c9f66fd425
This is the result of running `dprint fmt` after removing `src/` from
the list of excluded directories.
This also reformats the Rust code: we might want to tweak this a bit in
the future since some of the changes removes the hand-formatting. Of
course, this formatting can be seen as a mis-feature, so maybe this is
good overall.
Thanks to mdbook-i18n-helpers 0.2, the POT file is nearly unchanged
after this, meaning that all existing translations remain valid! A few
messages were changed because of stray whitespace characters:
msgid ""
"Slices always borrow from another object. In this example, `a` has to remain "
-"'alive' (in scope) for at least as long as our slice. "
+"'alive' (in scope) for at least as long as our slice."
msgstr ""
The formatting is enforced in CI and we will have to see how annoying
this is in practice for the many contributors. If it becomes annoying,
we should look into fixing dprint/check#11 so that `dprint` can annotate
the lines that need fixing directly, then I think we can consider more
strict formatting checks.
I added more customization to `rustfmt.toml`. This is to better emulate
the dense style used in the course:
- `max_width = 85` allows lines to take up the full width available in
our code blocks (when taking margins and the line numbers into account).
- `wrap_comments = true` ensures that we don't show very long comments
in the code examples. I edited some comments to shorten them and avoid
unnecessary line breaks — please trim other unnecessarily long comments
when you see them! Remember we're writing code for slides 😄
- `use_small_heuristics = "Max"` allows for things like struct literals
and if-statements to take up the full line width configured above.
The formatting settings apply to all our Rust code right now — I think
we could improve this with https://github.com/dprint/dprint/issues/711
which lets us add per-directory `dprint` configuration files. However,
the `inherit: true` setting is not yet implemented (as far as I can
tell), so a nested configuration file will have to copy most or all of
the top-level file.
1.9 KiB
Testing
Rust community typically authors unit tests in a module placed in the same source file as the code being tested. This was covered earlier in the course and looks like this:
#[cfg(test)]
mod tests {
#[test]
fn my_test() {
todo!()
}
}
In Chromium we place unit tests in a separate source file and we continue to
follow this practice for Rust --- this makes tests consistently discoverable and
helps to avoid rebuilding .rs files a second time (in the test
configuration).
This results in the following options for testing Rust code in Chromium:
- Native Rust tests (i.e.
#[test]). Discouraged outside of//third_party/rust. gtesttests authored in C++ and exercising Rust via FFI calls. Sufficient when Rust code is just a thin FFI layer and the existing unit tests provide sufficient coverage for the feature.gtesttests authored in Rust and using the crate under test through its public API (usingpub mod for_testing { ... }if needed). This is the subject of the next few slides.
Mention that native Rust tests of third-party crates should eventually be exercised by Chromium bots. (Such testing is needed rarely --- only after adding or updating third-party crates.)
Some examples may help illustrate when C++ gtest vs Rust gtest should be
used:
-
QR has very little functionality in the first-party Rust layer (it's just a thin FFI glue) and therefore uses the existing C++ unit tests for testing both the C++ and the Rust implementation (parameterizing the tests so they enable or disable Rust using a
ScopedFeatureList). -
Hypothetical/WIP PNG integration may need to implement memory-safe implementation of pixel transformations that are provided by
libpngbut missing in thepngcrate - e.g. RGBA => BGRA, or gamma correction. Such functionality may benefit from separate tests authored in Rust.