Details
-
Bug
-
Status: Closed (View Workflow)
-
Blocker
-
Resolution: Won't Fix
-
10.2.5
-
None
-
Fedora 25 - GCC 6
Description
Hello,
after release 10.2.5, I was finally able to atleast compile it and pack it, if ommiting all tests.
However when running the testsuite, it fails on the very first and most basic tests.
74% tests passed, 21 tests failed out of 81
|
Total Test time (real) = 48.87 sec |
The following tests FAILED:
|
23 - bulk1 (Failed)
|
24 - performance (Failed)
|
25 - basic-t (Failed)
|
26 - fetch (Failed)
|
27 - charset (Failed)
|
28 - logs (Failed)
|
29 - cursor (Failed)
|
30 - errors (Failed)
|
31 - view (Failed)
|
32 - ps (Failed) |
33 - ps_bugs (Failed)
|
34 - sp (Failed)
|
35 - result (Failed)
|
36 - connection (Failed)
|
37 - misc (Failed)
|
38 - ps_new (Failed)
|
39 - sqlite3 (Failed)
|
40 - thread (Failed)
|
41 - features-10_2 (Failed)
|
42 - dyncol (Failed)
|
43 - async (Failed)
|
Errors while running CTest |
Full build log can be found here.
Attachments
Issue Links
- duplicates
-
-
- Closed
-
- is duplicated by
-
-
- Closed
-
Activity
Ok
- By "that's exactly what we had earlier", you mean 10.1.21, for example?
In Fedora, we have both 'make test' and the 'mysql-test.run'.
I'll disable 'make test' tests, leaving 'mysql-test.run' testsuite do its work and I'll let you know about my progress.
No, not 10.1. Earlier releases of 10.2. We've replaced libmysqlclient with libmariadb in 10.2, libmariadb came with unit tests, mysql-test-run picked them up, they all passed, all was fine. And then we've realized that they "pass" because they return 0 (which means success) when they cannot connect to the server. After this was fixed, tests were not all passing anymore 
Connector/C unit tests require a running server, they cannot be run standalone.
We use mysql-test-run for that, it sets up the environment, starts the server, and then runs C/C unit tests.
See, for example this build log, C/C tests start from "unit.conc_"
I understand that your problem is that make test fails, right? We could make C/C tests to be skipped, if the server isn't running, but that creates a danger of everything looking good in mysql-test, even when it's not. In fact, that's exactly what we had earlier, C/C tests "passed" perfectly, because they silently returned 0 if the server wasn't found.