Disabling the redzone is required in x86-64's kernel mode to avoid interrupts trashing the stack.

Signed-off-by: NAnton Lofgren <alofgren@op5.com>

Use one or more of the following `-Z` flag options to tell the
graphviz renderer to include the corresponding dataflow sets (after
the iterative constraint propagation reaches a fixed-point solution):

  * `-Z flowgraph-print-loans` : loans computed via middle::borrowck
  * `-Z flowgraph-print-moves` : moves computed via middle::borrowck::move_data
  * `-Z flowgraph-print-assigns` : assignments, via middle::borrowck::move_data
  * `-Z flowgraph-print-all` : all of the available sets are included.

Fix #15016.

Use one or more of the following `-Z` flag options to tell the
graphviz renderer to include the corresponding dataflow sets (after
the iterative constraint propagation reaches a fixed-point solution):

  * `-Z flowgraph-print-loans` : loans computed via middle::borrowck
  * `-Z flowgraph-print-moves` : moves computed via middle::borrowck::move_data
  * `-Z flowgraph-print-assigns` : assignments, via middle::borrowck::move_data
  * `-Z flowgraph-print-all` : all of the available sets are included.

Fix #15016.

----

This also adds a module, `syntax::ast_map::blocks`, that captures a
common abstraction shared amongst code blocks and procedure-like
things.  As part of this, moved `ast_map.rs` to subdir
`ast_map/mod.rs`, to follow our directory layout conventions.

(incorporated review feedback from huon, acrichto.)

This was my weekend project, to start breaking up rustc. It first pulls out LLVM into `rustc_llvm`, then parts of `rustc::back` and `rustc::util` to `rustc_back`. The immediate intent is just to reduce the size of rustc, to reduce memory pressure when building rustc, but this is also a good starting point for further refactoring.

The `rustc_back` crate is definitely misnamed (`rustc::back` was never a very cohesive module anyway) - it's mostly just somewhere to stuff parts of rustc that don't have many deps. Right now it's main dep is `syntax`; it has no dep on `rustc_llvm`.

Some next steps might be to split `rustc_back` into `rustc_util` (with no `syntax` dep), and `rustc_syntax_util` (with a syntax dep); move the rest of `rustc::util` into `rustc_syntax_util`; move all of `rustc::front` to a new crate, `rustc_front`. At that point the refactoring necessary to keep extracting crates will get harder.

Currently when a read-only file has unlink() invoked on it on windows, the call
will fail. On unix, however, the call will succeed. In order to have a more
consistent behavior across platforms, this error is recognized on windows and
the file is changed to read-write before removal is attempted.

Currently when a read-only file has unlink() invoked on it on windows, the call
will fail. On unix, however, the call will succeed. In order to have a more
consistent behavior across platforms, this error is recognized on windows and
the file is changed to read-write before removal is attempted.

This makes it much saner for clients to use the library since
they don't have to worry about shadowing one llvm with another.

Just some leftover junk from extracting llvm.

Indicate that anything that can be extracted here should and that things with
syntax deps should be split out someday.

It is a wrapper around LLVM.

Not sure how to test this correctly I assume the current tests pass now because of the crate boundaries [and that this is fallout from private by default]?

I saw that it was bounded by `Show` but the implication is no guarantee (and had only 0.10 to test, where this behavior has been added to 0.11)

It is implied by the Show bound, but that implication can be missed.

  - for 3 implementations of into_maybe_owned()
  - is_slice()
  - is_owned()

This patch adds doc examples for the make_absolute, change_dir,
errors_string and args functions in the os module.

In order to have the spawning semantics be the same for unix/windows, the
child's PATH environment variable needs to be searched rather than the parent's
environment variable.

If the child is inheriting the parent's PATH, then no action need be taken as
windows will do the heavy lifting. If the child specifies its own PATH, then it
is searched beforehand for the target program and the result is favored if a hit
is found.

cc #15149, but does not close the issue because libgreen still needs to be
updated.

In order to have the spawning semantics be the same for unix/windows, the
child's PATH environment variable needs to be searched rather than the parent's
environment variable.

If the child is inheriting the parent's PATH, then no action need be taken as
windows will do the heavy lifting. If the child specifies its own PATH, then it
is searched beforehand for the target program and the result is favored if a hit
is found.

cc #15149, but does not close the issue because libgreen still needs to be
updated.

This PR is the outcome of the library stabilization meeting for the
`liballoc::owned` and `libcore::cell` modules.

Aside from the stability attributes, there are a few breaking changes:

* The `owned` modules is now named `boxed`, to better represent its
  contents. (`box` was unavailable, since it's a keyword.) This will
  help avoid the misconception that `Box` plays a special role wrt
  ownership.

* The `AnyOwnExt` extension trait is renamed to `BoxAny`, and its `move`
  method is renamed to `downcast`, in both cases to improve clarity.

* The recently-added `AnySendOwnExt` extension trait is removed; it was
  not being used and is unnecessary.

[breaking-change]

This PR is the outcome of the library stabilization meeting for the
`liballoc::owned` and `libcore::cell` modules.

Aside from the stability attributes, there are a few breaking changes:

* The `owned` modules is now named `boxed`, to better represent its
  contents. (`box` was unavailable, since it's a keyword.) This will
  help avoid the misconception that `Box` plays a special role wrt
  ownership.

* The `AnyOwnExt` extension trait is renamed to `BoxAny`, and its `move`
  method is renamed to `downcast`, in both cases to improve clarity.

* The recently-added `AnySendOwnExt` extension trait is removed; it was
  not being used and is unnecessary.

[breaking-change]

This patch adds doc examples for the make_absolute, change_dir,
errors_string and args functions in the os module.

This patch adds support for macros in method position. It follows roughly the template for Item macros, where an outer `Method` wrapper contains a `Method_` enum which can either be a macro invocation or a standard macro definition. 

One note; adding support for macros that expand into multiple methods is not included here, but should be a simple parser change, since this patch updates the type of fold_macro to return a smallvector of methods.

For reviewers, please pay special attention to the parser changes; these are the ones I'm most concerned about.

Because of the small change to the interface of fold_method, this is a ...

[breaking change]

Closes #4621