In https://github.com/rustls/rustls-ffi/pull/348 I neglected to update the callsites that were previously using make test so they would run the new Rust-based integration tests (behind make integration). It also seems like the Windows makefile test target wasn't being exercised at all (there was a missing required env var for the client.exe invocation).
This branch fixes that, and along the way uncovered that the MacOS and Windows linker parts have drifted from expected. Those are updated in this branch as well. Unfortunately it seems like the Windows CMake builds disagree on the linker parts compared to the Windows Makefile builds. For now I've made the CMake builds only run the client/server integration test, and let the Makefile builds enforce the expected linker parts. We could also consider maintaining two different expected_linker_parts returns, but doing that will be a bit tricky since right now the expected values are chosen by target_os alone, and don't know about the build tooling in-use.
In https://github.com/rustls/rustls-ffi/pull/348 I neglected to update the callsites that were previously using
make test
so they would run the new Rust-based integration tests (behindmake integration
). It also seems like the Windows makefiletest
target wasn't being exercised at all (there was a missing required env var for the client.exe invocation).This branch fixes that, and along the way uncovered that the MacOS and Windows linker parts have drifted from expected. Those are updated in this branch as well. Unfortunately it seems like the Windows CMake builds disagree on the linker parts compared to the Windows Makefile builds. For now I've made the CMake builds only run the client/server integration test, and let the Makefile builds enforce the expected linker parts. We could also consider maintaining two different
expected_linker_parts
returns, but doing that will be a bit tricky since right now the expected values are chosen bytarget_os
alone, and don't know about the build tooling in-use.