On Sun, May 4, 2025 at 2:31 PM Miguel Ojeda <[email protected]> wrote: > > On Sun, May 4, 2025 at 7:31 PM Tamir Duberstein <[email protected]> wrote: > > > > Is that true? The build host is often easier to work with. There's a > > number of host tests on the C side that exist precisely for this > > reason. > > Even for tests that could run in the host (pure functions), if you > test in the host, then you are not testing the actual kernel code, in > the sense of same compile flags, target, etc. > > Moreover, you have UML, which gives you access to other APIs. > > As for "easier to work with", I am not sure what you mean -- KUnit > does not really require anything special w.r.t. building the kernel > normally. In a way, these restricted host tests actually are an extra > hassle, in that you have to deal with yet another test environment and > special restrictions.
All good points. > But which host tests are you referring to? One example is https://github.com/torvalds/linux/blob/59c9ab3e8cc7f56cd65608f6e938b5ae96eb9cd2/tools/testing/radix-tree/xarray.c. It might be that these are necessary because the xarray tests don't use kunit, and so are pretty inconvenient to run. As you might have guessed, I discovered these host tests when my patch porting the xarray tests to kunit broke the host-side build :(
