Adds an `assume` intrinsic that gets translated to llvm.assume. It is
used on a boolean expression and allows the optimizer to assume that
the expression is true.
This implements #18051.
This reverts commit a0ec902e23 "Avoid
unnecessary temporary on assignments".
Leaving out the temporary for the functions return value can lead to a
situation that conflicts with rust's aliasing rules.
Given this:
````rust
fn func(f: &mut Foo) -> Foo { /* ... */ }
fn bar() {
let mut foo = Foo { /* ... */ };
foo = func(&mut foo);
}
````
We effectively get two mutable references to the same variable `foo` at
the same time. One for the parameter `f`, and one for the hidden
out-pointer. So we can't just `trans_into` the destination directly, but
must use `trans` to get a new temporary slot from which the result can
be copied.
Spring cleaning is here! In the Fall! This commit removes quite a large amount
of deprecated functionality from the standard libraries. I tried to ensure that
only old deprecated functionality was removed.
This is removing lots and lots of deprecated features, so this is a breaking
change. Please consult the deprecation messages of the deleted code to see how
to migrate code forward if it still needs migration.
[breaking-change]
When translating the unboxing shim, account for the fact that the shim translation has already performed the necessary unboxing of input types and values when forwarding to the shimmed function. This prevents ICEing or generating incorrect code.
Closes#16739
When translating the unboxing shim, account for the fact that the shim
translation has already performed the necessary unboxing of input
types and values when forwarding to the shimmed function. This
prevents ICEing or generating incorrect code.
Closes#16739
- Unify the representations of `cat_upvar` and `cat_copied_upvar`
- In `link_reborrowed_region`, account for the ability of upvars to
change their mutability due to later processing. A map of recursive
region links we may want to establish in the future is maintained,
with the links being established when the kind of the borrow is
adjusted.
- When categorizing upvars, add an explicit deref that represents the
closure environment pointer for closures that do not take the
environment by value. The region for the implicit pointer is an
anonymous free region type introduced for this purpose. This
creates the necessary constraint to prevent unsound reborrows from
the environment.
- Add a note to categorizations to make it easier to tell when extra
dereferences have been inserted by an upvar without having to
perform deep pattern matching.
- Adjust borrowck to deal with the changes. Where `cat_upvar` and
`cat_copied_upvar` were previously treated differently, they are
now both treated roughly like local variables within the closure
body, as the explicit derefs now ensure proper behavior. However,
error diagnostics had to be changed to explicitly look through the
extra dereferences to avoid producing confusing messages about
references not present in the source code.
Closes issue #17403. Remaining work:
- The error diagnostics that result from failed region inference are
pretty inscrutible and should be improved.
Code like the following is now rejected:
let mut x = 0u;
let f = || &mut x;
let y = f();
let z = f(); // multiple mutable references to the same location
This also breaks code that uses a similar construction even if it does
not go on to violate aliasability semantics. Such code will need to
be reworked in some way, such as by using a capture-by-value closure
type.
[breaking-change]
Adds an `assume` intrinsic that gets translated to llvm.assume. It is
used on a boolean expression and allows the optimizer to assume that
the expression is true.
This implements #18051.
LLVM generates wrong code (which may be an instance of compile-time UB) when
faced with types that take lots of memory - bigger than the address space.
Make using such types a trans error. While trans errors are bad, overbig
types are expected to be very rare.
Use the integer sizes LLVM uses, rather than having random projections
laying around. Sizes are u64, Alignments are u32, C_*int is target-dependent
but 64-bit is fine (the int -> C_int conversion is non-precision-losing,
but it can be preceded by `as int` conversions which are, so it is
somewhat ugly. However, being able to suffix a `u` to properly infer
integer types is nice).
RFC 344 proposes a set of naming conventions for lints. This commit
renames existing lints to follow the conventions.
Use the following sed script to bring your code up to date:
```
s/unnecessary_typecast/unused_typecasts/g
s/unsigned_negate/unsigned_negation/g
s/type_limits/unused_comparisons/g
s/type_overflow/overflowing_literals/g
s/ctypes/improper_ctypes/g
s/owned_heap_memory/box_pointers/g
s/unused_attribute/unused_attributes/g
s/path_statement/path_statements/g
s/unused_must_use/unused_must_use/g
s/unused_result/unused_results/g
s/non_uppercase_statics/non_upper_case_globals/g
s/unnecessary_parens/unused_parens/g
s/unnecessary_import_braces/unused_import_braces/g
s/unused_unsafe/unused_unsafe/g
s/unsafe_block/unsafe_blocks/g
s/unused_mut/unused_mut/g
s/unnecessary_allocation/unused_allocation/g
s/missing_doc/missing_docs/g
s/unused_imports/unused_imports/g
s/unused_extern_crate/unused_extern_crates/g
s/unnecessary_qualification/unused_qualifications/g
s/unrecognized_lint/unknown_lints/g
s/unused_variable/unused_variables/g
s/dead_assignment/unused_assignments/g
s/unknown_crate_type/unknown_crate_types/g
s/variant_size_difference/variant_size_differences/g
s/transmute_fat_ptr/fat_ptr_transmutes/g
```
Closes#16545Closes#17932
Due to deprecation, this is a:
[breaking-change]
Function arguments are (hopefully!) the last places where allocas don't
get proper markers for the end of their lifetimes. This means that this
code using 64 bytes of stack for the function arguments:
````rust
std::io::println("1");
std::io::println("2");
std::io::println("3");
std::io::println("4");
````
But with the proper lifetime markers, the slots can be reused, and
the arguments only need 16 bytes of stack.
Doing so would incur deeply nested expansion of the tree with no useful
side effects. This is problematic for "wide" data types such as structs
with dozens of fields but where only a few are actually being matched or bound.
Most notably, matching a fixed slice would use a number of stack frames that
grows with the number of elements in the slice.
Fixes#17877.
Implement multidispatch and conditional dispatch. Because we do not attempt to preserve crate concatenation, this is a backwards compatible change. This is not yet fully integrated into method dispatch, so "UFCS"-style wrappers must be used to take advantage of the new features (see the run-pass tests).
cc #17307 (multidispatch)
cc #5527 (trait reform -- conditional dispatch)
Because we no longer preserve crate concatenability, this deviates slightly from what was specified in the RFC. The motivation for this change is described in [this blog post](http://smallcultfollowing.com/babysteps/blog/2014/09/30/multi-and-conditional-dispatch-in-traits/). I will post an amendment to the RFC in due course but do not anticipate great controversy on this point -- particularly as the RFCs more important features (e.g., conditional dispatch) just don't work without the change.
This change is an implementation of [RFC 69][rfc] which adds a third kind of
global to the language, `const`. This global is most similar to what the old
`static` was, and if you're unsure about what to use then you should use a
`const`.
The semantics of these three kinds of globals are:
* A `const` does not represent a memory location, but only a value. Constants
are translated as rvalues, which means that their values are directly inlined
at usage location (similar to a #define in C/C++). Constant values are, well,
constant, and can not be modified. Any "modification" is actually a
modification to a local value on the stack rather than the actual constant
itself.
Almost all values are allowed inside constants, whether they have interior
mutability or not. There are a few minor restrictions listed in the RFC, but
they should in general not come up too often.
* A `static` now always represents a memory location (unconditionally). Any
references to the same `static` are actually a reference to the same memory
location. Only values whose types ascribe to `Sync` are allowed in a `static`.
This restriction is in place because many threads may access a `static`
concurrently. Lifting this restriction (and allowing unsafe access) is a
future extension not implemented at this time.
* A `static mut` continues to always represent a memory location. All references
to a `static mut` continue to be `unsafe`.
This is a large breaking change, and many programs will need to be updated
accordingly. A summary of the breaking changes is:
* Statics may no longer be used in patterns. Statics now always represent a
memory location, which can sometimes be modified. To fix code, repurpose the
matched-on-`static` to a `const`.
static FOO: uint = 4;
match n {
FOO => { /* ... */ }
_ => { /* ... */ }
}
change this code to:
const FOO: uint = 4;
match n {
FOO => { /* ... */ }
_ => { /* ... */ }
}
* Statics may no longer refer to other statics by value. Due to statics being
able to change at runtime, allowing them to reference one another could
possibly lead to confusing semantics. If you are in this situation, use a
constant initializer instead. Note, however, that statics may reference other
statics by address, however.
* Statics may no longer be used in constant expressions, such as array lengths.
This is due to the same restrictions as listed above. Use a `const` instead.
[breaking-change]
[rfc]: https://github.com/rust-lang/rfcs/pull/246
While booleans are represented as i1 in SSA values, LLVM expects them
to be stored/loaded as i8 values. Using i1 as we do now works, but
kills some optimizations, so we should switch to i8, just like we do
everywhere else.
Fixes#16959.
While booleans are represented as i1 in SSA values, LLVM expects them
to be stored/loaded as i8 values. Using i1 as we do now works, but
kills some optimizations, so we should switch to i8, just like we do
everywhere else.
Fixes#16959.
Apart from making the build system determine the LLDB version, this PR also fixes an issue with enums in LLDB pretty printers. In order for GDB's pretty printers to know for sure if a field of some value is an enum discriminant, I had rustc mark discriminant fields with the `artificial` DWARF tag. This worked out nicely for GDB but it turns out that one can't access artificial fields from LLDB. So I changed the debuginfo representation so that enum discriminants are marked by the special field name `RUST$ENUM$DISR` instead, which works in both cases.
The PR does not activate the LLDB test suite yet.
LLDB doesn't allow for reading 'artifical' fields (fields that are generated by the compiler). So do not mark, slice fields, enum discriminants, and GcBox value fields as artificial.
This is a quick fix that prevents an ICE by mimicing the visitor
glue for boxed closures and bare functions. Ideally, the `TyVisitor`
interface will be improved in the future to allow representing
more information about unboxed closures such as Fn/FnMut/FnOnce
status, capture mode, and captured free variable types and offsets.
Closes issue #17737
This began as an attempt to fix an ICE in borrowck (issue #17655), but the rabbit hole went pretty deep. I ended up plumbing support for capture-by-reference unboxed closures all the way into trans.
Closes issue #17655.
Store references to the freevars instead of copies when constructing
the environment and insert an additional load when reading them from
the environment.
Fixes that unit-like structs cannot be used if they are re-exported and used in another crate. (ICE)
The relevant changes are in `rustc::metadata::{decoder, encoder}` and `rustc::middle::ty`.
A test case is included.
The problem is that the expressoin `UnitStruct` is an `ExprPath` to an `DefFn`, which is of expr kind `RvalueDatumExpr`, but for unit-struct ctors the expr kind should be `RvalueDpsExpr`. I fixed this (in a I guess clean way) by introducing `CtorFn` in the metadata and including a `is_ctor` flag in `DefFn`.
Modify ast::ExprMatch to include a new value of type ast::MatchSource,
making it easy to tell whether the match was written literally or
produced via desugaring. This allows us to customize error messages
appropriately.
Fixes that unit-like structs cannot be used if they are reexported and
used in another crate. The compiler fails with an ICE, because unit-like
structs are exported as DefFn and the expression `UnitStruct` is
interpreted as function pointer instead of a call to the constructor.
To resolve this ambiguity tuple-like struct constructors are now exported
as CtorFn. When `rustc::metadata::decoder` finds a CtorFn it sets a new
flag `is_ctor` in DefFn to true.
Relevant changes are in `rustc::metadata::{encoder, decoder}` and in
`rustc::middle::ty`.
Closes#12660 and #16973.
This is the bare minimum to stop using split stacks on Windows, fixing https://github.com/rust-lang/rust/issues/13259 and #14742, by turning on stack probes for all functions and disabling compiler and runtime support for split stacks on Windows.
It does not restore the out-of-stack error message, which requires more runtime work.
This includes a test that the Windows TCB is no longer being clobbered, but the out-of-stack test itself is pretty weak, only testing that the program exits abnormally, not that it isn't writing to bogus memory, so I haven't truly verified that this is providing the safety we claim.
A more complete solution is in https://github.com/rust-lang/rust/pull/16388, which has some unresolved issues yet.
cc @Zoxc @klutzy @vadimcn
This PR makes rustc emit debug locations for *all* call and invoke statements in LLVM IR, if they are contained within a function that debuginfo is enabled for. This is important because LLVM does not handle the case where a function body containing debuginfo is inlined into another function with debuginfo, but the inlined call statement does not have a debug location. In this case, LLVM will not know where (in terms of source code coordinates) the function was inlined to and we end up with some statements still linked to the source locations in there original, non-inlined function without any indication that they are indeed an inline-copy. Later, when generating DWARF from the IR, LLVM will interpret this as corrupt IR and abort.
Unfortunately, the undesirable case described above can still occur when using LTO. If there is a crate compiled without debuginfo calling into a crate compiled with debuginfo, we again end up with the conditions triggering the error. This is why some LTO tests still fail with the dreaded assertion, if the standard library was built with debuginfo enabled. That is, `RUSTFLAGS_STAGE2=-g make rustc-stage2` will succeed but `RUSTFLAGS_STAGE2=-g make check` will still fail after this PR has been merged. I will open a separate issue for this problem.
This commit makes rustc emit debug locations for all call
and invoke statements in LLVM IR, if they are contained
within a function that debuginfo is enabled for. This is
important because LLVM does not handle the case where a
function body containing debuginfo is inlined into another
function with debuginfo, but the inlined call statement
does not have a debug location. In this case, LLVM will
not know where (in terms of source code coordinates) the
function was inlined to and we end up with some statements
still linked to the source locations in there original,
non-inlined function without any indication that they are
indeed an inline-copy. Later, when generating DWARF from
the IR, LLVM will interpret this as corrupt IR and abort.
Unfortunately, the undesirable case described above can
still occur when using LTO. If there is a crate compiled
without debuginfo calling into a crate compiled with
debuginfo, we again end up with the conditions triggering
the error. This is why some LTO tests still fail with the
dreaded assertion, if the standard library was built with
debuginfo enabled.
That is, `RUSTFLAGS_STAGE2=-g make rustc-stage2` will
succeed but `RUSTFLAGS_STAGE2=-g make check` will still
fail after this commit has been merged. This is a problem
that has to be dealt with separately.
Fixes#17201Fixes#15816Fixes#15156
Change to resolve and update compiler and libs for uses.
[breaking-change]
Enum variants are now in both the value and type namespaces. This means that
if you have a variant with the same name as a type in scope in a module, you
will get a name clash and thus an error. The solution is to either rename the
type or the variant.
The implementation essentially desugars during type collection and AST
type conversion time into the parameter scheme we have now. Only fully
qualified names--e.g. `<T as Foo>::Bar`--are supported.
