user32:win is a mess (was: Allow the backends to defer restoring minimized windows.)

[Francois Gouget]

On Fri, 2 Jul 2021, Arkadiusz Hiler wrote:
[...]
> > === debiant2 (32 bit WoW report) ===
> > 
> > user32:
> > win.c:10376: Test failed: GetForegroundWindow() = 00000000
> > win.c:10381: Test failed: Expected foreground window 00150128, got 00E400F2
> > win.c:10383: Test failed: GetActiveWindow() = 00000000
> > win.c:10383: Test failed: GetFocus() = 00000000
> > win.c:10384: Test failed: Received WM_ACTIVATEAPP(1), did not expect it.
> > win.c:10385: Test failed: Received WM_ACTIVATEAPP(0), did not expect it.
> > win.c:10393: Test failed: Expected foreground window 00150128, got 00000000
> > win.c:10395: Test failed: GetActiveWindow() = 00000000
> > win.c:10395: Test failed: GetFocus() = 00000000
> > win.c:10403: Test failed: Received WM_ACTIVATEAPP(1), did not expect it.
> 
> Those are just flaky tests. I was seeing anywhere between 3-17 failures
> in user32 win tests during the development + I don't see how the change
> in question could cause any of those.

Right. user32:win is a complete mess (and it's not even the worst of 
the user32 tests):

https://test.winehq.org/data/patterns.html#user32:win

But even if the task is daunting we'll have to start cleaning it up at 
some point rather than systematically turning away (and I'm not saying 
this specifically for you, just seizing the occasion).

Unfortunately all I can say about the failures above is that they are 
random. But I diagnosed some of the others and maybe these can be useful 
starting points:

* On Vista to Windows 8.1 part of mscoree:mscoree fails to run, breaks user32:win
  https://bugs.winehq.org/show_bug.cgi?id=51390

* On Windows 7 and 8.1 ntoskrnl.exe:ntoskrnl triggers a network firewall dialog, breaks user32:win
  https://bugs.winehq.org/show_bug.cgi?id=51391

* user32:monitor breaks user32:win
  https://bugs.winehq.org/show_bug.cgi?id=51392

* user32:win gets an unexpected 0x7fff message in the Korean locale
  https://bugs.winehq.org/show_bug.cgi?id=51395

* There's also this set of failures which only happen in 64-bit [1] 
  tests on the cw-gtx560 and cw-rx460 machines for some, as yet,
  unknown reason.
  win.c:2284: Test failed: expected !100
  win.c:2284: Test failed: expected !100


[1] Actually it happened in the 32-bit tests exactly once on 2021-05-04 
    on cw-gtx560-1909-32:
    https://test.winehq.org/data/cda4abac9859ed42c29e8cb2746201a578431a5e/win1909_cw-gtx560-1909-32/user32:win.html

-- 
Francois Gouget <fgouget at codeweavers.com>