It appears that the QVM tests are broken somehow but it isn't clear to me what might be wrong. It shows "pass" next to __boost_check_library__, but there are no results for the actual tests. It used to work fine earlier -- this issue might be related somehow to a previous problem, where I had a build directory without realizing that the test framework recognized that name and did things I did not intend; so I renamed the build directory to bld. Can someone help me fix this? Thanks, Emil
On Tue, Jul 5, 2016 at 1:05 PM, Thomas Trummer
On 05 Jul 2016, at 21:21, Emil Dotchevski
wrote: Can someone help me fix this?
The tests seem to run fine. I’ve attached the results, maybe it’s useful to someone.
Yes, the tests do work when I run them. This is what used to work fine, but isn't working right now: http://www.boost.org/development/tests/master/developer/qvm.html. Emil
On 05 Jul 2016, at 22:52, Emil Dotchevski
wrote: Yes, the tests do work when I run them. This is what used to work fine, but isn't working right now: http://www.boost.org/development/tests/master/developer/qvm.html.
I do understand this. What I meant to say was that the problem is not that the regression testers don’t produce a result but that the code that creates the test matrix doesn’t like it for some reason. Thomas
On Tue, Jul 5, 2016 at 9:52 PM, Thomas Trummer
On 05 Jul 2016, at 22:52, Emil Dotchevski
wrote: Yes, the tests do work when I run them. This is what used to work fine, but isn't working right now: http://www.boost.org/development/tests/master/developer/qvm.html.
I do understand this. What I meant to say was that the problem is not that the regression testers don’t produce a result but that the code that creates the test matrix doesn’t like it for some reason.
Ah, yes. That's what I'm trying to understand too. Emil
On 05 Jul 2016, at 22:05, Thomas Trummer
wrote: The tests seem to run fine. I’ve attached the results, maybe it’s useful to someone.
I did a second run with an older revision of QVM to see if there is a difference within the results. Obvious differences are that the test-log tags are missing values for some attributes and the output directories differ as well. I’ve attached the results for the run of the older revision, maybe someone with a better understanding of the testing system can chime in. Thomas
On Tue, Jul 5, 2016 at 2:21 PM, Emil Dotchevski
It appears that the QVM tests are broken somehow but it isn't clear to me what might be wrong. It shows "pass" next to __boost_check_library__, but there are no results for the actual tests. It used to work fine earlier -- this issue might be related somehow to a previous problem, where I had a build directory without realizing that the test framework recognized that name and did things I did not intend; so I renamed the build directory to bld.
Can someone help me fix this?
The problem stems from this change < https://github.com/boostorg/qvm/commit/b22b5868205653efc96b4c42c31c0294b0bc43b8>. The test & report system uses the test paths to determine various things including the library the test belongs to and the *sub-library* it belongs to. When you moved the test targets (what's in the Jamfile) the path changed. And from that point onwards the test paths, instead of looking like the library tests, started looking like sub-library tests. Specifically the test info references used to look like: boost-test(COMPILE) "qvm/hdrqvm-all~hpp" : "libs/qvm/test/header-test.cpp" And now they look like: boost-test(COMPILE) "qvm/bld/hdrqvm-all~hpp" : "libs/qvm/test/header-test.cpp" I did looked at the test and reporting code and did some experiments... And I don't see any good way to adjust to how you structured your tests without adversely impacting the support for sub-libraries. Best I can suggest for now is to move the test targets back into the qvm/test/Jamfile.v2 file. -- -- Rene Rivera -- Grafik - Don't Assume Anything -- Robot Dreams - http://robot-dreams.net -- rrivera/acm.org (msn) - grafikrobot/aim,yahoo,skype,efnet,gmail
The problem stems from this change <
. The test & report system uses the test paths to determine various things including the library the test belongs to and the *sub-library* it belongs to. When you moved the test targets (what's in the Jamfile) the path changed. And from that point onwards the test paths, instead of looking
https://github.com/boostorg/qvm/commit/b22b5868205653efc96b4c42c31c0294b0bc4... like the library tests, started looking like sub-library tests. Specifically the test info references used to look like:
boost-test(COMPILE) "qvm/hdrqvm-all~hpp" : "libs/qvm/test/header-test.cpp"
And now they look like:
boost-test(COMPILE) "qvm/bld/hdrqvm-all~hpp" : "libs/qvm/test/header-test.cpp"
I did looked at the test and reporting code and did some experiments... And I don't see any good way to adjust to how you structured your tests without adversely impacting the support for sub-libraries. Best I can suggest for now is to move the test targets back into the qvm/test/Jamfile.v2 file.
Yes I wanted to move the targets out from qvm/test, I guess that won't work. I'll move them back into qvm/test. Thanks! Emil
participants (3)
-
Emil Dotchevski
-
Rene Rivera
-
Thomas Trummer