Wow, that's fantastic progress! (And I'm happy to see that my Windows scripts
were close to working.)
ctest has a --timeout flag, so in build_and_test.bash, let's change this line:
ctest -C ${CMAKE_BUILD_TYPE} -E broken --force-new-ctest-process
--output-on-failure --timeout 120
I don't think any test should take longer than 2 mins, but if it turns out that
any do, we can adjust this number (globally or for an individual test), it
doesn't need to be a tight bound, it just needs to kill a test if it's
deadlocked.
Then we can start the process of debugging why some things are hanging. Is it
consistently certain tests that seem to deadlock?
I don't expect all tests to pass when you run them manually ("make test" or
"ctest") -- there are a couple of problematic ones that we should track down.
But those tests are intentionally excluded for the CI tests (they are excluded
when it knows it's running on GHA, the logic is in testing.cmake), so I do
expect every test to pass for CI.
The general nature of each test should be apparent by name, but every one is
really controlled by testsuite/<testname>/run.py, and from that short script
it's usually pretty obvious what is being run and why.
I think that the way to proceed is to get the timeout working (even if some
tests fail or hang until the timeout causes them to be killed), and then
prepare a PR with what you have so far. Once that gets merged and we can all
inspect the failed tests, I think it may be more apparent what things are
broken and how to prioritize those tasks.
-- lg
> On Dec 23, 2020, at 10:32 AM, Dieter De Baets <[email protected]> wrote:
>
> Hey all,
>
> I finally found some time to try getting the CI tests going on Windows.
>
> The good news is that after some typical windows vs unix fixes (paths, DLL
> loading paths,.. see
> https://github.com/debaetsd/oiio/commit/1a3b9c8a8a412bda6597dc956b0414c996c93de9
>
> <https://github.com/debaetsd/oiio/commit/1a3b9c8a8a412bda6597dc956b0414c996c93de9>
> in case you want to know more), about 70% of the test pass
> The bad news is obviously that still 41 tests currently fail (including a 75%
> change with a MT deadlock that completely trips up the GH actions).
>
> Is it safe to assume that all tests are expected to pass?
> I can start looking into fixing some of them later.
> Is there a list or some documentation about what each test is expected to do?
>
> Best Regards,
> Dieter
>
>
> _______________________________________________
> Oiio-dev mailing list
> [email protected]
> http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
--
Larry Gritz
[email protected]
_______________________________________________
Oiio-dev mailing list
[email protected]
http://lists.openimageio.org/listinfo.cgi/oiio-dev-openimageio.org
