I believe this patch incorporates all expected syntax changes from extern
function reform (#3678). You can now write things like:
extern "<abi>" fn foo(s: S) -> T { ... }
extern "<abi>" mod { ... }
extern "<abi>" fn(S) -> T
The ABI for foreign functions is taken from this syntax (rather than from an
annotation). We support the full ABI specification I described on the mailing
list. The correct ABI is chosen based on the target architecture.
Calls by pointer to C functions are not yet supported, and the Rust type of
crust fns is still *u8.
Before it wouldn't warn about unused imports in the list if something in the list was used. These commits fix that case, add a test, and remove all unused imports in lists of imports throughout the compiler.
re bug that @nikomatsakis was hitting: when you define a `static` (old: `const`) containing a `&` or `&[]` expression, it will create temporaries (the underlying pointee) by creating a throwaway symbol for each temporary, each with _global_ linkage, and each named `"const"`. LLVM will helpfully rename multiple copies of this throwaway symbol to `"const1"` and `"const2"` and so forth in the _same_ library. But if you have _2 libraries_ -- say, libcore and librustc -- that both do this, the dynamic linker (at least on linux) will happily do horrible things like make the slice in one library point to the bytes of the vector from the other library. This is obviously a recipe for much hilarity and head-scratching.
The solution is to change the linkage to something else, internal or (in the case of this patch) _private_.
It will require a snapshot to integrate this into stage0 and thereby fix the problem / unblock patches that were hitting this in stage1.
Hey folks,
This patch series does some work on the json decoder, specifically with auto decoding of enums. Previously, we would take this code:
```
enum A {
B,
C(~str, uint)
}
```
and would encode a value of this enum to either `["B", []]` or `["C", ["D", 123]]`. I've changed this to `"B"` or `["C", "D", 123]`. This matches the style of the O'Caml json library [json-wheel](http://mjambon.com/json-wheel.html). I've added tests to make sure all this work.
In order to make this change, I added passing a `&[&str]` vec to `Decode::emit_enum_variant` so the json decoder can convert the name of a variant into it's position. I also changed the impl of `Encodable` for `Option<T>` to have the right upper casing.
I also did some work on the parser, which allows for `fn foo<T: ::cmp::Eq>() { ... }` statements (#5572), fixed the pretty printer properly expanding `debug!("...")` expressions, and removed `ast::expr_vstore_fixed`, which doesn't appear to be used anymore.
the types. Initially I thought it would be necessary to thread this data
through not only the AST but the types themselves, but then I remembered that
the pretty printer only cares about the AST. Regardless, I have elected to
leave the changes to the types intact since they will eventually be needed. I
left a few FIXMEs where it didn't seem worth finishing up since the code wasn't
crucial yet.
Various FIXME comments added around to denote copies which when removed cause
the compiler to segfault at some point before stage2. None of these copies
should even be necessary.
and from typeck, which is verboten. We are supposed to write inference results
into the FnCtxt and then these get copied over in writeback. Add assertions
that no inference by-products are added to this table.
Fixes#3888Fixes#4036Fixes#4492
r? @nikomatsakis The typechecker previously passed around a boolean return flag to
indicate whether it saw something with type _|_ (that is, something
it knows at compile-time will definitely diverge) and also had some
manual checks for the `ty_err` pseudo-type that represents a previous
type error. This was because the typing rules implemented by the
typechecker didn't properly propagate _|_ and ty_err. I fixed it.
This also required changing expected error messages in a few tests,
as now we're printing out fewer derived errors -- in fact, at this
point we should print out no derived errors, so report any that
you see (ones that include "[type error]") as bugs.
The typechecker previously passed around a boolean return flag to
indicate whether it saw something with type _|_ (that is, something
it knows at compile-time will definitely diverge) and also had some
manual checks for the `ty_err` pseudo-type that represents a previous
type error. This was because the typing rules implemented by the
typechecker didn't properly propagate _|_ and ty_err. I fixed it.
This also required changing expected error messages in a few tests,
as now we're printing out fewer derived errors -- in fact, at this
point we should print out no derived errors, so report any that
you see (ones that include "[type error]") as bugs.
Removes a lot of instances of `/*bad*/ copy` throughout libsyntax/librustc. On the plus side, this shaves about 2s off of the runtime when compiling `librustc` with optimizations.
Ideally I would have run a profiler to figure out which copies are the most critical to remove, but in reality there was a liberal amount of `git grep`s along with some spot checking and removing the easy ones.
For bootstrapping purposes, this commit does not remove all uses of
the keyword "pure" -- doing so would cause the compiler to no longer
bootstrap due to some syntax extensions ("deriving" in particular).
Instead, it makes the compiler ignore "pure". Post-snapshot, we can
remove "pure" from the language.
There are quite a few (~100) borrow check errors that were essentially
all the result of mutable fields or partial borrows of `@mut`. Per
discussions with Niko I think we want to allow partial borrows of
`@mut` but detect obvious footguns. We should also improve the error
message when `@mut` is erroneously reborrowed.
This is done in two steps:
First, we make foreign functions not consider modes at all. This is because previously ++ mode was the only way to pass structs to foreign functions and so forth. We also add a lint mode warning if you use `&&` mode in a foreign function, since the semantics of that change (it used to pass a pointer to the C function, now it doesn't).
Then, we remove by value and make it equivalent to `+` mode. At the same time, we stop parsing `-` mode and convert all uses of it to `+` mode (it was already being parsed to `+` mode anyhow).
This obsoletes pull request #5298.
r? @brson
adjusting a few foreign functions that were declared with by-ref
mode. This also allows us to remove by-val mode in the near future.
With copy mode, though, we have to be careful because Rust will implicitly pass
somethings by pointer but this may not be the C ABI rules. For example, rust
will pass a struct Foo as a Foo*. So I added some code into the adapters to
fix this (though the C ABI rules may put the pointer back, oh well).
This patch also includes a lint mode for the use of by-ref mode
in foreign functions as the semantics of this have changed.
Continuing #5140
For the sake of getting this merged I've disabled debuginfo tests on mac (where running gdb needs root). Please feel free to follow up with further improvements.
r? @graydon
This removes `log` from the language. Because we can't quite implement it as a syntax extension (probably need globals at the least) it simply renames the keyword to `__log` and hides it behind macros.
After this the only way to log is with `debug!`, `info!`, etc. I figure that if there is demand for `log!` we can add it back later.
I am not sure that we ever agreed on this course of action, though I *think* there is consensus that `log` shouldn't be a statement.
These changes make const translation use adjustments (autodereference, autoreference, bare-fn-to-closure), like normal code does, replacing some ad-hoc logic that wasn't always right.
As a convenient side-effect, explicit dereference (both of pointers and of newtypes) is also supported in const expressions.
There is also a “bonus fix” for a bug in the pretty-printer exposed by one of the added tests.
Struct and enum representations have some complicatedness that's no longer needed. Now that everything's in one place and has access to anything we'd want to know about the type, flatten some of that out. Slight changes to representations in some cases.
The only thing we really lose is that C-like enums with one variant and a
non-zero discriminant now take up space, but I do not think this is a
common usage. As previously noted, that was mostly there for
transitional compatibility with the pre-adt.rs codebase.
Out goes the extra layer of struct wrapping; the destructedness flag is
added to the end of the struct. This means that, if the struct
previously had alignment padding at the end, the flag will live there
instead of increasing the struct size.
Changes the ad-hoc closure adjustment into using adjustment info instead
of being separately driven from types, and likewise for autoderef.
Also takes care of autoref (the cases we should be seeing in consts,
at least, since we can't be doing method calls which would need the
ref-to-vec mode), which didn't quite work right previously.
However, "dereference" of a newtype isn't handled yet....
Also converts const cast-from-enum, because it used the same routine to
get the discriminant as what's renovated to construct the enums.
Also fixes ICE on struct-like variants as consts, and provides a slightly
less bad ICE for functional-update-like struct expressions in consts.
Note that in the ByValue case (which can't happen? yet?) we're still
effectively bitcasting, I think. So this change adds a way to assert
that that's safe.
Note also, for future reference, that LLVM's instcombine pass will turn
a bitcast into a GEP(0, 0, ...) if possible.
This change remains separate from the addition of adt.rs, even though
it's necessary for compatibility with pre-trans::adt representation,
to serve as an example of a change to the representation logic.
Later changes on this branch adapt the rest of rustc::middle::trans
to use this module instead of scattered hard-coded knowledge of
representations; a few of them also have improvements or cleanup for
adt.rs (and many added comments) that weren't drastic enough to justify
changing history to move them into this commit.
Work towards #4846.
- Institute new region defaults where all omitted regions get a fresh lifetime.
- Require explicit region names except in functions.
- Fix a bug in region parameterization inference. I've been putting this off because it will not be important when we remove RP inference in favor of explicit declarations, but then it was blocking this patch.
r? @pcwalton
Before:
````
test.rs:3:21: 3:30 error: expected constant integer for repeat count but found variable
test.rs:3 let a = ~[0, ..n]; //~ ERROR expected constant integer for repeat count but found variable
^~~~~~~~~
````
After:
````
test.rs:3:27: 3:28 error: expected constant integer for repeat count but found variable
test.rs:3 let a = ~[0, ..n]; //~ ERROR expected constant integer for repeat count but found variable
^
````
Before:
````
test.rs:3:21: 3:30 error: expected constant integer for repeat count but found variable
test.rs:3 let a = ~[0, ..n]; //~ ERROR expected constant integer for repeat count but found variable
^~~~~~~~~
````
After:
````
test.rs:3:27: 3:28 error: expected constant integer for repeat count but found variable
test.rs:3 let a = ~[0, ..n]; //~ ERROR expected constant integer for repeat count but found variable
^
````
The basic idea is that we add a new kind of adjustment, AutoAddEnv, that pads
an extern fn into a closure by adding the extra NULL word. Then there are a few
misc changes in trans to get the LLVM types to match up.
Fixes#4808.
This removes all but 6 uses of `drop {}` from the entire codebase. Removing any of the remaining uses causes various non-trivial bugs; I'll start reporting them once this gets merged.
Major changes are:
- replace ~[ty_param] with Generics structure, which includes
both OptVec<TyParam> and OptVec<Lifetime>;
- the use of syntax::opt_vec to avoid allocation for empty lists;
cc #4846
r? @graydon
Major changes are:
- replace ~[ty_param] with Generics structure, which includes
both OptVec<TyParam> and OptVec<Lifetime>;
- the use of syntax::opt_vec to avoid allocation for empty lists;
cc #4846
r?
#3406
Pretty straightforward. I'm using opaque pointers instead trying to get trans and core to agree on the types of the main function and crate map. One oddity is that this required changing the order of the `-lrustrt` argument to the linker in order to resolve `upcall_new_stack`. Linkers are mysterious.
r?
After this patch, macros declared in a module, function, or block can only be used inside of that module, function or block, with the exception of modules declared with the #[macro_escape] attribute; these modules allow macros to escape, and can be used as a limited macro export mechanism.
This pull request also includes miscellaneous comments, lots of new test cases, a few renamings, and a few as-yet-unused data definitions for hygiene.
Note on `struct_elt`: the comment is wrong, it actually dereferences the nth element of LLVM struct type if it is a pointer. That's why `T_ptr` is removed in `callee.rs`.
It appears that using deriving_eq/auto_encode on ASTs bumps up against the "gee this looks like infinite unfolding" limit of 10 in monomorphization. Increasing it to 30 seems to solve this problem for me....
Also, commenting and a few renames.
Consts of such enums are aligned correctly, so we could either misalign
them to match the type_of, or fix the type_of. The latter seems like a
better idea.
`simplify_type` was bogus, as there was no way for it to handle enums
properly. It was also slow, because it created many Rust types at runtime. In
general creating Rust types during trans is a source of slowness, and I'd like
to avoid doing it as much as possible. (It is probably not possible to
eliminate it entirely, due to `subst`, but we should get rid of as much of it
as we can.) So this patch replaces `simplify_type` with `sizing_type_of`,
which creates a size-equivalent LLVM type directly without going through a
Rust type first.
Because this is causing an ICE in Servo, I'm rubber stamping it.
Issue #3869
review? @nikomatsakis
Convert all uses of vec::slice to vec::view Issue #3869
Rename const_view to const_slice
Renamed mut_view to mut_slice
Fix windows build error. `buf` is borrowed by the call to
`as_mut_buf()` and so we must invoke `slice()` outside of that
call.
This patch finishes removing inner vector mutability from the vast majority of the compiler. Exceptions:
* core::dvec: ideally this entire type will be able to be replaced by `~[]`, but Niko asked me to hold off on removing Dvecs until he makes some fixes to borrowed pointers.
* liveness: liveness.rs is an impenetrable neutron star of deprecated semantics.
* compile-fail: I'm not sure if a lot of these tests are testing inner mutability or mutability in general. I figure that RIMOVing this folder should wait until this syntax is removed from the parser.
I also took this chance to remove many of the inner-mutability-related functions from core::vec, or as many uses of those functions as possible where still necessary. consume_mut and append_mut have been axed. cast_to_mut and cast_from_mut are still needed in a few places.
If an enum is isomorphic to unit, there's no need to use any bits to
represent it. The only obvious reason this wasn't the case was because
the enum could be C-like and have a user-specified discriminant -- but
that value is constant, so it doesn't need to be stored.
This change means that all newtype-like enums have the same size (and
layout) as their underlying type, which might be a useful property to
have, at least in terms of making programs' low-level behavior less
surprising.
The number of operands of the LLVM node initializing the array
underlying a const vector isn't always the array length -- if the
array is of a sufficiently primitive type and all the elements' values
are known (or something like that), LLVM uses a specialized Constant
subclass that stores the data packed, and thus has no operands. Oops.
But, because llsize_of now gives us a ConstantInt, we can just fix
mozilla/rust#3169 and this all goes away.
This is wasted space if the const is just an enum, but optimizing that
case without breaking everything else is an issue that can be addressed
separately.
- Moved ToStr implementation of integers to int-template.rs.
- Marked the `str()` function as deprecated.
- Forwarded all conversion functions to `core::num::to_str_common()`
and `core::num::from_str_common()`.
- Fixed most places in the codebase where `to_str()` is being used.
- Added int-template to_str and from_str overflow tests.
LinearMap is quite a bit faster, and is fully owned/sendable without
requiring copies. The older std::map also doesn't use explicit self and
relies on mutable fields.
Changes:
- Refactor move mode computation
- Removes move mode arguments, unary move, capture clauses
(though they still parse for backwards compatibility)
- Simplify how moves are handled in trans
- Fix a number of illegal copies that cropped up
- Workaround for bug involving def-ids in params
(see details below)
Future work (I'll open bugs for these...):
- Improve error messages for moves that are due
to bindings
- Add support for moving owned content like a.b.c
to borrow check, test in trans (but I think it'll
"just work")
- Proper fix for def-ids in params
Def ids in params:
Move captures into a map instead of recomputing.
This is a workaround for a larger bug having to do with the def-ids associated
with ty_params, which are not always properly preserved when inlining. I am
not sure of my preferred fix for the larger bug yet. This current fix removes
the only code in trans that I know of which relies on ty_param def-ids, but
feels fragile.
