Re: [PATCH] advapi32: test and fix ChangeServiceConfig2 when given a zero length or a null description (try2, resend)
Hi,
While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=15399
Your paranoid android.
=== w7u (32 bit service) === service.c:1421: Test failed: Active services mismatch 1 service.c:1422: Test failed: Inactive services mismatch 4294967295
Am 17.07.2015 um 15:35 schrieb Marvin:
Hi,
While running your changed tests on Windows, I think I found new failures. Being a bot and all I'm not very good at pattern recognition, so I might be wrong, but could you please double-check? Full results can be found at https://testbot.winehq.org/JobDetails.pl?Key=15399
Your paranoid android.
=== w7u (32 bit service) === service.c:1421: Test failed: Active services mismatch 1 service.c:1422: Test failed: Inactive services mismatch 4294967295
Hello,
On two attempts to send a patch modifying the service tests I got a failing test in the "w7u" VM while the other VMs were succeeding [1][2][3].
This test first gets the number of active and inactive services. Then enumerates all services and decrements the corresponding counter. At the end both counter should be zero.
I am now under the impression that this is just a race condition of this VM when the patch arrives via patchwatcher. The same patch gave no error when sent via the testbot web interface [4][5].
Is there something against this theory about one service stopping between the different calls to EnumServicesStatusA?
Would the attached patch be acceptable for inclusion?
It gets the counter before and after enumeration and just tests if the original values have not changed in between.
Kind regards, Bernhard
[1] https://testbot.winehq.org/JobDetails.pl?Key=15399 [2] http://source.winehq.org/patches/data/112996.testfail [3] http://source.winehq.org/patches/data/111835.testfail [4] https://testbot.winehq.org/JobDetails.pl?Key=15334 [5] https://testbot.winehq.org/JobDetails.pl?Key=15406
On 17 Jul 2015, at 20:15, Bernhard Übelacker wrote:
On two attempts to send a patch modifying the service tests I got a failing test in the "w7u" VM while the other VMs were succeeding [1][2][3].
This test first gets the number of active and inactive services. Then enumerates all services and decrements the corresponding counter. At the end both counter should be zero.
Yes, there's clearly a race in the tests. What's more, this failure happens before your new code and is clearly not related to it.
Of course it would be nice to fix the race, but that doesn't seem possible the way the existing tests are written.
Huw.
Am 20.07.2015 um 09:55 schrieb Huw Davies:
On 17 Jul 2015, at 20:15, Bernhard Übelacker wrote:
On two attempts to send a patch modifying the service tests I got a failing test in the "w7u" VM while the other VMs were succeeding [1][2][3].
This test first gets the number of active and inactive services. Then enumerates all services and decrements the corresponding counter. At the end both counter should be zero.
Yes, there's clearly a race in the tests. What's more, this failure happens before your new code and is clearly not related to it.
Of course it would be nice to fix the race, but that doesn't seem possible the way the existing tests are written.
Huw.
Hello Huw, thanks for your answer.
So the patch attached to my last mail does really not fix it, but it tries to skip these tests when it detects that number of active/inactive services has changed from before to after enumeration.
Kind regards, Bernhard
-
Bernhard Übelacker -
Huw Davies -
Marvin