With this change, rustc creates a unique type identifier for types in debuginfo. These type identifiers are used by LLVM to correctly handle link-time-optimization scenarios but also help rustc with dealing with inlining from other crates. For more information, see the documentation block at the top of librustc/middle/trans/debuginfo.rs and also [my blog post about the topic](http://michaelwoerister.github.io/2014/06/05/rust-debuginfo-and-unique-type-identifiers.html). This should fix the LTO issues that have been popping up lately.
The changes to the debuginfo module's inner workings are also improved by this. Metadata uniquing of pointer types is not handled explicitly instead of relying on LLVM doing the right thing behind the scenes, and region parameters on types should not lead to metadata duplication anymore.
There are two things that I'd like to get some feedback on:
1. IDs for named items consist of two parts: The [Strict Version Hash](https://github.com/mozilla/rust/blob/0.10/src/librustc/back/svh.rs#L11) of their defining crate and the AST node id of their definition within that crate. My question is: Is the SVH a good choice for identifying the crate? Is it even going to stay? The [crate-id RFC](https://github.com/rust-lang/rfcs/pull/109) got me confused.
2. Unique Type Identifiers can be arbitrary strings and right now the format is rather verbose. For debugging this is nice, because one can infer a lot about a type from the type id alone (it's more or less a signature). For deeply nested generics, id strings could get rather long though. One option to limit the id size would be to use some hashcode instead of the full id (anything that avoids collision as much as possible). Another option would be to use a more compact representation, like ty_encode. This reduces size but also readability.
Since these ID's only show up in LLVM IR, I'm inclined to just leave in the verbose format for now, and only act if sizes of rlibs become a problem.
With this change, rustc creates a unique type identifier for types in debuginfo. These type identifiers are used by LLVM to correctly handle link-time-optimization scenarios but also help rustc with dealing with inlining from other crates. For more information, see the documentation block at the top of librustc/middle/trans/debuginfo.rs.
Fixes#13681.
Previously, the type system's restrictions on borrowing were summarized as
> The previous example showed that the type system forbids any borrowing of owned boxes found in aliasable, mutable memory.
This did not jive with the example, which allowed mutations so long as the borrowed reference had been returned. Also, the language has changed to no longer allow aliasable mutable locations. This changes the summary to read
> The previous example showed that the type system forbids mutations of owned boxed values while they are being borrowed. In general, the type system also forbids borrowing a value as mutable if it is already being borrowed - either as a mutable reference or an immutable one.
This adds more general information for the experienced reader as well, to offer a more complete understanding.
The guide previously stated:
> The compiler will automatically convert a box box point to a reference like &point.
This fixes the doubled word `box`, so the statement reads
> The compiler will automatically convert a box point to a reference like &point.
The code it is referring to is `compute_distance(&on_the_stack, on_the_heap);`, so a single `box` is appropriate.
This configures the makefiles to copy a local jemalloc/libuv library into place instead of building the local copy of one. Additionally, this switches our travis builds to using the system-provided jemalloc instead of a custom-built jemalloc to exercise this functionality.
This adds a new configure option, --jemalloc-root, which will specify a location
at which libjemalloc_pic.a must live. This library is then used for the build
triple as the jemalloc library to link.
* The select/plural methods from format strings are removed
* The # character no longer needs to be escaped
* The \-based escapes have been removed
* '{{' is now an escape for '{'
* '}}' is now an escape for '}'
Closes#14810
[breaking-change]
The following features have been removed
* `box [a, b, c]`
* `~[a, b, c]`
* `box [a, ..N]`
* `~[a, ..N]`
* `~[T]` (as a type)
* deprecated_owned_vector lint
All users of ~[T] should move to using Vec<T> instead.
The following features have been removed
* box [a, b, c]
* ~[a, b, c]
* box [a, ..N]
* ~[a, ..N]
* ~[T] (as a type)
* deprecated_owned_vector lint
All users of ~[T] should move to using Vec<T> instead.
This commit is the final step in the libstd facade, #13851. The purpose of this
commit is to move libsync underneath the standard library, behind the facade.
This will allow core primitives like channels, queues, and atomics to all live
in the same location.
There were a few notable changes and a few breaking changes as part of this
movement:
* The `Vec` and `String` types are reexported at the top level of libcollections
* The `unreachable!()` macro was copied to libcore
* The `std::rt::thread` module was moved to librustrt, but it is still
reexported at the same location.
* The `std::comm` module was moved to libsync
* The `sync::comm` module was moved under `sync::comm`, and renamed to `duplex`.
It is now a private module with types/functions being reexported under
`sync::comm`. This is a breaking change for any existing users of duplex
streams.
* All concurrent queues/deques were moved directly under libsync. They are also
all marked with #![experimental] for now if they are public.
* The `task_pool` and `future` modules no longer live in libsync, but rather
live under `std::sync`. They will forever live at this location, but they may
move to libsync if the `std::task` module moves as well.
[breaking-change]
This commit removes `@T` from the compiler by moving the AST to using `Gc<T>`. This also starts treating `Gc<T>` as `@T` in the same way that `Box<T>` is the same as `~T` in the compiler.
After this hits a snapshot, the `@T` syntax should be able to be removed completely.
This commit is the final step in the libstd facade, #13851. The purpose of this
commit is to move libsync underneath the standard library, behind the facade.
This will allow core primitives like channels, queues, and atomics to all live
in the same location.
There were a few notable changes and a few breaking changes as part of this
movement:
* The `Vec` and `String` types are reexported at the top level of libcollections
* The `unreachable!()` macro was copied to libcore
* The `std::rt::thread` module was moved to librustrt, but it is still
reexported at the same location.
* The `std::comm` module was moved to libsync
* The `sync::comm` module was moved under `sync::comm`, and renamed to `duplex`.
It is now a private module with types/functions being reexported under
`sync::comm`. This is a breaking change for any existing users of duplex
streams.
* All concurrent queues/deques were moved directly under libsync. They are also
all marked with #![experimental] for now if they are public.
* The `task_pool` and `future` modules no longer live in libsync, but rather
live under `std::sync`. They will forever live at this location, but they may
move to libsync if the `std::task` module moves as well.
[breaking-change]
This commit uses the same trick as ~/Box to map Gc<T> to @T internally inside
the compiler. This moves a number of implementations of traits to the `gc`
module in the standard library.
This removes functions such as `Gc::new`, `Gc::borrow`, and `Gc::ptr_eq` in
favor of the more modern equivalents, `box(GC)`, `Deref`, and pointer equality.
The Gc pointer itself should be much more useful now, and subsequent commits
will move the compiler away from @T towards Gc<T>
[breaking-change]
If this breaks your code, take a deep breath, go for a walk, and
consider why you're relying on the sign extension semantics of
enum-to-float casts.
[breaking-change]
Closes#8230.
When generating documentation, rustdoc has the ability to generate relative
links within the current distribution of crates to one another. To do this, it
must recognize when a crate's documentation is in the same output directory. The
current threshold for "local documentation for crate X being available" is
whether the directory "doc/X" exists.
This change modifies the build system to have new dependencies for each
directory of upstream crates for a rustdoc invocation. This will ensure that
when building documentation that all the crates in the standard distribution are
guaranteed to have relative links to one another.
This change is prompted by guaranteeing that offline docs always work with one
another. Before this change, races could mean that some docs were built before
others, and hence may have http links when relative links would suffice.
Closes#14747
**Update**
I've reimplemented this using `Cell` and `RefCell`, as suggested by @alexcrichton. By taking care with the duration of the borrows, I was able to maintain the recursive allocation feature (now covered by a test) without the use of `Unsafe`, and without breaking the non-aliasing `&mut` invariant.
**Original**
Changes both `Arena` and `TypedArena` to contain an inner struct wrapped in a `Unsafe`, and change field access to go through those instead of transmuting `&self` to `&mut self`.
Part of #13933
