[PATCH v2] user32/tests: Tests for DM_REPOSITION message
Francois Gouget
fgouget at codeweavers.com
Thu Nov 7 21:09:30 CST 2019
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 at codeweavers.com>
More information about the wine-devel
mailing list