On Wed, Aug 24, 2011 at 03:22, Hans Leidekker hans@codeweavers.com wrote:
dlls/msi/tests/action.c | 6 ++++++ 1 files changed, 6 insertions(+), 0 deletions(-)
diff --git a/dlls/msi/tests/action.c b/dlls/msi/tests/action.c index aac601f..90832af 100644 --- a/dlls/msi/tests/action.c +++ b/dlls/msi/tests/action.c @@ -4721,6 +4721,12 @@ static void test_create_remove_folder(void) { UINT r;
- if (is_process_limited())
- {
- skip("process is limited\n");
- return;
- }
CreateDirectoryA("msitest", NULL); CreateDirectoryA("msitest\first", NULL); CreateDirectoryA("msitest\second", NULL); -- 1.7.4.1
Shouldn't that be a win_skip()? Wine always runs as administrator, and if the process is running as a limited user, something may be broken? Unless you're preparing for when wine supports non-admin mode.. :)
On Wed, 2011-08-24 at 03:26 -0500, Austin English wrote:
Shouldn't that be a win_skip()? Wine always runs as administrator, and if the process is running as a limited user, something may be broken? Unless you're preparing for when wine supports non-admin mode.. :)
No, win_skip is for behavior observed on Windows that Wine should not emulate, which is not the case here. And testing if a process is correctly limited should be done elsewhere.
On Wed, 24 Aug 2011, Hans Leidekker wrote:
On Wed, 2011-08-24 at 03:26 -0500, Austin English wrote:
Shouldn't that be a win_skip()? Wine always runs as administrator, and if the process is running as a limited user, something may be broken? Unless you're preparing for when wine supports non-admin mode.. :)
No, win_skip is for behavior observed on Windows that Wine should not emulate, which is not the case here. And testing if a process is correctly limited should be done elsewhere.
My understanding is that win_skip() is for tests that it's ok to skip on Windows but not ok to skip in Wine. I don't think there's more semantics attached to it than that.
From my understanding, in this case using a simple skip() means that
anyone running WineTest on Windows in a non-administrator account will get a test failure. Since we don't seem to have a policy mandating that WineTest be run in an administrator account that seems wrong and thus a win_skip() seems in order.
See also my email about best practices for running WineTest for a chance at changing/setting WineTest policy: http://www.winehq.org/pipermail/wine-devel/2011-August/091703.html
On Wed, 2011-08-24 at 15:11 +0200, Francois Gouget wrote:
From my understanding, in this case using a simple skip() means that anyone running WineTest on Windows in a non-administrator account will get a test failure. Since we don't seem to have a policy mandating that
Quite the opposite, running it as a regular user causes a test failure now, my patch avoids that.
On Wed, 24 Aug 2011, Francois Gouget wrote: [...]
From my understanding, in this case using a simple skip() means that anyone running WineTest on Windows in a non-administrator account will get a test failure. Since we don't seem to have a policy mandating that WineTest be run in an administrator account that seems wrong and thus a win_skip() seems in order.
Sorry I got that wrong and Hans set me straight: * skip() always skips no matter what the platform is. So it cannot cause test errors.
* win_skip() skips on Windows and fails in Wine. So it's stricter than skip().
On 08/24/2011 10:47 AM, Hans Leidekker wrote:
No, win_skip is for behavior observed on Windows that Wine should not emulate, which is not the case here.
Is there any page on the wiki explaining the meaning of broken(), skip(), win_skip() ? Because your definition of win_skip is my definition of broken that is : something seen on windows but broken (wrong behaviour fixed in a more recent windows version or caused by a buggy driver).
On Wed, 24 Aug 2011, GOUJON Alexandre wrote:
On 08/24/2011 10:47 AM, Hans Leidekker wrote:
No, win_skip is for behavior observed on Windows that Wine should not emulate, which is not the case here.
Is there any page on the wiki explaining the meaning of broken(), skip(), win_skip() ? Because your definition of win_skip is my definition of broken that is : something seen on windows but broken (wrong behaviour fixed in a more recent windows version or caused by a buggy driver).
As far as I know it's not documented anywhere. Also I'd personally argue that if it's documented wine/test.h would be a better place than the Wiki (but Alexandre may disagree).
In any case there's a fundamental difference which is that skip() and win_skip() issue a trace noting that some tests are skipped and causing such a trace to be counted in the statistics, while broken() just returns a boolean and has no side-effect.
-
Austin English -
Francois Gouget -
GOUJON Alexandre -
Hans Leidekker