labath wrote: > @skipIfRemote is too much. These tests are still usable for Linux->Linux and > such.
I don't really care about this, but I'll note that while these tests will run in a remote configuration, they will not actually test any meaningful property of the remote setup. They create a local pty endpoint, connect and talk over it, completely ignoring the the remote connection that the general test infra has set up for them. In fact, I might go so far as to say that if this test does ever fail in a remote configuration, then that's a bug in the test, because it's not insulated enough from the environment. I realize this if confusing, and that's because the API test suite is really two test suites jumbled into one: - on one side we have really generic tests, which test that debugging writ large works, and are usable in a wide variety of configurations. To achieve this, the tests try to make as few assumptions about the environment as possible. - on the other one, we have tests for some very specific scenarios/corner cases/bugs/features. Because the thing they are testing occurs only in very specific circumstances, these tests usually try to insulate themselves from the environment as much as possible. And since these things often require very elaborate setups, it's usually not possible or very difficult to test them using one of other other test suites. We're not currently doing a good job at differentiating the two. It might be the best two split the test suite into two, but that would be a fairly big undertaking, and would involve a lot of hairsplitting, as the line is not always very clear. https://github.com/llvm/llvm-project/pull/92090 _______________________________________________ lldb-commits mailing list lldb-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/lldb-commits
