Update sccache to its master branch
Ideally I'd like to soon enable sccache for rustbuild itself and some of the
stage0 tools, but for that to work we'll need some better Rust support than the
pretty old version we were previously using!
The docs team has decided that we're framing resources in three ways:
"learning Rust," "using Rust," "mastering Rust." This is a more useful
split than "beginner/intermediate/advanced." As we add more resources
in the future, we expect "using Rust" to grow. "the bookshelf" as a
concept is great, but isn't really organized along these lines. As such,
this reorganizes the docs along these lines.
This is too likely to cause spurious bounces on CI; what we run may be
dependent on what ran successfully before hand (e.g. RLS features with
Clippy), which makes this not tenable. There's no good way to ignore
specifically these problematic steps so we'll just ignore everything for
the time being. We still test that a dry run worked though so largely
this is the same from a ensure-that-tests-work perspective.
Eventually we'll want to undo this commit, though, to make our tests
more accurate.
This ensures that each build will support the testing design of "dry
running" builds. It's also checked that a dry run build is equivalent
step-wise to a "wet" run build; the graphs we generate when running are
directly compared node/node and edge/edge, both for order and contents.
In order to run tests, previous commits have cfg'd out various parts of
rustbuild. Generally speaking, these are filesystem-related operations
and process-spawning related parts. Then, rustbuild is run "as normal"
and the various steps that where run are retrieved from the cache and
checked against the expected results.
Note that this means that the current implementation primarily tests
"what" we build, but doesn't actually test that what we build *will*
build. In other words, it doesn't do any form of dependency verification
for any crate. This is possible to implement, but is considered future
work.
This implementation strives to cfg out as little code as possible; it
also does not currently test anywhere near all of rustbuild. The current
tests are also not checked for "correctness," rather, they simply
represent what we do as of this commit, which may be wrong.
Test cases are drawn from the old implementation of rustbuild, though
the expected results may vary.
This ensures that the working directory of rustbuild has no effect on
it's run; since tests will run with a different cwd this is required for
consistent behavior.
Remove adjacent all-const match arm hack.
An old fix for moves-in-guards had a hack for adjacent all-const match arms.
The hack was explained in a comment, which you can see here:
https://github.com/rust-lang/rust/pull/22580/files#diff-402a0fa4b3c6755c5650027c6d4cf1efR497
But hack was incomplete (and thus unsound), as pointed out here:
https://github.com/rust-lang/rust/issues/47295#issuecomment-357108458
Plus, it is likely to be at least tricky to reimplement this hack in
the new NLL borrowck.
So rather than try to preserve the hack, we want to try to just remove
it outright. (At least to see the results of a crater run.)
[breaking-change]
This is a breaking-change, but our hope is that no one is actually
relying on such an extreme special case. (We hypothesize the hack was
originally added to accommodate a file in our own test suite, not code
in the wild.)
Extend two-phase borrows to apply to method receiver autorefs
Fixes#48598 by permitting two-phase borrows on the autorefs created when functions and methods.
Add Iterator::find_map
I'd like to propose to add `find_map` method to the `Iterator`: an occasionally useful utility, which relates to `filter_map` in the same way that `find` relates to `filter`.
`find_map` takes an `Option`-returning function, applies it to the elements of the iterator, and returns the first non-`None` result. In other words, `find_map(f) == filter_map(f).next()`.
Why do we want to add a function to the `Iterator`, which can be trivially expressed as a combination of existing ones? Observe that `find(f) == filter(f).next()`, so, by the same logic, `find` itself is unnecessary!
The more positive argument is that desugaring of `find[_map]` in terms of `filter[_map]().next()` is not super obvious, because the `filter` operation reads as if it is applies to the whole collection, although in reality we are interested only in the first element. That is, the jump from "I need a **single** result" to "let's use a function which maps **many** values to **many** values" is a non-trivial speed-bump, and causes friction when reading and writing code.
Does the need for `find_map` arise in practice? Yes!
* Anecdotally, I've more than once searched the docs for the function with `[T] -> (T -> Option<U>) -> Option<U>` signature.
* The direct cause for this PR was [this](1291c50e86 (r174934173)) discussion in Cargo, which boils down to "there's some pattern that we try to express here, but current approaches looks non-pretty" (and the pattern is `filter_map`
* There are several `filter_map().next` combos in Cargo: [[1]](545a4a2c93/src/cargo/ops/cargo_new.rs (L585)), [[2]](545a4a2c93/src/cargo/core/resolver/mod.rs (L1130)), [[3]](545a4a2c93/src/cargo/ops/cargo_rustc/mod.rs (L1086)).
* I've also needed similar functionality in `Kotlin` several times. There, it is expressed as `mapNotNull {}.firstOrNull`, as can be seen [here](ee8bdb4e07/src/main/kotlin/org/rust/cargo/project/model/impl/CargoProjectImpl.kt (L154)), [here](ee8bdb4e07/src/main/kotlin/org/rust/lang/core/resolve/ImplLookup.kt (L444)) [here](ee8bdb4e07/src/main/kotlin/org/rust/ide/inspections/RsLint.kt (L38)) and [here](ee8bdb4e07/src/main/kotlin/org/rust/cargo/toolchain/RustToolchain.kt (L74)) (and maybe in some other cases as well)
Note that it is definitely not among the most popular functions (it definitely is less popular than `find`), but, for example it (in case of Cargo) seems to be more popular than `rposition` (1 occurrence), `step_by` (zero occurrences) and `nth` (three occurrences as `nth(0)` which probably should be replaced with `next`).
Do we necessary need this function in `std`? Could we move it to itertools? That is possible, but observe that `filter`, `filter_map`, `find` and `find_map` together really form a complete table:
|||
|-------|---------|
| filter| find|
|filter_map|find_map|
It would be somewhat unsatisfying to have one quarter of this table live elsewhere :) Also, if `Itertools` adds an `find_map` method, it would be more difficult to move it to std due to name collision.
Hm, at this point I've searched for `filter_map` the umpteenth time, and, strangely, this time I do find this RFC: https://github.com/rust-lang/rfcs/issues/1801. I guess this could be an implementation though? :)
To sum up:
Pro:
- complete the symmetry with existing method
- codify a somewhat common non-obvious pattern
Contra:
- niche use case
- we can, and do, live without it
Bump to 1.27.0
Also update some `Cargo.lock` dependencies, finishing up some final steps of our
[release process]!
This doesn't update the bootstrap compiler just yet but that will come in a
follow-up PR.
[release process]: https://forge.rust-lang.org/release-process.html
This commit is a reorganization of the `proc_macro` crate's public user-facing
API. This is the result of a number of discussions at the recent Rust All-Hands
where we're hoping to get the `proc_macro` crate into ship shape for
stabilization of a subset of its functionality in the Rust 2018 release.
The reorganization here is motivated by experiences from the `proc-macro2`,
`quote`, and `syn` crates on crates.io (and other crates which depend on them).
The main focus is future flexibility along with making a few more operations
consistent and/or fixing bugs. A summary of the changes made from today's
`proc_macro` API is:
* The `TokenNode` enum has been removed and the public fields of `TokenTree`
have also been removed. Instead the `TokenTree` type is now a public enum
(what `TokenNode` was) and each variant is an opaque struct which internally
contains `Span` information. This makes the various tokens a bit more
consistent, require fewer wrappers, and otherwise provides good
future-compatibility as opaque structs are easy to modify later on.
* `Literal` integer constructors have been expanded to be unambiguous as to what
they're doing and also allow for more future flexibility. Previously
constructors like `Literal::float` and `Literal::integer` were used to create
unsuffixed literals and the concrete methods like `Literal::i32` would create
a suffixed token. This wasn't immediately clear to all users (the
suffixed/unsuffixed aspect) and having *one* constructor for unsuffixed
literals required us to pick a largest type which may not always be true. To
fix these issues all constructors are now of the form
`Literal::i32_unsuffixed` or `Literal::i32_suffixed` (for all integral types).
This should allow future compatibility as well as being immediately clear
what's suffixed and what isn't.
* Each variant of `TokenTree` internally contains a `Span` which can also be
configured via `set_span`. For example `Literal` and `Term` now both
internally contain a `Span` rather than having it stored in an auxiliary
location.
* Constructors of all tokens are called `new` now (aka `Term::intern` is gone)
and most do not take spans. Manufactured tokens typically don't have a fresh
span to go with them and the span is purely used for error-reporting
**except** the span for `Term`, which currently affects hygiene. The default
spans for all these constructed tokens is `Span::call_site()` for now.
The `Term` type's constructor explicitly requires passing in a `Span` to
provide future-proofing against possible hygiene changes. It's intended that a
first pass of stabilization will likely only stabilize `Span::call_site()`
which is an explicit opt-in for "I would like no hygiene here please". The
intention here is to make this explicit in procedural macros to be
forwards-compatible with a hygiene-specifying solution.
* Some of the conversions for `TokenStream` have been simplified a little.
* The `TokenTreeIter` iterator was renamed to `token_stream::IntoIter`.
Overall the hope is that this is the "final pass" at the API of `TokenStream`
and most of `TokenTree` before stabilization. Explicitly left out here is any
changes to `Span`'s API which will likely need to be re-evaluated before
stabilization.
All changes in this PR have already been reflected to the [`proc-macro2`],
`quote`, and `syn` crates. New versions of all these crates have also been
published to crates.io.
Once this lands in nightly I plan on making an internals post again summarizing
the changes made here and also calling on all macro authors to give the APIs a
spin and see how they work. Hopefully pending no major issues we can then have
an FCP to stabilize later this cycle!
[`proc-macro2`]: https://docs.rs/proc-macro2/0.3.1/proc_macro2/
Also update some `Cargo.lock` dependencies, finishing up some final steps of our
[release process]!
This doesn't update the bootstrap compiler just yet but that will come in a
follow-up PR.
[release process]: https://forge.rust-lang.org/release-process.html
Ideally I'd like to soon enable sccache for rustbuild itself and some of the
stage0 tools, but for that to work we'll need some better Rust support than the
pretty old version we were previously using!
