On Wed, 30 Oct 2024 at 13:11, Boqun Feng <boqun.feng@xxxxxxxxx> wrote: > > On Wed, Oct 30, 2024 at 12:57:14PM +0800, David Gow wrote: > > From: José Expósito <jose.exposito89@xxxxxxxxx> > > Add a new procedural macro (`#[kunit_tests(kunit_test_suit_name)]`) to > > run KUnit tests using a user-space like syntax. > > > > The macro, that should be used on modules, transforms every `#[test]` > > in a `kunit_case!` and adds a `kunit_unsafe_test_suite!` registering > > all of them. > > > > The only difference with user-space tests is that instead of using > > `#[cfg(test)]`, `#[kunit_tests(kunit_test_suit_name)]` is used. > > > > Note that `#[cfg(CONFIG_KUNIT)]` is added so the test module is not > > compiled when `CONFIG_KUNIT` is set to `n`. > > > > Reviewed-by: David Gow <davidgow@xxxxxxxxxx> > > Signed-off-by: José Expósito <jose.exposito89@xxxxxxxxx> > > [Updated to use new const fn.] > > Signed-off-by: David Gow <davidgow@xxxxxxxxxx> > > --- > > > > Changes since v2: > > https://lore.kernel.org/linux-kselftest/20241029092422.2884505-3-davidgow@xxxxxxxxxx/ > > - Include missing rust/macros/kunit.rs file from v2. (Thanks Boqun!) > > - The proc macro now emits an error if the suite name is too long. > > > > Changes since v1: > > https://lore.kernel.org/lkml/20230720-rustbind-v1-2-c80db349e3b5@xxxxxxxxxx/ > > - Rebased on top of rust-next > > - Make use of the new const functions, rather than the kunit_case!() > > macro. > > > > --- > > MAINTAINERS | 1 + > > rust/kernel/kunit.rs | 11 ++++ > > rust/macros/kunit.rs | 153 +++++++++++++++++++++++++++++++++++++++++++ > > rust/macros/lib.rs | 29 ++++++++ > > 4 files changed, 194 insertions(+) > > create mode 100644 rust/macros/kunit.rs > > (...snip...) > > + let new_body: TokenStream = vec![body.stream(), kunit_macros.parse().unwrap()] > > + .into_iter() > > + .collect(); > > + > > + // Remove the `#[test]` macros. > > + let new_body = new_body.to_string().replace("#[test]", ""); > > Yeah, I want to see how you do it this time ;-) So if you do a > `.to_string()` on a `TokenStream`, you lose all the span [1] information > ("span information" is a term invited by me, hope I get it right ;-)) > e.g. if there is a compile error in the test code, the compiler cannot > report the exact line of the error, it can only report there is an > error. > > Last time I find how to preserve the Span: > > https://lore.kernel.org/rust-for-linux/ZMba0_XXZuTgWyWY@boqun-archlinux/ > > Hope it helps! > > [1]: https://doc.rust-lang.org/proc_macro/struct.Span.html > > Regards, > Boqun Thanks. I managed to get this working, but just ended up with an uglier version of your change, so I've copied it in and added you as a co-developer for this patch in v4. It made clippy catch a couple of warnings in the example tests, too, so it's clearly working well. Cheers, -- David
Attachment:
smime.p7s
Description: S/MIME Cryptographic Signature
