This is enough for dynamic libraries, but not executables because MinGW
does not output a .reloc section even with `--dynamicbase`. It could
either be worked around by exporting a DLL symbol from the executable or
fixed in MinGW itself.
This is enough for dynamic libraries, but not executables because MinGW
does not output a .reloc section even with `--dynamicbase`. It could
either be worked around by exporting a DLL symbol from the executable or
fixed in MinGW itself.
declared with the same name in the same scope.
This breaks several common patterns. First are unused imports:
use foo::bar;
use baz::bar;
Change this code to the following:
use baz::bar;
Second, this patch breaks globs that import names that are shadowed by
subsequent imports. For example:
use foo::*; // including `bar`
use baz::bar;
Change this code to remove the glob:
use foo::{boo, quux};
use baz::bar;
Or qualify all uses of `bar`:
use foo::{boo, quux};
use baz;
... baz::bar ...
Finally, this patch breaks code that, at top level, explicitly imports
`std` and doesn't disable the prelude.
extern crate std;
Because the prelude imports `std` implicitly, there is no need to
explicitly import it; just remove such directives.
The old behavior can be opted into via the `import_shadowing` feature
gate. Use of this feature gate is discouraged.
This implements RFC #116.
Closes#16464.
[breaking-change]
Crates that are resolved normally have their path canonicalized and all
symlinks resolved. This does currently not happen for paths specified
using the --extern option to rustc, which can lead to rustc thinking
that it encountered two different versions of a crate, when it's
actually the same version found through different paths.
Fixes#16496
These are already marked as `noalias` due to the immutability guarantee
(see 4c2d4cd3de), but more information can
be bubbled up to the caller via `readonly`.
When a struct implements Drop, its fields should still drop in
declaration order (just as they do when the struct does not implement
Drop).
Fixes#16492.
When a struct implements Drop, its fields should still drop in
declaration order (just as they do when the struct does not implement
Drop).
Fixes#16492.
The discriminant for Option values is either 0 or 1, so we can just
truncate the value to an i1, which ends up as a no-op for Options
containing pointers.
These are already marked as `noalias` due to the immutability guarantee
(see 4c2d4cd3de), but more information can
be bubbled up to the caller via `readonly`.
Crates that are resolved normally have their path canonicalized and all
symlinks resolved. This does currently not happen for paths specified
using the --extern option to rustc, which can lead to rustc thinking
that it encountered two different versions of a crate, when it's
actually the same version found through different paths.
To fix this, we must store the canonical path for crates found via
--extern and also use the canonical path when comparing paths.
Fixes#16496
The discriminant for Option values is either 0 or 1, so we can just
truncate the value to an i1, which ends up as a no-op for Options
containing pointers.
These `where` clauses are accepted everywhere generics are currently
accepted and desugar during type collection to the type parameter bounds
we have today.
A new keyword, `where`, has been added. Therefore, this is a breaking
change. Change uses of `where` to other identifiers.
[breaking-change]
r? @nikomatsakis (or whoever)
These `where` clauses are accepted everywhere generics are currently
accepted and desugar during type collection to the type parameter bounds
we have today.
A new keyword, `where`, has been added. Therefore, this is a breaking
change. Change uses of `where` to other identifiers.
[breaking-change]
methods.
This paves the way to associated items by introducing an extra level of
abstraction ("impl-or-trait item") between traits/implementations and
methods. This new abstraction is encoded in the metadata and used
throughout the compiler where appropriate.
There are no functional changes; this is purely a refactoring.
This patch primarily does two things: (1) it prevents lifetimes from
leaking out of unboxed closures; (2) it allows unboxed closure type
notation, call notation, and construction notation to construct closures
matching any of the three traits.
This breaks code that looked like:
let mut f;
{
let x = &5i;
f = |&mut:| *x + 10;
}
Change this code to avoid having a reference escape. For example:
{
let x = &5i;
let mut f; // <-- move here to avoid dangling reference
f = |&mut:| *x + 10;
}
I believe this is enough to consider unboxed closures essentially
implemented. Further issues (for example, higher-rank lifetimes) should
be filed as followups.
Closes#14449.
[breaking-change]
r? @pnkfelix
This patch primarily does two things: (1) it prevents lifetimes from
leaking out of unboxed closures; (2) it allows unboxed closure type
notation, call notation, and construction notation to construct closures
matching any of the three traits.
This breaks code that looked like:
let mut f;
{
let x = &5i;
f = |&mut:| *x + 10;
}
Change this code to avoid having a reference escape. For example:
{
let x = &5i;
let mut f; // <-- move here to avoid dangling reference
f = |&mut:| *x + 10;
}
I believe this is enough to consider unboxed closures essentially
implemented. Further issues (for example, higher-rank lifetimes) should
be filed as followups.
Closes#14449.
[breaking-change]
by-reference upvars.
This partially implements RFC 38. A snapshot will be needed to turn this
on, because stage0 cannot yet parse the keyword.
Part of #12381.
This required some contortions because importing both raw::Slice
and slice::Slice makes rustc crash.
Since `Slice` is in the prelude, this renaming is unlikely to
casue breakage.
[breaking-change]
This fixes borrow checking for closures. Code like this will break:
struct Foo {
x: int,
}
pub fn main() {
let mut this = &mut Foo {
x: 1,
};
let r = || {
let p = &this.x;
&mut this.x;
};
r()
}
Change this code to not take multiple mutable references to the same value. For
example:
struct Foo {
x: int,
}
pub fn main() {
let mut this = &mut Foo {
x: 1,
};
let r = || {
&mut this.x;
};
r()
}
Closes#16361.
[breaking-change]
r? @nikomatsakis
`for` loop heads.
This breaks code like:
let x = Some(box 1i);
for &a in x.iter() {
}
Change this code to obey the borrow checking rules. For example:
let x = Some(box 1i);
for &ref a in x.iter() {
}
Closes#16205.
[breaking-change]
r? @nikomatsakis
This code produces an ICE:
```rust
#![crate_type = "rlib"]
fn main() {
if true { return }
// remaining code is unreachable
match () {
() => { static MAGIC: uint = 0; }
}
}
```
([playpen](http://is.gd/iwOISB))
The error is "encode_symbol: id not found 18", where 18 is the `NodeId` of the declaration of `MAGIC`. The problem is that `rustc` tries to emit metadata for `MAGIC`, but some of the information is missing because `MAGIC` never gets translated by `trans_item` - the entire body of the `match` gets skipped because the `match` itself is unreachable.
This branch simplifies the handling of inner items by always processing them using the `trans_item` visitor, instead of sometimes using the visitor and sometimes waiting until `trans_stmt` encounters the item. This fixes the ICE by making the translation of the item no longer depend on the declaration being reachable code. This branch also reverts #16059 and #16359, since the new change to item translation fixes the same problems as those but is simpler.
`for` loop heads.
This breaks code like:
let x = Some(box 1i);
for &a in x.iter() {
}
Change this code to obey the borrow checking rules. For example:
let x = Some(box 1i);
for &ref a in x.iter() {
}
Closes#16205.
[breaking-change]
This fixes borrow checking for closures. Code like this will break:
struct Foo {
x: int,
}
pub fn main() {
let mut this = &mut Foo {
x: 1,
};
let r = || {
let p = &this.x;
&mut this.x;
};
r()
}
Change this code to not take multiple mutable references to the same value. For
example:
struct Foo {
x: int,
}
pub fn main() {
let mut this = &mut Foo {
x: 1,
};
let r = || {
&mut this.x;
};
r()
}
Closes#16361.
[breaking-change]
Previously the lint considered cross-crate items only. That's
appropriate for unstable and experimental levels, but not for
deprecation.
Closes#16409
Due to deny(deprecation), this is a:
[breaking-change]
For historical reasons, "Win32" has been used in Rust codebase to mean "Windows OS in general".
This is confusing, especially now, that Rust supports Win64 builds.
[breaking-change]
The fail macro defines some function/static items internally, which got
a dead_code warning when `fail!()` is used inside a dead function. This
is ugly and unnecessarily reveals implementation details, so the
warnings can be squashed.
Fixes#16192.
The fail macro defines some function/static items internally, which got
a dead_code warning when `fail!()` is used inside a dead function. This
is ugly and unnecessarily reveals implementation details, so the
warnings can be squashed.
Fixes#16192.
This generalises the behaviour with struct fields (which recieve no
dead_code warning if they have a leading _), and other similar lints, to
all items, e.g. `fn _foo() {} fn main() {}` has no warnings.
Rust already builds all code as position independent by default, so the
linker can be told to build a position independent executable if it's
not disabled with `-C relocation-model=dynamic-no-pic`. Position
independent code does have a significant cost on i686 (not on x86_64 or
ARM) but there's no significant cost to linking code that's already
position independent as a position independent executable.
Address space layout randomization makes exploiting vulnerabilities much
more difficult by providing a statistical defence against an attempt to
find or modify existing code / data. Without ASLR, it's trivial to use a
vulnerability to take over control of the process via return-oriented
programming.
Rust code can be used for return-oriented programming whether it is safe
or unsafe, so even a fully safe application needs to be built as a
position independent executable to defend against vulnerabilities in
unsafe blocks or C libraries.
Sample program:
extern crate libc;
use std::mem;
static mut global: u32 = 5;
static constant: u32 = 5;
fn foo() {}
fn main() {
let local = 5;
println!("stack: {}, global: {}, constant: {}, fn: {}, lib fn: {}",
&local as *const u32,
unsafe { &global as *const u32 },
&constant as *const u32,
unsafe { mem::transmute::<_, *const ()>(foo) },
unsafe { mem::transmute::<_, *const ()>(libc::mprotect) });
}
Before:
stack: 0x3ff15eb9f94, global: 0x6ab488, constant: 0x47db40, fn: 0x4030e0, lib fn: 0x32749547530
stack: 0x3b5d47d80e4, global: 0x6ab488, constant: 0x47db40, fn: 0x4030e0, lib fn: 0x394469a7530
stack: 0x3fe2c4e5564, global: 0x6ab488, constant: 0x47db40, fn: 0x4030e0, lib fn: 0x399734a2530
stack: 0x3e525e0fb24, global: 0x6ab488, constant: 0x47db40, fn: 0x4030e0, lib fn: 0x2f62a810530
stack: 0x3b50fb3eae4, global: 0x6ab488, constant: 0x47db40, fn: 0x4030e0, lib fn: 0x2e590e86530
After:
stack: 0x38cf12c90a4, global: 0x3e2d46b488, constant: 0x3e2d23cf80, fn: 0x3e2d1c2510, lib fn: 0x2617d3b4530
stack: 0x3d733faf474, global: 0x7eb1839488, constant: 0x7eb160af80, fn: 0x7eb1590510, lib fn: 0x32d30c1f530
stack: 0x3bb42212ec4, global: 0x5bbb365488, constant: 0x5bbb136f80, fn: 0x5bbb0bc510, lib fn: 0x3595e6c1530
stack: 0x39f678c1ab4, global: 0x22c4e3c488, constant: 0x22c4c0df80, fn: 0x22c4b93510, lib fn: 0x3835b727530
stack: 0x3afb25bd394, global: 0x493eab2488, constant: 0x493e883f80, fn: 0x493e809510, lib fn: 0x3478d6a7530
This may also be necessary on other platforms, but I can only test on
Linux right now. Note that GDB gained support for debugging position
independent executables in version 7.1 (March 2010).
Extended `ast_map::Map` with an iterator over all node id's that match a path suffix.
Extended pretty printer to let users choose particular items to pretty print, either by indicating an integer node-id, or by providing a path suffix.
* Example 1: the suffix `typeck::check::check_struct` matches the item with the path `rustc::middle::typeck::check::check_struct` when compiling the `rustc` crate.
* Example 2: the suffix `and` matches `core::option::Option::and` and `core::result::Result::and` when compiling the `core` crate.
Refactored `pprust` slightly to support the pretty printer changes.
(See individual commits for more description.)
With this change:
* `--pretty variant=<node-id>` will print the item associated with
`<node-id>` (where `<node-id>` is an integer for some node-id in
the AST, and `variant` means one of {`normal`,`expanded`,...}).
* `--pretty variant=<path-suffix>` will print all of the items that
match the `<path-suffix>` (where `<path-suffix>` is a suffix of a
path, and `variant` again means one of {`normal`,`expanded`,...}).
Example 1: the suffix `typeck::check::check_struct` matches the
item with the path `rustc::middle::typeck::check::check_struct`
when compiling the `rustc` crate.
Example 2: the suffix `and` matches `core::option::Option::and`
and `core::result::Result::and` when compiling the `core` crate.
Both of the `--pretty variant=...` modes will include the full path to
the item in a comment that follows the item.
Note that when multiple paths match, then either:
1. all matching items are printed, in series; this is what happens in
the usual pretty-print variants, or
2. the compiler signals an error; this is what happens in flowgraph
printing.
----
Some drive-by improvements:
Heavily refactored the pretty-printing glue in driver.rs, introducing
a couple local traits to avoid cut-and-pasting very code segments that
differed only in how they accessed the `Session` or the
`ast_map::Map`. (Note the previous code had three similar calls to
`print_crate` which have all been unified in this revision; the
addition of printing individual node-ids exacerbated the situation
beyond tolerance.) We may want to consider promoting some of these
traits, e.g. `SessionCarrier`, for use more generally elsewhere in the
compiler; right now I have to double check how to access the `Session`
depending on what context I am hacking in.
Refactored `PpMode` to make the data directly reflect the fundamental
difference in the categories (in terms of printing source-code with
various annotations, versus printing a control-flow graph).
(also, addressed review feedback.)
This requires avoiding `quote_...!` for constructing the parts of the
__test module, since that stringifies and reinterns the idents, losing
the special gensym'd nature of them. (#15962.)
This leaves the `Share` trait at `std::kinds` via a `#[deprecated]` `pub use`
statement, but the `NoShare` struct is no longer part of `std::kinds::marker`
due to #12660 (the build cannot bootstrap otherwise).
All code referencing the `Share` trait should now reference the `Sync` trait,
and all code referencing the `NoShare` type should now reference the `NoSync`
type. The functionality and meaning of this trait have not changed, only the
naming.
Closes#16281
[breaking-change]
Rust already builds all code as position independent by default, so the
linker can be told to build a position independent executable if it's
not disabled with `-C relocation-model=dynamic-no-pic`. Position
independent code does have a significant cost on i686 (not on x86_64 or
ARM) but there's no significant cost to linking code that's already
position independent as a position independent executable.
Address space layout randomization makes exploiting vulnerabilities much
more difficult by providing a statistical defence against an attempt to
find or modify existing code / data. Without ASLR, it's trivial to use a
vulnerability to take over control of the process via return-oriented
programming.
Rust code can be used for return-oriented programming whether it is safe
or unsafe, so even a fully safe application needs to be built as a
position independent executable to defend against vulnerabilities in
unsafe blocks or C libraries.
Sample program:
extern crate libc;
use std::mem;
static mut global: u32 = 5;
static constant: u32 = 5;
fn foo() {}
fn main() {
let local = 5;
println!("stack: {}, global: {}, constant: {}, fn: {}, lib fn: {}",
&local as *const u32,
unsafe { &global as *const u32 },
&constant as *const u32,
unsafe { mem::transmute::<_, *const ()>(foo) },
unsafe { mem::transmute::<_, *const ()>(libc::mprotect) });
}
Before:
stack: 0x3ff15eb9f94, global: 0x6ab488, constant: 0x47db40, fn: 0x4030e0, lib fn: 0x32749547530
stack: 0x3b5d47d80e4, global: 0x6ab488, constant: 0x47db40, fn: 0x4030e0, lib fn: 0x394469a7530
stack: 0x3fe2c4e5564, global: 0x6ab488, constant: 0x47db40, fn: 0x4030e0, lib fn: 0x399734a2530
stack: 0x3e525e0fb24, global: 0x6ab488, constant: 0x47db40, fn: 0x4030e0, lib fn: 0x2f62a810530
stack: 0x3b50fb3eae4, global: 0x6ab488, constant: 0x47db40, fn: 0x4030e0, lib fn: 0x2e590e86530
After:
stack: 0x38cf12c90a4, global: 0x3e2d46b488, constant: 0x3e2d23cf80, fn: 0x3e2d1c2510, lib fn: 0x2617d3b4530
stack: 0x3d733faf474, global: 0x7eb1839488, constant: 0x7eb160af80, fn: 0x7eb1590510, lib fn: 0x32d30c1f530
stack: 0x3bb42212ec4, global: 0x5bbb365488, constant: 0x5bbb136f80, fn: 0x5bbb0bc510, lib fn: 0x3595e6c1530
stack: 0x39f678c1ab4, global: 0x22c4e3c488, constant: 0x22c4c0df80, fn: 0x22c4b93510, lib fn: 0x3835b727530
stack: 0x3afb25bd394, global: 0x493eab2488, constant: 0x493e883f80, fn: 0x493e809510, lib fn: 0x3478d6a7530
This may also be necessary on other platforms, but I can only test on
Linux right now. Note that GDB gained support for debugging position
independent executables in version 7.1 (March 2010).
This leaves the `Share` trait at `std::kinds` via a `#[deprecated]` `pub use`
statement, but the `NoShare` struct is no longer part of `std::kinds::marker`
due to #12660 (the build cannot bootstrap otherwise).
All code referencing the `Share` trait should now reference the `Sync` trait,
and all code referencing the `NoShare` type should now reference the `NoSync`
type. The functionality and meaning of this trait have not changed, only the
naming.
Closes#16281
[breaking-change]
This requires avoiding `quote_...!` for constructing the parts of the
__test module, since that stringifies and reinterns the idents, losing
the special gensym'd nature of them. (#15962.)
meaning `'b outlives 'a`. Syntax currently does nothing but is needed for full
fix to #5763. To use this syntax, the issue_5763_bootstrap feature guard is
required.
Generic extern functions written in Rust have their names mangled, as well as their internal clownshoe __rust_abi functions. This allows e.g. specific monomorphizations of these functions to be used as callbacks.
Closes#12502.
Fixes missing overflow lint for i64 #14269
The `type_overflow` lint, doesn't catch the overflow for `i64` because the overflow happens earlier in the parse phase when the `u64` as biggest possible int gets casted to `i64` , without checking the for
overflows.
We can't lint in the parse phase, so we emit a compiler error, as we do for overflowing `u64`
Perhaps a consistent behaviour would be to emit a parse error for *all* overflowing integer types.
See #14269
The `type_overflow` lint, doesn't catch the overflow for `i64` because
the overflow happens earlier in the parse phase when the `u64` as biggest
possible int gets casted to `i64` , without checking the for overflows.
We can't lint in the parse phase, so a refactoring of the `LitInt` type
was necessary.
The types `LitInt`, `LitUint` and `LitIntUnsuffixed` where merged to one
type `LitInt` which stores it's value as `u64`. An additional parameter was
added which indicate the signedness of the type and the sign of the value.
As discovered in #15460, a particular #[link(kind = "static", ...)] line is not
actually guaranteed to link the library at all. The reason for this is that if
the external library doesn't have any referenced symbols in the object generated
by rustc, the entire library is dropped by the linker.
For dynamic native libraries, this is solved by passing -lfoo for all downstream
compilations unconditionally. For static libraries in rlibs this is solved
because the entire archive is bundled in the rlib. The only situation in which
this was a problem was when a static native library was linked to a rust dynamic
library.
This commit brings the behavior of dylibs in line with rlibs by passing the
--whole-archive flag to the linker when linking native libraries. On OSX, this
uses the -force_load flag. This flag ensures that the entire archive is
considered candidate for being linked into the final dynamic library.
This is a breaking change because if any static library is included twice in the
same compilation unit then the linker will start emitting errors about duplicate
definitions now. The fix for this would involve only statically linking to a
library once.
Closes#15460
[breaking-change]
Using the Show impl for Names created global symbols with names like
`"str\"str\"(1027)"`. This adjusts strings, binaries and vtables to
avoid using that impl.
Using the Show impl for Names created global symbols with names like
`"str\"str\"(1027)"`. This adjusts strings, binaries and vtables to
avoid using that impl.
Fixes#15799.
Before this commit, the LLVM IR of exported items was simply zip-compressed and stored as an object file inside rlib archives. This commit adds a header to this "object" containing a file identifier and a format version number so the compiler can deal with changes in the way bytecode objects are stored within rlibs.
While updating the format of bytecode objects, this commit also worksaround a problem in LLDB which could not handle odd-sized objects within archives before mid-2014.
This is an alternative to upgrading the way rvalues are handled in the
borrow check. Making rvalues handled more like lvalues in the borrow
check caused numerous problems related to double mutable borrows and
rvalue scopes. Rather than come up with more borrow check rules to try
to solve these problems, I decided to just forbid pattern bindings after
`@`. This affected fewer than 10 lines of code in the compiler and
libraries.
This breaks code like:
match x {
y @ z => { ... }
}
match a {
b @ Some(c) => { ... }
}
Change this code to use nested `match` or `let` expressions. For
example:
match x {
y => {
let z = y;
...
}
}
match a {
Some(c) => {
let b = Some(c);
...
}
}
Closes#14587.
[breaking-change]
When generating a unique symbol for things like closures or glue_drop,
we call token::gensym() to create a crate-unique Name. Recently, Name
changed its Show impl so it no longer prints as a number. This caused
symbols like glue_drop:1542 to become glue_drop:"glue_drop"(1542), or in
mangled form, glue_drop.$x22glue_drop$x22$LP$1542$RP$.
When generating a unique symbol for things like closures or glue_drop,
we call token::gensym() to create a crate-unique Name. Recently, Name
changed its Show impl so it no longer prints as a number. This caused
symbols like glue_drop:1542 to become glue_drop:"glue_drop"(1542), or in
mangled form, glue_drop.$x22glue_drop$x22$LP$1542$RP$.
Note: This PR is motivated by an attempt to write an custom syntax extension that tried to use `syntax::fold`, and that could only do so by fixing bugs in it and copying out private functions.
---
Refactored `syntax::fold`
Prior to this, the code there had a few issues:
- Default implementations inconsistenly either had the prefix `noop_` or
not.
- Some default methods where implemented in terms of a public noop function
for user code to call, others where implemented directly on the trait
and did not allow users of the trait to reuse the code.
- Some of the default implementations where private, and thus not reusable
for other implementors.
- There where some bugs where default implemntations called other default
implementations directly, rather than to the underlying Folder, with the
result of some ast nodes never being visted even if the user implemented that
method. (For example, the current Folder never folded struct fields)
This commit solves this situation somewhat radically by making __all__
`fold_...` functions in the module into Folder methods, and implementing
them all in terms of public `noop_...` functions for other implementors to
call out to.
Some public functions had to be renamed to fit the new system, so this is a
breaking change.
---
Also added a few trait implementations to `ast` types
Not included are two required patches:
* LLVM: segmented stack support for DragonFly [1]
* jemalloc: simple configure patches
[1]: http://reviews.llvm.org/D4705
Our implementation of ebml has diverged from the standard in order
to better serve the needs of the compiler, so it doesn't make much
sense to call what we have ebml anyore. Furthermore, our implementation
is pretty crufty, and should eventually be rewritten into a format
that better suits the needs of the compiler. This patch factors out
serialize::ebml into librbml, otherwise known as the Really Bad
Markup Language. This is a stopgap library that shouldn't be used
by end users, and will eventually be replaced by something better.
[breaking-change]
Currently, each time a function is monomorphized, all items within that function are translated. This is unnecessary work because the inner items already get translated when the function declaration is visited by `trans_item`. This patch adds a flag to the `FunctionContext` to prevent translation of items during monomorphization.
Remove the ability of the borrow checker to determine that repeated
dereferences of a Box<T> refer to the same memory object. This will
usually require one of two workarounds:
1) The interior of a Box<T> will sometimes need to be moved / borrowed
into a temporary before moving / borrowing individual derived paths.
2) A `ref x` pattern will have to be replaced with a `box ref x`
pattern.
Fixes#16094.
[breaking-change]
When dealing with HTTP request or responses, many tokens are case-insensitive in the ASCII range but the bytes from the network are not necessarily valid UTF-8.
**[breaking-change]** Rather than adding new very similar traits, this re-uses the `std::ascii::OwnedStrAsciiExt` and `std::ascii::StrAsciiExt` traits, but rename to remove `Str` since that does not apply for bytes.
This PR also makes `std::ascii::ASCII_UPPER_MAP` and `std::ascii::ASCII_LOWER_MAP`, the lookup table all these methods are based on, public. In case there is something else related to ASCII case we haven’t thought of yet, that can be implemented outside of libstd without duplicating the tables.
Although this is a breaking change, I thought this could do without an RFC since the relevant traits are not in the prelude.
r? @alexcrichton
When rustc produces an rlib, it includes the contents of each static library required by the crate. Currently each static library is added individually, by extracting the library with `ar x` and adding the objects to the rlib using `ar r`. Each `ar r` has significant overhead - it appears to scan through the full contents of the rlib before adding the new files. This patch avoids most of the overhead by adding all library objects (and other rlib components) at once using a single `ar r`.
When building `librustc` (on Linux, using GNU ar), this patch gives a 60-80% reduction in linking time, from 90s to 10s one machine I tried and 25s to 8s on another. (Though `librustc` is a bit of a special case - it's a very large crate, so the rlib is large to begin with, and it also relies on a total of 45 static libraries due to the way LLVM is organized.) More reasonable crates such as `libstd` and `libcore` also get a small reduction in linking time (just from adding metadata, bitcode, and object code in one `ar` invocation instead of three), but this is not very noticeable since the time there is small to begin with (around 1s).
I think this is an improvement of the previous warning message, which
- like the comment that I removed implies - is in need of some
improvement.
I've opted to point the user in the right direction w.r.t how to fix the
problem, which I think is good form.
Not being familiar with the repr(...) attribute, I personally had to
check the lint rules myself to figure out what was wrong. Hopefully,
this will save he next person some time and headache.
Signed-off-by: Anton Lofgren <alofgren@op5.com>
the CFG for match statements.
There were two bugs in issue #14684. One was simply that the borrow
check didn't know about the correct CFG for match statements: the
pattern must be a predecessor of the guard. This disallows the bad
behavior if there are bindings in the pattern. But it isn't enough to
prevent the memory safety problem, because of wildcards; thus, this
patch introduces a more restrictive rule, which disallows assignments
and mutable borrows inside guards outright.
I discussed this with Niko and we decided this was the best plan of
action.
This breaks code that performs mutable borrows in pattern guards. Most
commonly, the code looks like this:
impl Foo {
fn f(&mut self, ...) {}
fn g(&mut self, ...) {
match bar {
Baz if self.f(...) => { ... }
_ => { ... }
}
}
}
Change this code to not use a guard. For example:
impl Foo {
fn f(&mut self, ...) {}
fn g(&mut self, ...) {
match bar {
Baz => {
if self.f(...) {
...
} else {
...
}
}
_ => { ... }
}
}
}
Sometimes this can result in code duplication, but often it illustrates
a hidden memory safety problem.
Closes#14684.
[breaking-change]
r? @pnkfelix
Not included are two required patches:
* LLVM: segmented stack support for DragonFly [1]
* jemalloc: simple configure patches
[1]: http://reviews.llvm.org/D4705
Prior to this, the code there had a few issues:
- Default implementations inconsistently either had the prefix `noop_` or
not.
- Some default methods where implemented in terms of a public noop function
for user code to call, others where implemented directly on the trait
and did not allow users of the trait to reuse the code.
- Some of the default implementations where private, and thus not reusable
for other implementors.
- There where some bugs where default implementations called other default
implementations directly, rather than to the underlying Folder, with the
result of some AST nodes never being visited even if the user implemented that
method. (For example, the current Folder never folded struct fields)
This commit solves this situation somewhat radically by making _all_
`fold_...` functions in the module into Folder methods, and implementing
them all in terms of public `noop_...` functions for other implementors to
call out to.
Some public functions had to be renamed to fit the new system, so this is a
breaking change.
[breaking-change]
We previously reexported entire modules, which caused private things to
become reachable and trip the dead code and private items in public API
lints.
Closes#15912
Some minor changes to the compiler to expose this information. Very
inconvenient since struct fields aren't an item. Adds (yet another) table to
metadata.
Closes#15739
We previously reexported entire modules, which caused private things to
become reachable and trip the dead code and private items in public API
lints.
Closes#15912
This commit applies stability attributes to the contents of these modules,
summarized here:
* The `unit` and `bool` modules have become #[unstable] as they are purely meant
for documentation purposes and are candidates for removal.
* The `ty` module has been deprecated, and the inner `Unsafe` type has been
renamed to `UnsafeCell` and moved to the `cell` module. The `marker1` field
has been removed as the compiler now always infers `UnsafeCell` to be
invariant. The `new` method i stable, but the `value` field, `get` and
`unwrap` methods are all unstable.
* The `tuple` module has its name as stable, the naming of the `TupleN` traits
as stable while the methods are all #[unstable]. The other impls in the module
have appropriate stability for the corresponding trait.
* The `arc` module has received the exact same treatment as the `rc` module
previously did.
* The `any` module has its name as stable. The `Any` trait is also stable, with
a new private supertrait which now contains the `get_type_id` method. This is
to make the method a private implementation detail rather than a public-facing
detail.
The two extension traits in the module are marked #[unstable] as they will not
be necessary with DST. The `is` method is #[stable], the as_{mut,ref} methods
have been renamed to downcast_{mut,ref} and are #[unstable].
The extension trait `BoxAny` has been clarified as to why it is unstable as it
will not be necessary with DST.
This is a breaking change because the `marker1` field was removed from the
`UnsafeCell` type. To deal with this change, you can simply delete the field and
only specify the value of the `data` field in static initializers.
[breaking-change]
Rename and gensym the runtime on import, so that users
can't refer to the `native` crate.
This is unlikely to break code, but users should import the "native" crate directly.
[breaking-change]
cc @alexcrichton
Currently we don't emit lifetime end markers when translating the
unwinding code. I omitted that when I added the support for lifetime
intrinsics, because I initially made the mistake of just returning true
in clean_on_unwind(). That caused almost all calls to be translated as
invokes, leading to quite awful results.
To correctly emit the lifetime end markers, we must differentiate
between cleanup that requires unwinding and such cleanup that just wants
to emit code during unwinding.
the CFG for match statements.
There were two bugs in issue #14684. One was simply that the borrow
check didn't know about the correct CFG for match statements: the
pattern must be a predecessor of the guard. This disallows the bad
behavior if there are bindings in the pattern. But it isn't enough to
prevent the memory safety problem, because of wildcards; thus, this
patch introduces a more restrictive rule, which disallows assignments
and mutable borrows inside guards outright.
I discussed this with Niko and we decided this was the best plan of
action.
This breaks code that performs mutable borrows in pattern guards. Most
commonly, the code looks like this:
impl Foo {
fn f(&mut self, ...) {}
fn g(&mut self, ...) {
match bar {
Baz if self.f(...) => { ... }
_ => { ... }
}
}
}
Change this code to not use a guard. For example:
impl Foo {
fn f(&mut self, ...) {}
fn g(&mut self, ...) {
match bar {
Baz => {
if self.f(...) {
...
} else {
...
}
}
_ => { ... }
}
}
}
Sometimes this can result in code duplication, but often it illustrates
a hidden memory safety problem.
Closes#14684.
[breaking-change]
Rename and gensym the runtime on import, so that users
can't refer to the `native` crate.
This is unlikely to break code, but users should import the "native" crate directly.
[breaking-change]
The right hand side of the comparison in these checks are values of type
Option<&Path> which are normalized versions of the left-hand side, so they're
not guaranteed to be byte-for-byte equivalent even though they're the same path.
For this reasons, the command line arguments are promoted to paths for
comparison of equality.
This fixes a bug on windows where if a library was specified with --extern it
would then be picked up twice because it was not considered to have been
previously registered.
Currently we don't emit lifetime end markers when translating the
unwinding code. I omitted that when I added the support for lifetime
intrinsics, because I initially made the mistake of just returning true
in clean_on_unwind(). That caused almost all calls to be translated as
invokes, leading to quite awful results.
To correctly emit the lifetime end markers, we must differentiate
between cleanup that requires unwinding and such cleanup that just wants
to emit code during unwinding.
method calls are involved.
This breaks code like:
impl<T:Copy> Foo for T { ... }
fn take_param<T:Foo>(foo: &T) { ... }
fn main() {
let x = box 3i; // note no `Copy` bound
take_param(&x);
}
Change this code to not contain a type error. For example:
impl<T:Copy> Foo for T { ... }
fn take_param<T:Foo>(foo: &T) { ... }
fn main() {
let x = 3i; // satisfies `Copy` bound
take_param(&x);
}
Closes#15860.
[breaking-change]
r? @alexcrichton
method calls are involved.
This breaks code like:
impl<T:Copy> Foo for T { ... }
fn take_param<T:Foo>(foo: &T) { ... }
fn main() {
let x = box 3i; // note no `Copy` bound
take_param(&x);
}
Change this code to not contain a type error. For example:
impl<T:Copy> Foo for T { ... }
fn take_param<T:Foo>(foo: &T) { ... }
fn main() {
let x = 3i; // satisfies `Copy` bound
take_param(&x);
}
Closes#15860.
[breaking-change]
The right hand side of the comparison in these checks are values of type
Option<&Path> which are normalized versions of the left-hand side, so they're
not guaranteed to be byte-for-byte equivalent even though they're the same path.
For this reasons, the command line arguments are promoted to paths for
comparison of equality.
This fixes a bug on windows where if a library was specified with --extern it
would then be picked up twice because it was not considered to have been
previously registered.
librustc: Stop desugaring `for` expressions and translate them directly.
This makes edge cases in which the `Iterator` trait was not in scope
and/or `Option` or its variants were not in scope work properly.
This breaks code that looks like:
struct MyStruct { ... }
impl MyStruct {
fn next(&mut self) -> Option<int> { ... }
}
for x in MyStruct { ... } { ... }
Change ad-hoc `next` methods like the above to implementations of the
`Iterator` trait. For example:
impl Iterator<int> for MyStruct {
fn next(&mut self) -> Option<int> { ... }
}
Closes#15392.
[breaking-change]
This makes edge cases in which the `Iterator` trait was not in scope
and/or `Option` or its variants were not in scope work properly.
This breaks code that looks like:
struct MyStruct { ... }
impl MyStruct {
fn next(&mut self) -> Option<int> { ... }
}
for x in MyStruct { ... } { ... }
Change ad-hoc `next` methods like the above to implementations of the
`Iterator` trait. For example:
impl Iterator<int> for MyStruct {
fn next(&mut self) -> Option<int> { ... }
}
Closes#15392.
[breaking-change]
This is done entirely in the libraries for functions up to 16 arguments.
A macro is used so that more arguments can be easily added if we need.
Note that I had to adjust the overloaded call algorithm to not try
calling the overloaded call operator if the callee is a built-in
function type, to prevent loops.
Closes#15448.
This eliminates the last vestige of the `~` syntax.
Instead of `~self`, write `self: Box<TypeOfSelf>`; instead of `mut
~self`, write `mut self: Box<TypeOfSelf>`, replacing `TypeOfSelf` with
the self-type parameter as specified in the implementation.
Closes#13885.
[breaking-change]
The allocas used in match expression currently don't get good lifetime
markers, in fact they only get lifetime start markers, because their
lifetimes don't match to cleanup scopes.
While the bindings themselves are bog standard and just need a matching
pair of start and end markers, they might need them twice, once for a
guard clause and once for the match body.
The __llmatch alloca OTOH needs a single lifetime start marker, but
when there's a guard clause, it needs two end markers, because its
lifetime ends either when the guard doesn't match or after the match
body.
With these intrinsics in place, LLVM can now, for example, optimize
code like this:
````rust
enum E {
A1(int),
A2(int),
A3(int),
A4(int),
}
pub fn variants(x: E) {
match x {
A1(m) => bar(&m),
A2(m) => bar(&m),
A3(m) => bar(&m),
A4(m) => bar(&m),
}
}
````
To a single call to bar, using only a single stack slot. It still fails
to eliminate some of checks.
````gas
.Ltmp5:
.cfi_def_cfa_offset 16
movb (%rdi), %al
testb %al, %al
je .LBB3_5
movzbl %al, %eax
cmpl $1, %eax
je .LBB3_5
cmpl $2, %eax
.LBB3_5:
movq 8(%rdi), %rax
movq %rax, (%rsp)
leaq (%rsp), %rdi
callq _ZN3bar20hcb7a0d8be8e17e37daaE@PLT
popq %rax
retq
````
Refs #15665
The allocas used in match expression currently don't get good lifetime
markers, in fact they only get lifetime start markers, because their
lifetimes don't match to cleanup scopes.
While the bindings themselves are bog standard and just need a matching
pair of start and end markers, they might need them twice, once for a
guard clause and once for the match body.
The __llmatch alloca OTOH needs a single lifetime start marker, but
when there's a guard clause, it needs two end markers, because its
lifetime ends either when the guard doesn't match or after the match
body.
With these intrinsics in place, LLVM can now, for example, optimize
code like this:
````rust
enum E {
A1(int),
A2(int),
A3(int),
A4(int),
}
pub fn variants(x: E) {
match x {
A1(m) => bar(&m),
A2(m) => bar(&m),
A3(m) => bar(&m),
A4(m) => bar(&m),
}
}
````
To a single call to bar, using only a single stack slot. It still fails
to eliminate some of checks.
````gas
.Ltmp5:
.cfi_def_cfa_offset 16
movb (%rdi), %al
testb %al, %al
je .LBB3_5
movzbl %al, %eax
cmpl $1, %eax
je .LBB3_5
cmpl $2, %eax
.LBB3_5:
movq 8(%rdi), %rax
movq %rax, (%rsp)
leaq (%rsp), %rdi
callq _ZN3bar20hcb7a0d8be8e17e37daaE@PLT
popq %rax
retq
````
Lifetime intrinsics help to reduce stack usage, because LLVM can apply
stack coloring to reuse the stack slots of dead allocas for new ones.
For example these functions now both use the same amount of stack, while
previous `bar()` used five times as much as `foo()`:
````rust
fn foo() {
println("{}", 5);
}
fn bar() {
println("{}", 5);
println("{}", 5);
println("{}", 5);
println("{}", 5);
println("{}", 5);
}
````
On top of that, LLVM can also optimize out certain operations when it
knows that memory is dead after a certain point. For example, it can
sometimes remove the zeroing used to cancel the drop glue. This is
possible when the glue drop itself was already removed because the
zeroing dominated the drop glue call. For example in:
````rust
pub fn bar(x: (Box<int>, int)) -> (Box<int>, int) {
x
}
````
With optimizations, this currently results in:
````llvm
define void @_ZN3bar20h330fa42547df8179niaE({ i64*, i64 }* noalias nocapture nonnull sret, { i64*, i64 }* noalias nocapture nonnull) unnamed_addr #0 {
"_ZN29_$LP$Box$LT$int$GT$$C$int$RP$39glue_drop.$x22glue_drop$x22$LP$1347$RP$17h88cf42702e5a322aE.exit":
%2 = bitcast { i64*, i64 }* %1 to i8*
%3 = bitcast { i64*, i64 }* %0 to i8*
tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* %2, i64 16, i32 8, i1 false)
tail call void @llvm.memset.p0i8.i64(i8* %2, i8 0, i64 16, i32 8, i1 false)
ret void
}
````
But with lifetime intrinsics we get:
````llvm
define void @_ZN3bar20h330fa42547df8179niaE({ i64*, i64 }* noalias nocapture nonnull sret, { i64*, i64 }* noalias nocapture nonnull) unnamed_addr #0 {
"_ZN29_$LP$Box$LT$int$GT$$C$int$RP$39glue_drop.$x22glue_drop$x22$LP$1347$RP$17h88cf42702e5a322aE.exit":
%2 = bitcast { i64*, i64 }* %1 to i8*
%3 = bitcast { i64*, i64 }* %0 to i8*
tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* %2, i64 16, i32 8, i1 false)
tail call void @llvm.lifetime.end(i64 16, i8* %2)
ret void
}
````
Fixes#15665
Lifetime intrinsics help to reduce stack usage, because LLVM can apply
stack coloring to reuse the stack slots of dead allocas for new ones.
For example these functions now both use the same amount of stack, while
previous `bar()` used five times as much as `foo()`:
````rust
fn foo() {
println("{}", 5);
}
fn bar() {
println("{}", 5);
println("{}", 5);
println("{}", 5);
println("{}", 5);
println("{}", 5);
}
````
On top of that, LLVM can also optimize out certain operations when it
knows that memory is dead after a certain point. For example, it can
sometimes remove the zeroing used to cancel the drop glue. This is
possible when the glue drop itself was already removed because the
zeroing dominated the drop glue call. For example in:
````rust
pub fn bar(x: (Box<int>, int)) -> (Box<int>, int) {
x
}
````
With optimizations, this currently results in:
````llvm
define void @_ZN3bar20h330fa42547df8179niaE({ i64*, i64 }* noalias nocapture nonnull sret, { i64*, i64 }* noalias nocapture nonnull) unnamed_addr #0 {
"_ZN29_$LP$Box$LT$int$GT$$C$int$RP$39glue_drop.$x22glue_drop$x22$LP$1347$RP$17h88cf42702e5a322aE.exit":
%2 = bitcast { i64*, i64 }* %1 to i8*
%3 = bitcast { i64*, i64 }* %0 to i8*
tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* %2, i64 16, i32 8, i1 false)
tail call void @llvm.memset.p0i8.i64(i8* %2, i8 0, i64 16, i32 8, i1 false)
ret void
}
````
But with lifetime intrinsics we get:
````llvm
define void @_ZN3bar20h330fa42547df8179niaE({ i64*, i64 }* noalias nocapture nonnull sret, { i64*, i64 }* noalias nocapture nonnull) unnamed_addr #0 {
"_ZN29_$LP$Box$LT$int$GT$$C$int$RP$39glue_drop.$x22glue_drop$x22$LP$1347$RP$17h88cf42702e5a322aE.exit":
%2 = bitcast { i64*, i64 }* %1 to i8*
%3 = bitcast { i64*, i64 }* %0 to i8*
tail call void @llvm.memcpy.p0i8.p0i8.i64(i8* %3, i8* %2, i64 16, i32 8, i1 false)
tail call void @llvm.lifetime.end(i64 16, i8* %2)
ret void
}
````
Fixes#15665
In f1ad425199, I changed the handling
of macros, to prevent macro invocations from occurring in fully expanded
source. Instead, I added a side table. It contained only the
spans of the macros, because this was the only information required
in order to make macro export work.
However, librustdoc was also affected by this change, since it
extracts macro information in a similar way. As a result of the earlier
change, exported macros were no longer documented.
In order to repair this, I've adjusted the side table to contain whole
items, rather than just the spans.
`call_visit_glue` is only ever called from trans_intrinsic, and the
block won't be unreachable there. Also, the comment doesn't make sense
anymore. When the code was introduced in 38fee9526a the function was
also responsible for the cleanup glue, which is no longer the case.
While we're at it, also fixed the debug message to output the right
function name.
This implements RFC 39. Omitted lifetimes in return values will now be
inferred to more useful defaults, and an error is reported if a lifetime
in a return type is omitted and one of the two lifetime elision rules
does not specify what it should be.
This primarily breaks two uncommon code patterns. The first is this:
unsafe fn get_foo_out_of_thin_air() -> &Foo {
...
}
This should be changed to:
unsafe fn get_foo_out_of_thin_air() -> &'static Foo {
...
}
The second pattern that needs to be changed is this:
enum MaybeBorrowed<'a> {
Borrowed(&'a str),
Owned(String),
}
fn foo() -> MaybeBorrowed {
Owned(format!("hello world"))
}
Change code like this to:
enum MaybeBorrowed<'a> {
Borrowed(&'a str),
Owned(String),
}
fn foo() -> MaybeBorrowed<'static> {
Owned(format!("hello world"))
}
Closes#15552.
[breaking-change]
r? @nick29581
This is accomplished by rewriting static expressions into equivalent patterns.
This way, patterns referencing static variables can both participate
in exhaustiveness analysis as well as be compiled down into the appropriate
branch of the decision trees that match expressions are codegened to.
Fixes#6533.
Fixes#13626.
Fixes#13731.
Fixes#14576.
Fixes#15393.
This implements RFC 39. Omitted lifetimes in return values will now be
inferred to more useful defaults, and an error is reported if a lifetime
in a return type is omitted and one of the two lifetime elision rules
does not specify what it should be.
This primarily breaks two uncommon code patterns. The first is this:
unsafe fn get_foo_out_of_thin_air() -> &Foo {
...
}
This should be changed to:
unsafe fn get_foo_out_of_thin_air() -> &'static Foo {
...
}
The second pattern that needs to be changed is this:
enum MaybeBorrowed<'a> {
Borrowed(&'a str),
Owned(String),
}
fn foo() -> MaybeBorrowed {
Owned(format!("hello world"))
}
Change code like this to:
enum MaybeBorrowed<'a> {
Borrowed(&'a str),
Owned(String),
}
fn foo() -> MaybeBorrowed<'static> {
Owned(format!("hello world"))
}
Closes#15552.
[breaking-change]
The first is to require that `#[crate_name]` and `--crate-name` always match (if both are specified). The second is to fix parallel compilation in cargo by mixing in `-C extra-filename` into the temporary outputs of the compiler.
When invoking the compiler in parallel, the intermediate output of the object
files and bytecode can stomp over one another if two crates with the same name
are being compiled.
The output file is already being disambiguated with `-C extra-filename`, so this
commit alters the naming of the temporary files to also mix in the extra
filename to ensure that file names don't clash.
This is accomplished by rewriting static expressions into equivalent patterns.
This way, patterns referencing static variables can both participate
in exhaustiveness analysis as well as be compiled down into the appropriate
branch of the decision trees that match expressions are codegened to.
Fixes#6533.
Fixes#13626.
Fixes#13731.
Fixes#14576.
Fixes#15393.
