[PATCH 4/7] user32: Implement GetRegisteredRawInputDevices

[Francois Gouget]

On Tue, 27 Aug 2019, Rémi Bernon wrote:
[...]
> I'm relieved to learn that mutter is not used for testing :)

I think it's an often overlooked feature; one can get information about 
the test VMs configuration in two places:

* On the VM selection page when submitting a job: click on the 
  short VM description (such as Debian 10 / vm4) in the Description 
  column.

* And similarly on the test result page in each VM section clicking on 
  the short description will similarly reveal more details.

For instance clicking on "Debian 10 / vm4" shows the following:

Debian 10 configured for Wine. Runs fvwm. [CPU:4*host RAM:4GB disk:virtio-scsi eth:virtio snd:ich6 GPU:qxl display:spice testagentd:1.7]
Missions: win32,test=module:win32,lang=fr_FR:win32,lang=ja_JP:win32,test=module,lang=zh_CN

The first line describes the software and hardware configuration of the 
VM. If more information about the software side would be useful I could 
add some more.

The second line describes the tests that the VM is configured to 
perform. It's a colon separated list.
* win32,test=module
  The VM will perform tests in a regular 32 bit Wine build (so not a 
  Windows on Windows setup) and that if a patch touches a module, such 
  as user32, then all of that module's tests should be rerun. This is 
  meant to detect test breakage caused by changes in the Wine 
  implementation.

* win32,lang=fr_FR and win32,lang=ja_JP
  Same build but in a French and Japanese locale respectively. And since 
  'test' is not specified it means a test is only run if it is patched 
  directly. 

* win32,test=module,lang=zh_CN
  Still the same build but in a Chinese locale and again running all of 
  a dll's tests if that dll is patched. This helps provide a minimum 
  level of locale testing when patching Wine dlls but it can also be 
  helpful to detect random failures for tests that don't depend on 
  locales.

Each of these tests is run in its own wineprefix so they are mostly 
independent. But they run out of the same account on the same X session 
so if a test borks the X server it would impact the ones that come 
after.

-- 
Francois Gouget <fgouget at codeweavers.com>