We need to create a new class of test cases (a mix of C++ gtest and Java junit) that run for long periods of time. I don't know what the right way to do this in terms of both build/test infrastructure (jenkins) and in the source code.
We should always be building these test cases as part of the regular build process but we don't want these to run when someone invokes make check. There might be a way to label tests as "long running" in C++ just by the class name (e.g., LongTest versus Test).
The long running tests should be running all the time in Jenkins. It can just run on the master branch. When one test finishes, it should automatically fire up the next one. If a test run fails, it should send an email to the mailing list.
The first test case should be a test that hammers the Bw-Tree index for hours (e.g., insert 1b keys, delete 1b keys, repeat). We should also run @tli2's fuzz tester in this manner as well.
crd477
commented
6 years ago
We need to create a new class of test cases (a mix of C++ gtest and Java junit) that run for long periods of time. I don't know what the right way to do this in terms of both build/test infrastructure (jenkins) and in the source code.
We should always be building these test cases as part of the regular build process but we don't want these to run when someone invokes
make check. There might be a way to label tests as "long running" in C++ just by the class name (e.g.,LongTestversusTest).The long running tests should be running all the time in Jenkins. It can just run on the master branch. When one test finishes, it should automatically fire up the next one. If a test run fails, it should send an email to the mailing list.
The first test case should be a test that hammers the Bw-Tree index for hours (e.g., insert 1b keys, delete 1b keys, repeat). We should also run @tli2's fuzz tester in this manner as well.