +DebugSendMessage (HWND hWnd, UINT Msg, WPARAM wParam, LPARAM lParam)
>> > +{
>> > +  return SendMessageTimeoutA (hWnd, Msg, wParam, lParam, 0, 100, NULL);
>> > +}
>> > +
>> >  /* Start update of window W.  */
>> >
>> >  static void
>> >
>> > Ken
>>
>> Sorry Ken, I can't sabotage myself like that, I have work to do.
>
> This could be a misunderstanding: the above change is not supposed to
> sabotage anything, it's supposed to be a 100% compatible change for
> the current behavior when all threads are running, and also provide a
> "fire escape" when the addressee of the message is for some reason
> stuck, as we think happens in your scenario.

From the docs for SendMessageTimeout:
  "If the function succeeds, the return value is nonzero.".
We're going to cast that to HWND and pretend it's a scrollbar?
(See `my_create_vscrollbar()' in "w32term.c".)
Then what happens? Ken, what happened when you tested this?

> If you are unwilling to make such a sweeping change, could you at
> least replace the call SendMessage in my_show_window with
> SendMessageTimeoutA, using the above patch as a template?

I will think about it, but I'll ignore the patch :)