Re: [PATCH v2] user32/tests: Tests for DM_REPOSITION message

[katahiromz .]

> user32:
> dialog.c:1987: Test failed: dialog doesn't use message box font
> Why it failed? This is not related to message box.

[PATCH v2] user32/tests: Tests for DM_REPOSITION message

[Francois Gouget]

On Thu, 7 Nov 2019, katahiromz . wrote:
> > user32:
> > dialog.c:1987: Test failed: dialog doesn't use message box font
> > Why it failed? This is not related to message box.

This is another test performed by user32:dialog.

Your patch modifies the user32:dialog test unit so the TestBot reran it. 
As it happens the other tests in user32:dialog fail and the TestBot then 
has to determine if these failures are related to your patch or not; 
which it does by comparing the task's result to the latest WineTest 
results.
> === w1064v1809_ar (32 bit report) ===
> &
> === w1064v1809_zh_CN (32 bit report) ===
>
> user32:
> dialog.c:1987: Test failed: dialog doesn't use message box font

This test fails systematically on Windows 10 1709+. However the TestBot 
did not find reference test results for user32:dialog for the 
w1064v1809_ar and w1064v1809_zh_CN VM configurations. I looked into this 
and there are two reasons for this:

* The WineTest reports on record date back to November 1st and 
  October 22 respectively. However during these runs get_subtests() 
  failed to get the list of user32 tests with error 1359 
  (ERROR_INTERNAL_ERROR, see programs/winetest/main.c).

  There are a couple of possible sources for this error code but the 
  most likely one is that WineTest.exe failed to parse the output of 
  user32_test.exe which would be pretty abnormal and intriguing.

  A deeper analysis of the test.winehq.org reports indicates that type 
  of error still happens on a regular basis to random test binaries. So 
  this is something that must be investigated (see bug 48061).

* And since then these WineTest runs have been timing out which means 
  the reports have not been updated since they would be incomplete.

So as a result of these two factors the TestBot did not know that this 
error systematically occurs on w1064v1809_ar and w1064v1809_zh_CN and 
thought it was caused by your patch.
> === debian10 (32 bit report) ===
> === debian10 (32 bit French report) ===
> === debian10 (32 bit Japanese:Japan report) ===
> === debian10 (32 bit Chinese:China report) ===
> === debian10 (32 bit WoW report) ===
> === debian10 (64 bit WoW report) ===
>
> user32:
> dialog.c:2197: Test failed: rc.left was -100
> dialog.c:2198: Test failed: rc.top was -100

These however don't seem to happen on unpatched Wine. The following page 
shows Wine getting 8 todos but 0 failures for user32:dialog (except 72 
failures once a long time ago).

http://winetest.dolphin/data/tests/user32:dialog.html

So I think these two failures could indeed be caused by your patch.


-- 
Francois Gouget <fgouget(a)codeweavers.com>