Joerg-Cyril.Hoehle@t-systems.com writes:
-START_TEST(mci) +#ifndef FULL_TEST +/* Some tests hang or crash Wine. Enable once fixed. */ +#define FULL_TEST 0 +#endif
Please don't introduce dead code. If the tests really can't work then don't add them at all until you have the corresponding fixes.
Hi,
+/* Some tests hang or crash Wine. Enable once fixed. */ +#define FULL_TEST 0
Please don't introduce dead code. If the tests really can't work then don't add them at all until you have the corresponding fixes.
Will do. Although I'd very much prefer having tests than no tests. I've been fixing MCI regressions that were as old as 2001 that would hopefully not have occured if there had been tests in the first place. The #if FULL_TEST is very useful when compiling on MS-W* to immediately spot what Wine should eventually do, and as direction for others when I'll be gone.
How do other people deal with that situation? - use some patch manager, e.g. quilt? - maintain seprate test files for running under MS-W*? - ignore, write no tests and wait for bugs.winehq? - ...?
Regards, Jörg Höhle PS: despite my 20 patches to MCI, there's still need for #if !FULL_TEST.
2009/10/26 Joerg-Cyril.Hoehle@t-systems.com:
Hi,
+/* Some tests hang or crash Wine. Enable once fixed. */ +#define FULL_TEST 0
Please don't introduce dead code. If the tests really can't work then don't add them at all until you have the corresponding fixes.
Will do. Although I'd very much prefer having tests than no tests. I've been fixing MCI regressions that were as old as 2001 that would hopefully not have occured if there had been tests in the first place. The #if FULL_TEST is very useful when compiling on MS-W* to immediately spot what Wine should eventually do, and as direction for others when I'll be gone.
How do other people deal with that situation? - use some patch manager, e.g. quilt? - maintain seprate test files for running under MS-W*? - ignore, write no tests and wait for bugs.winehq? - ...?
If the tests are returning incorrect results in Wine, the practice when writing the tests is to add todo_wine before the ok statement. Once the fix has been added, the todo_wine bit can be removed. E.g.
before fix:
todo_wine ok(hr == S_OK, "should return S_OK, got: %08x", hr);
after fix:
ok(hr == S_OK, "should return S_OK, got: %08x", hr);
Alternatively, you can fix a Wine bug and add the test for it in the same patch (or with the tests added in the next patch in the series). These should ideally be tested on Wine and at least one Windows platform. Other platforms will be picked up on the tests.winehq.org site.
- Reece
Reece dunn wrote:
Once the fix has been added, the todo_wine bit can be removed.
I know about it and skip and win_skip thanks to Michael Karcher's excellent text: http://www.winehq.org/pipermail/wine-devel/2008-October/069721.html
The problem is that running the full tests crashes or hangs Wine, but not MS-Windows platforms. A hang or crash yields less useable results on test.winehq and is not nice to users.
Other platforms will be picked up on the tests.winehq.org site.
What I'm missing is sort of the converse of skip: Perform a test on all platforms *except* Wine, such that I can later observe on test.winehq the results of a variety of original MS-Windows systems - many more than the few I can access.
How to achieve this?
Thanks, Jörg Höhle
On 10/27/2009 10:13 AM, Joerg-Cyril.Hoehle@t-systems.com wrote:
Reece dunn wrote:
Once the fix has been added, the todo_wine bit can be removed.
I know about it and skip and win_skip thanks to Michael Karcher's excellent text: http://www.winehq.org/pipermail/wine-devel/2008-October/069721.html
The problem is that running the full tests crashes or hangs Wine, but not MS-Windows platforms. A hang or crash yields less useable results on test.winehq and is not nice to users.
Other platforms will be picked up on the tests.winehq.org site.
What I'm missing is sort of the converse of skip: Perform a test on all platforms *except* Wine, such that I can later observe on test.winehq the results of a variety of original MS-Windows systems
- many more than the few I can access.
How to achieve this?
Thanks, Jörg Höhle
Isn't there anything in your current test suite that succeeds on Windows but fails on Wine? This could be as simple as returning a different last error, or not exporting a function (a one not used yet, of course).
Paul Virnes wrote:
What I'm missing is sort of the converse of skip: Perform a test on all platforms *except* Wine
Isn't there anything in your current test suite that succeeds on Windows but fails on Wine?
This is it! So simple, right in front of my eyes. Thanks, Jörg Höhle.