"self" is always passed as an opaque box, so there's no point in using
the concrete self type when translating the argument. All it does it
causing the value to be casted back to an opaque box right away.
The commit f9a5453 is meant to be a temporary hold-over. Whether or not there is added a way for the compiler to "implicitly borrow" stack closures in this way, there should be a codegen optimization that prevents having to traverse possibly-very-many function pointers to find the function you ultimately wanted to call. I tried to separate out the changes so this particular commit could be straight-up reverted if auto-borrowing happens in the future.
r? @nikomatsakis
@graydon suggested that once closures not be part of the language for 1.0, but that they might be hidden behind a -Z compile flag as an "experimental feature" in case people decide they need them.
Regardless of whether ```-Z once-fns``` is set, this PR will parse the ```once``` keyword and will prevent closures labelled with it from being called more than once. It will also permit moving out of captured vars in heap closures, just to let the runtime writers stop using ```Cell``` sooner. Setting ```-Z once-fns``` only toggles whether the move-out-from-capture privilege is also given for stack closures.
r? @nikomatsakis
This fixes a bunch of default method bugs and restructures how vtable resolutions are represented.
(It also adds a depth counter to llvm::type_to_str as a hacky work around for our circular llvm types. This is related in the sense that I needed to do it to make debug tracing not cause rustc to crash after running out of stack space.)
Instead of determining paths from the path tag, we iterate through
modules' children recursively in the metadata. This will allow for
lazy external module resolution.
This removes the `namegen` thunk that was in `common.rs`. I also take the opportunity to refactor a few uses where we had a `str -> ident -> str` chain that seemed somewhat redundant to me.
Also cleans up some warnings that made their way in already.
Mostly just low-haning fruit, i.e. function arguments that were @ even
though & would work just as well.
Reduces librustc.so size by 200k when compiling without -O, by 100k when
compiling with -O.
Mostly just low-haning fruit, i.e. function arguments that were @ even
though & would work just as well.
Reduces librustc.so size by 200k when compiling without -O, by 100k when
compiling with -O.
r? @brson (or @graydon if available) rustpkg/api.rs provides functions intended for package scripts to call.
It will probably need more functionality added to it later, but this is
a start.
Added a test case checking that a package script can use the API.
Closes#6401
This allows macros to both be conditionally defined, and expand
to items with #[cfg]'s.
This seems to have a performance improvement, e.g. for `std`:
```
# Before
time: 1.660 s expansion
time: 0.125 s configuration
# After
time: 0.080 s configuration 1
time: 1.127 s expansion
time: 0.132 s configuration 2
```
And for `extra`:
```
# Before
time: 0.593 s expansion
time: 0.062 s configuration
# After
time: 0.047 s configuration 1
time: 0.147 s expansion
time: 0.058 s configuration 2
```
(This seems a little peculiar, but it is possibly because the expansion AST traversal is very slow, so removing as much as possible as early as possible has big benefits.)
This PR contains no real code changes. Just some documentation additions in the form of comments and some internal reordering of functions within debuginfo.rs.
rustpkg/api.rs provides functions intended for package scripts to call.
It will probably need more functionality added to it later, but this is
a start.
Added a test case checking that a package script can use the API.
Closes#6401
Fix#7322.
I started out with a band-aid approach to special-case the duplicate module error using `is_duplicate_module`, but thought this would be better in the long term.
The "first definition of ..." error string reported by add_child() looks
different from similar messages reported by other functions. Fix this.
Signed-off-by: Ramkumar Ramachandra <artagnon@gmail.com>
add_child() is responsible for reporting errors about type, value, and
module duplicate definitions. Although it checks for all three, it uses
namespace_to_str() to convert a Namespace value into a string before
printing an error like:
error: duplicate definition of type `foo`
^^^^
note: first definition of type foo here:
^^^^
Unfortunately, this string can only be one of "type" or
"value" (corresponding to TypeNS and ValueNS respectively), and it
reports duplicate modules as duplicate types.
To alleviate the problem, define a special NamespaceError enum to define
more specialized errors, and use it instead of attempting to reuse the
Namespace enum.
Reported-by: Corey Richardson <corey@octayn.net>
Signed-off-by: Ramkumar Ramachandra <artagnon@gmail.com>
This adds a `#[no_drop_flag]` attribute. This attribute tells the compiler to omit the drop flag from the struct, if it has a destructor. When the destructor is run, instead of setting the drop flag, it instead zeroes-out the struct. This means the destructor can run multiple times and therefore it is up to the developer to use it safely.
The primary usage case for this is smart-pointer types like `Rc<T>` as the extra flag caused the struct to be 1 word larger because of alignment.
This closes#7271 and #7138
This sets the `get_tydesc()` return type correctly and removes the intrinsic module. See #3730, #3475.
Update: this now also removes the unused shape fields in tydescs.
This adds both `static mut` items and `static mut` foreign items. This involved changing far less code than I thought it was going to, but the tests seem to pass and the variables seem functional.
I'm more than willing to write more tests, so suggestions are welcome!
Closes#553
the `test/run-pass/class-trait-bounded-param.rs` test was xfailed and
written in an ancient dialect of Rust so I've just removed it
this also removes `to_vec` from DList because it's provided by
`std::iter::to_vec`
an Iterator implementation is added for OptVec but some transitional
internal iterator methods are still left
To achieve this, the following changes were made:
* Move TyDesc, TyVisitor and Opaque to std::unstable::intrinsics
* Convert TyDesc, TyVisitor and Opaque to lang items instead of specially
handling the intrinsics module
* Removed TypeDesc, FreeGlue and get_type_desc() from sys
Fixes#3475.
This fixes part of #3730, but not all.
Also changes the TyDesc struct to be equivalent with the generated
code, with the hope that the above issue may one day be closed for good,
i.e. that the TyDesc type can completely be specified in the Rust
sources and not be generated.
I removed the `static-method-test.rs` test because it was heavily based
on `BaseIter` and there are plenty of other more complex uses of static
methods anyway.
The only really tricky change is that a long chain of ifs, and elses
was turned into a single if, and a match in astencode.rs. Some methods
can only be called in certain cases, and so have to come after the if.
This is another big refactoring of `trans` though this is unlikely to have much of an
impact on code size or speed.
The major change here is the implementation of a `Type` struct which is the new
home for all your LLVM `TypeRef` needs. It's a simple wrapper struct, with static
methods for constructing types, then regular methods for
manipulating/interrogating them. The purpose of this is mostly to make the code
surrounding them somewhat more ideomatic. A line like: `T_ptr(T_ptr(T_i8()))` is
now `Type::i8().ptr_to().ptr_to()`,which is much more like regular Rust code.
There are a variety of smaller changes here and there:
* Remove address spaces. At least it doesn't generate them, I haven't spent much
time looking for related code.
* Use a macro for declaring the LLVM intrinsics, makes it look much nicer.
* Make the type for a string slice actually create a named `str_slice` type in LLVM,
this makes reading the appropriate code much easier.
* Change the way struct and enum type names are generated. This just means
that a struct like `struct Foo { a: int }` now produces the IR
`%struct.Foo = type { i64 }`, which is much easier to read. Similarly, other structs
are a bit tighter to make it easier to read.
--- --- ---
This PR did get away from me a little, as I occasionally got distracted or as I fixed
up problems with unrelated code that were stopping me from continuing. One major
thing is that this PR contains the work from #7168, since that would have conflicted
with this and it was broken anyway. Sorry for bundling it like this.
Fixes#3670 and #7063
--- --- ---
EDIT: This no longer removes the llvm insn stats.
This makes the handling of atomic operations more generic, which
does impose a specific naming convention for the intrinsics, but
that seems ok with me, rather than having an individual case for
each name.
It also adds the intrinsics to the the intrinsics file.
The changes in these commits improve the IR codegen by removing unnecessary copies for certain function call arguments and stopping to allocate return values for functions returning nil. They reduce compile times by about 10% in total.
By using "void" instead of "{}" as the LLVM type for nil, we can avoid
the alloca/store/load sequence for the return value, resulting in less
and simpler IR code.
This reduces compile times by about 10%.
The removed test for issue #2611 is well covered by the `std::iterator`
module itself.
This adds the `count` method to `IteratorUtil` to replace `EqIter`.
This fixes the large number of problems that prevented cross crate
methods from ever working. It also fixes a couple lingering bugs with
polymorphic default methods and cleans up some of the code paths.
Closes#4102. Closes#4103.
r? nikomatsakis
This fixes the large number of problems that prevented cross crate
methods from ever working. It also fixes a couple lingering bugs with
polymorphic default methods and cleans up some of the code paths.
Closes#4102. Closes#4103.
Currently, by-copy function arguments are always stored into a scratch
datum, which serves two purposes. First, it is required to be able to
have a temporary cleanup, in case that the call fails before the callee
actually takes ownership of the value. Second, if the argument is to be
passed by reference, the copy is required, so that the function doesn't
get a reference to the original value.
But in case that the datum does not need a drop glue call and it is
passed by value, there's no need to perform the extra copy.
This commit fixes rustc's debug info generation and turns debug-info tests back on.
The old generator used to write out LLVM metadata directly, however it seems that debug metadata format is not stable and keeps changing from release to release. So I wrapped LLVM's official debug info API - the DIBuilder class, and now rustc will use that.
One bit of old functionality that still doesn't work, is debug info for function arguments. Someone more familiar with the compiler guts will need to look into that.
Also, unfortunately, debug info is still won't work on Windows,- due to a LLVM bug (http://llvm.org/bugs/show_bug.cgi?id=16249).
Resolves issues #5836, #5848, #6814
Currently, cleanup blocks are only reused when there are nested scopes, the
child scope's cleanup block will terminate with a jump to the parent
scope's cleanup block. But within a single scope, adding or revoking
any cleanup will force a fresh cleanup block. This means quadratic
growth with the number of allocations in a scope, because each
allocation needs a landing pad.
Instead of forcing a fresh cleanup block, we can keep a list chained
cleanup blocks that form a prefix of the currently required cleanups.
That way, the next cleanup block only has to handle newly added
cleanups. And by keeping the whole list instead of just the latest
block, we can also handle revocations more efficiently, by only
dropping those blocks that are no longer required, instead of all of
them.
Reduces the size of librustc by about 5% and the time required to build
it by about 10%.
Currently, cleanup blocks are only reused when there are nested scopes, the
child scope's cleanup block will terminate with a jump to the parent
scope's cleanup block. But within a single scope, adding or revoking
any cleanup will force a fresh cleanup block. This means quadratic
growth with the number of allocations in a scope, because each
allocation needs a landing pad.
Instead of forcing a fresh cleanup block, we can keep a list chained
cleanup blocks that form a prefix of the currently required cleanups.
That way, the next cleanup block only has to handle newly added
cleanups. And by keeping the whole list instead of just the latest
block, we can also handle revocations more efficiently, by only
dropping those blocks that are no longer required, instead of all of
them.
Reduces the size of librustc by about 5% and the time required to build
it by about 10%.
Moves all the remaining functions that could reasonably be methods to be methods, except for some FFI ones (which I believe @erickt is working on, possibly) and `each_split_within`, since I'm not really sure the details of it (I believe @kimundi wrote the current implementation, so maybe he could convert it to an external iterator method on `StrSlice`, e.g. `word_wrap_iter(&self) -> WordWrapIterator<'self>`, where `WordWrapIterator` impls `Iterator<&'self str>`. It probably won't be too hard, since it's already a state machine.)
This also cleans up the comparison impls for the string types, except I'm not sure how the lang items `eq_str` and `eq_str_uniq` need to be handled, so they (`eq_slice` and `eq`) remain stand-alone functions.
This is a bad default, because the binaries will point at an absolute
path regardless of where they are moved. This opens up a security issue
for packages, because they will attempt to load libraries from a path
that's often owned by a regular user.
Every Rust binary is currently flagged by Debian, Fedora and Arch lint
checkers as having dangerous rpaths. They don't meet the requirements to
be placed in the repositories without manually stripping this from each
binary.
The relative rpath is still enough to keep the binaries working until
they are moved relative to the crates they're linked against.
http://wiki.debian.org/RpathIssuehttps://fedoraproject.org/wiki/Packaging:Guidelines#Beware_of_Rpath
This is a bad default, because the binaries will point at an absolute
path regardless of where they are moved. This opens up a security issue
for packages, because they will attempt to load libraries from a path
that's often owned by a regular user.
Every Rust binary is currently flagged by Debian, Fedora and Arch lint
checkers as having dangerous rpaths. They don't meet the requirements to
be placed in the repositories without manually stripping this from each
binary.
The relative rpath is still enough to keep the binaries working until
they are moved relative to the crates they're linked against.
http://wiki.debian.org/RpathIssuehttps://fedoraproject.org/wiki/Packaging:Guidelines#Beware_of_Rpath
Remove all the explicit @mut-fields from CrateContext, though many
fields are still @-ptrs.
This required changing every single function call that explicitly
took a @CrateContext, so I took advantage and changed as many as I
could get away with to &-ptrs or &mut ptrs.
Currently, when calling glue functions, we cast the function to match
the argument type. This interacts very badly with LLVM and breaks
inlining of the glue code.
It's more efficient to use a unified function type for the glue
functions and always cast the function argument instead of the function.
The resulting code for rustc is about 13% faster (measured up to and
including the "trans" pass) and the resulting librustc is about 5%
smaller.
Currently, when calling glue functions, we cast the function to match
the argument type. This interacts very badly with LLVM and breaks
inlining of the glue code.
It's more efficient to use a unified function type for the glue
functions and always cast the function argument instead of the function.
The resulting code for rustc is about 13% faster (measured up to and
including the "trans" pass) and the resulting librustc is about 5%
smaller.
r? @graydon Automate more tests described in the commands.txt file,
and add infrastructure for running them. Right now, tests shell
out to call rustpkg. This is not ideal.
Goes part of the way towards addressing #5683
This un-reverts the reverts of the rusti commits made awhile back. These were reverted for an LLVM failure in rustpkg. I believe that this is not a problem with these commits, but rather that rustc is being used in parallel for rustpkg tests (in-process). This is not working yet (almost! see #7011), so I serialized all the tests to run one after another.
@brson, I'm mainly just guessing as to the cause of the LLVM failures in rustpkg tests. I'm confident that running tests in parallel is more likely to be the problem than those commits I made.
Additionally, this fixes two recently reported issues with rusti.
The lookups for these items in external crates currently cause repeated
decoding of the EBML metadata, which is pretty slow. Adding caches to
avoid the repeated decoding reduces the time required for the type
checking of librustc by about 25%.
The lookups for these items in external crates currently cause repeated
decoding of the EBML metadata, which is pretty slow. Adding caches to
avoid the repeated decoding reduces the time required for the type
checking of librustc by about 25%.
This fixes the strange random crashes in compile-fail tests.
This reverts commit 96cd61ad03.
Conflicts:
src/librustc/driver/driver.rs
src/libstd/str.rs
src/libsyntax/ext/quote.rs
This almost removes the StringRef wrapper, since all strings are
Equiv-alent now. Removes a lot of `/* bad */ copy *`'s, and converts
several things to be &'static str (the lint table and the intrinsics
table).
There are many instances of .to_managed(), unfortunately.
There are now only half-a-dozen or so functions left `std::str` that should be methods.
Highlights:
- `.substr` was removed, since most of the uses of it in the code base were actually incorrect (it had a weird mixing of a byte index and a unicode character count), adding `.slice_chars` if one wants to handle characters, and the normal `.slice` method to handle bytes.
- Code duplication between the two impls for `connect` and `concat` was removed via a new `Str` trait, that is purely designed to allow an explicit -> `&str` conversion (`.as_slice()`)
- Deconfuse the 5 different functions for converting to `[u8]` (3 of which had actually incorrect documentation: implying that they didn't have the null terminator), into 3: `as_bytes` (all strings), `as_bytes_with_null` (`&'static str`, `@str` and `~str`) and `as_bytes_with_null_consume` (`~str`). None of these allocate, unlike the old versions.
(cc @thestinger)
The confusing mixture of byte index and character count meant that every
use of .substr was incorrect; replaced by slice_chars which only uses
character indices. The old behaviour of `.substr(start, n)` can be emulated
via `.slice_from(start).slice_chars(0, n)`.
This was a lot more painful than just changing `x.each` to `x.iter().advance` . I ran into my old friend #5898 and had to add underscores to some method names as a temporary workaround.
The borrow checker also had other ideas because rvalues aren't handled very well yet so temporary variables had to be added. However, storing the temporary in a variable led to dynamic `@mut` failures, so those had to be wrapped in blocks except where the scope ends immediately.
Anyway, the ugliness will be fixed as the compiler issues are fixed and this change will amount to `for x.each |x|` becoming `for x.iter |x|` and making all the iterator adaptors available.
I dropped the run-pass tests for `old_iter` because there's not much point in fixing a module that's on the way out in the next week or so.
This is a reopening of #6570, and almost fixes#6511.
Note that this doesn't actually enable building a threadsafe LLVM, because that will require an LLVM rebuild which will be bundled with the upgrades in #6713.
What this does do, however, is removes all thread-unsafe usage of LLVM from the compiler.
Minimally fixes#7017, we were overwriting the result and thus ignoring attributes before the last.
csearch::get_item_attrs and decoder::get_item_attrs should probably also be changed to each_item_attrs using the for protocol, but that's just a minor performance/style issue.
This removes some unnecessary allocations in the lexer, the typechecker and the metadata decoder. Reduces the time spent in the parsing and typechecking passes by about 10% for me.
Handle more characters that appear in types, most notably <>): were
missing. Also the new scheme takes care that no two different input
strings result in the same mangled string, which was not the case before.
Fixes#6921
Handle more characters that appear in types, most notably <>): were
missing. Also the new scheme takes care that no two different input
strings result in the same mangled string, which was not the case before.
Fixes#6921
For types that are passed by value, we can't just cast the value to a
pointer, but have to use an alloca and copy the value there. This
handling is already present for all other arguments, but was missing
for "self".
Fixes#6682, #4850 and #4878
This commit won't be quite as useful until I implement RUST_PATH and
until we change `extern mod` to take a general string instead of
an identifier (#5682 and #6407).
With that said, now if you're using rustpkg and a program contains:
extern mod foo;
rustpkg will attempt to search for `foo`, so that you don't have to
provide a -L directory explicitly. In addition, rustpkg will
actually try to build and install `foo`, unless it's already
installed (specifically, I tested that `extern mod extra;` would
not cause it to try to find source for `extra` and compile it
again).
This is as per #5681.
Incidentally, I changed some driver code to infer the link name
from the crate link_meta attributes. If that change isn't ok, say
something. Also, I changed the addl_lib_search_paths field in the
session options to be an @mut ~[Path] so that it can be modified
after expansion but before later phases.
The `callee_id` in `ast::expr` in only used in a couple expression variants. This moves the `callee_id` into those branches to make it more clear when its should be used.
Also, it fixes a bug in a std::run test when there is a symlink in the path rust where was checked out.
This fixes#6745, which itself relates to #4202. Slightly ham-fisted -- feel particularly funny about using the typeck phase to gather the base -> impl mapping, and the separate code paths for traits vs. "real" bases feels like it could be avoided -- but it seems to work.
As always, open to suggestions if there's a better way to accomplish what I'm trying to do.
@catamorphism r?
This is a better pipeline, both faster-running and produces faster code.
For some reason the `mergefunc` pass screws over resolve. I have no idea why though.
Closes#5090 by using the excellent new generic deriving code
Promotes the unreachable code attribute to a lint attribute (instead of always being a warning)
Fixes some edge cases when creating hashmaps/hashsets and also when consuming them. (fixes#5998)
Fix a laundry list of warnings involving unused imports that glutted
up compilation output. There are more, but there seems to be some
false positives (where 'remedy' appears to break the build), but this
particular set of fixes seems safe.
Fix a laundry list of warnings involving unused imports that glutted
up compilation output. There are more, but there seems to be some
false positives (where 'remedy' appears to break the build), but this
particular set of fixes seems safe.
This refactors pass handling to use the argument names, so it can be used
in a similar manner to `opt`. This may be slightly less efficient than the
previous version, but it is much easier to maintain.
It also adds in the ability to specify a custom pipeline on the command
line, this overrides the normal passes, however. This should completely
close#2396.
Most of the relevant information can be found in the commit messages.
r? @brson - I just wanted to make sure the make changes aren't completely bogus
This would close#2400, #6517, and #6489 (although a run through incoming-full on linux would have to confirm the latter two)
This refactors pass handling to use the argument names, so it can be used
in a similar manner to `opt`. This may be slightly less efficient than the
previous version, but it is much easier to maintain.
It also adds in the ability to specify a custom pipeline on the command
line, this overrides the normal passes, however. This should completely
close#2396.
Move the computation of what data is moved out of `liveness` and into `borrowck`. The resulting code is cleaner, since before we had a split distribution of responsibilities, and also this avoids having multiple implementations of the dataflow code. Liveness is still used to report warnings about useless writes. This will go away when we get the control-flow graph code landed (working on that).
Also adds borrow checker documentation.
Fixes#4384.
Required to support once fns and to properly fix closures (#2202).
First step to generalize our treatment of moves somewhat as well.
This way a cross-compiled rustc's answer to host_triple() is correct. The return
value of host_triple() reflects the actual host triple that the compiler was
build for, not the triple the compiler is being built on
Refactor the optimization passes to explicitly use the passes. This commit
just re-implements the same passes as were already being run.
It also adds an option (behind `-Z`) to run the LLVM lint pass on the
unoptimized IR.
Fix for #6575. In the trans phase, rustc emits code for a function parameter that goes completely unused in the event the return type of the function in question happens to be an immediate.
This patch modifies rustc & parts of rustrt to ensure that the vestigial parameter is no longer present in compiled code.
borrow checker and generalize what moves are allowed. Fixes a nasty
bug or two in the pattern move checking code. Unifies dataflow code
used for initialization and other things. First step towards
once fns. Everybody wins.
Fixes#4384. Fixes#4715. cc once fns (#2202), optimizing local moves (#5016).
The compiler guarantees that there are no other references to a unique pointer when it's passed by-value to a function.
The existence of the header and annihilator don't matter since it's not relevant to the call:
> For a call to the parent function, dependencies between memory references from before or after the call and from those during the call are “irrelevant” to the noalias keyword for the arguments and return value used in that call.
@graydon's tracing garbage collector stores the metadata outside of the boxes, so that won't be a problem. I'm unsure if updating the header while inside a function where it's marked as `noalias` would be a problem anyway since you never actually read or write to the header.
@nikomatsakis: r?
mentioned in #2625.
This change makes the module more oriented around
Process values instead of having to deal with process ids
directly.
Apart from issues mentioned in #2625, other changes include:
- Changing the naming to be more consistent - Process/process
is now used instead of a mixture of Program/program and
Process/process.
- More docs/tests.
Some io/scheduler related issues remain (mentioned in #2625).
