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).
Simple patch series to fix up all the warnings a rustc compile is giving at the moment. It also fixes a NOTE in `to_bytes.rs` to remove the `to_bytes::iter_bytes_<N>` functions.
The way we deal with unreachable expressions in trans is pretty ad hoc,
but this at least doesn't make it worse, and eliminates the LLVM
assertion failure reported in #5741.
Fixes https://github.com/mozilla/rust/issues/6578 by merging the 3 different ways to build an AST into a single `AstBuilder` trait, creating a more uniform and briefer interface.
Also, converts the `ext_ctxt` trait-object to be a plain struct, as well as renaming it to `ExtCtxt`.
Seems to make expansion slightly faster for the normal case (e.g. `libcore` and `libstd`), but slower for `librustc` (slightly) and `libsyntax` (0.3s -> 0.8s! I'm investigating this, but I'd prefer this patch to land relatively quickly.).
`git blame` suggests maybe @graydon or @erickt are familiar with this area of the code. r?
This adds a lint mode for detecting unnecessary allocations on the heap. This isn't super fancy, currently it only has two rules
1. For a function's arguments, if you allocate a `[~|@]str` literal, when the type of the argument is a `&str`, emit a warning.
2. For the same case, emit warnings for boxed vectors when slices are required.
After adding the lint, I rampaged through the libraries and removed all the unnecessary allocations I could find.
Currently, trait_info is a hashmap that allows a quick lookup of all
methods contained in a given trait, but we actually only use it to
lookup traits that contain a given method. Adjusting the map to support
the lookup we actually need gives a nice speed boost, reducing the time
required for the resolution step for librustc from ~2.6s to ~1.0s on my
box.
With this, the build is almost 100% warning free.
One more can be fixed after the next snapshot, and there's one other that I filed an issue about already.
It uses the private field of TCB head to store stack limit. I tested on my Raspberry PI. A simple hello world program ran without any problem. However, for a more complex program, it segfaulted as #6231.
Closes#2647
This way it's much easier to add lints throughout compilation correctly, and
functions on impls can alter the way lints are emitted. This involved pretty much rewriting how lints are emitted. Beforehand, only items could alter the lint settings, so whenever a lint was added it had to be associated with whatever item id it was coming from. I removed this (possibly questionably) in favor of just specifying a span and a message when adding a lint. When lint checking comes around, it looks at all the lints and sees which node with attributes best encloses it and uses that level of linting. This means that all consumer code doesn't have to deal with what item things came from (especially because functions on impls aren't items). More details of this can be found in the code (and comments).
As a bonus, I managed to greatly simplify emission of lints in resolve.rs about unused imports. Now instead of it manually tracking what the lint level is, it's all moved over into the lint module (as is to be expected).
r? @nikomatsakis Impls can implement either zero or one traits; this has been true
more or less since we removed classes. So I got rid of the comments
saying "we should support multiple traits" and changed the code to
make it clear that we don't. This is just cleanup, and doesn't break
any existing tests.
Impls can implement either zero or one traits; this has been true
more or less since we removed classes. So I got rid of the comments
saying "we should support multiple traits" and changed the code to
make it clear that we don't. This is just cleanup, and doesn't break
any existing tests.
`str::from_slice` and `vec::from_slice` are changed to `str::to_owned` and `vec::to_owned`. (#6389)
Replace `at_vec::from_owned` and `at_vec::from_slice` with `at_vec::to_managed_consume` and `at_vec::to_managed`.
Replace all instances of #[auto_*code] with the appropriate #[deriving] attribute
and remove the majority of the actual auto_* code, leaving stubs to refer the user to
the new syntax.
Also, moves the useful contents of auto_encode.rs to more appropriate spots: tests and comments to deriving/encodable.rs, and the ExtCtxtMethods trait to build.rs (unused so far, but the method syntax might be nicer than using the mk_* fns in many instances).
Replace all instances of #[auto_*code] with the appropriate #[deriving] attribute
and remove the majority of the actual code, leaving stubs to refer the user to
the new syntax.
fail!() used to require owned strings but can handle static strings
now. Also, it can pass its arguments to fmt!() on its own, no need for
the caller to call fmt!() itself.
When trying to import nonexistent items from existing modules, specify that
that is what happened, rather than just reporting "unresolved name".
Ideally the error would be reported on the span of the import... but I do not see a way to get a span there. Help appreciated 😄
This pull request adds 4 atomic intrinsics to the compiler, in preparation for #5042.
* `atomic_load(src: &int) -> int` performs an atomic sequentially consistent load.
* `atomic_load_acq(src: &int) -> int` performs an atomic acquiring load.
* `atomic_store(dst: &mut int, val: int)` performs an atomic sequentially consistent store.
* `atomic_store_rel(dst: &mut int, val: int)` performs an atomic releasing store.
For more information about the whole acquire/release thing: http://llvm.org/docs/Atomics.html
r?
Every unresolved import is reported. An additional error message isn't useful
and obscures (imo) the real errors: I need to take it into account when
looking at the error count.
The default versions (atomic_load and atomic_store) are sequentially consistent.
The atomic_load_acq intrinsic acquires as described in [1].
The atomic_store_rel intrinsic releases as described in [1].
[1]: http://llvm.org/docs/Atomics.html
r? @nikomatsakis In #6319, several people mentioned they ran into a "computing
fictitious type" ICE in trans. This turns out to be because some
of my recent changes to typeck::check::_match resulted in type errors
getting reported with ty_err as the expected type, which meant the errors
were suppressed, and typechecking incorrectly succeeded (since the errors
weren't recorded).
Changed the error messages in these cases not to use an expected type at all,
rather, printing out a string describing the type that was expected (which is
what the code originally did). The result is a bit repetitive and the
proliferation of error-reporting functions in typeck::infer is a bit annoying,
but I thought it was important to fix this now; more cleanup can happen later.
In #6319, several people mentioned they ran into a "computing
fictitious type" ICE in trans. This turns out to be because some
of my recent changes to typeck::check::_match resulted in type errors
getting reported with ty_err as the expected type, which meant the errors
were suppressed, and typechecking incorrectly succeeded (since the errors
weren't recorded).
Changed the error messages in these cases not to use an expected type at all,
rather, printing out a string describing the type that was expected (which is
what the code originally did). The result is a bit repetitive and the
proliferation of error-reporting functions in typeck::infer is a bit annoying,
but I thought it was important to fix this now; more cleanup can happen later.
Fixes#6378
Don't pass the binary name to the LLVMRustExecuteJIT closure, otherwise it will leak memory; the binary name doesn't seem to be needed, anyhow.
**Caveat**: With the current commit, this check only works for `match`s, the checks (incorrectly) do not run for patterns in `let`s, and invalid/unsafe code compiles.
I don't know how to fix this, I experimented with some things to try to make let patterns and match patterns run on the same code (since this would presumably fix many of the other unsoundness issues of let-patterns, e.g. #6225), but I don't understand enough of the code. (I think I heard someone talking about a fix for `let` being in progress?)
Fixes#6344 and #6341.
This allows macros to create tests and benchmarks.
This is possibly unsound (I've got no idea, but it seemed to work), and being able to programmatically generate benchmarks to compare several implementations of similar algorithms is nice.
This fixes the issue described in #4202.
From what I understood of the code, when we reexport a trait in a submodule using e.g. "pub use foo::SomeTrait", we were not previously making an effort to reexport the static methods on that trait.
I'm new to the Rust code base (and the Rust language itself) so my approach may not be kosher, but this patch works by changing the encoder to include the static methods associated with traits.
I couldn't see any tests for this area of the code, so I didn't really have any examples to go by. If tests are needed, I'm happy to work through that if I can get some assistance to do so.
There may be a more efficient implementation of `core::util::swap_ptr`. The issue mentioned using `move_val_init`, but I couldn't figure out what that did, so I just used `copy_memory` a few times instead.
I'm not exactly the best at reading LLVM generated by rust, but this does appear to be optimized away just as expected (when possible).
Closes#6183.
The first commit changes the compiler's method of treating a `for` loop, and all the remaining commits are just dealing with the fallout.
The biggest fallout was the `IterBytes` trait, although it's really a whole lot nicer now because all of the `iter_bytes_XX` methods are just and-ed together. Sadly there was a huge amount of stuff that's `cfg(stage0)` gated, but whoever lands the next snapshot is going to have a lot of fun deleting all this code!
&str can be turned into @~str on demand, using to_owned(), so for
strings, we can create a specialized interner that accepts &str for
intern() and find() but stores and returns @~str.
This improves error reporting for the following class of imports:
```rust
use foo::bar;
```
Where foo, the topmost module, is unresolved. It now results in:
```text
/tmp/foo.rs:1:4: 1:7 error: unresolved import. perhapsyou forgot an 'extern mod foo'?
/tmp/foo.rs:1 use foo::bar;
^~~
/tmp/foo.rs:1:4: 1:12 error: failed to resolve import: foo::bar
/tmp/foo.rs:1 use foo::bar;
^~~~~~~~
error: failed to resolve imports
error: aborting due to 3 previous errors
```
This is the first of a series of changes I plan on making to unresolved name error messages.
Use a bitset to represent built-in bounds. There are several places in the language where only builtin bounds (aka kinds) will be accepted, e.g. on closures, destructor type parameters perhaps, and on trait types.
r? @brson
Fix#6355 and #6272---we were not giving the correct index to the derefs that occur as part of the rooting process, resulting in extra copies and generally bogus behavior. Haven't quite produced the right test for this, but I thought I'd push the fix in the meantime. Test will follow shortly.
r? @graydon
Adds an `uninit` intrinsic.
It's just an empty function, so llvm optimizes it down to nothing.
I changed all of the `init` intrinsic usages to `uninit` where it seemed appropriate to.
its own type. Use a bitset to represent built-in bounds. There
are several places in the language where only builtin bounds (aka kinds)
will be accepted, e.g. on closures, destructor type parameters perhaps,
and on trait types.
&str can be turned into @~str on demand, using to_owned(), so for
strings, we can create a specialized interner that accepts &str for
intern() and find() but stores and returns @~str.
Hi there,
Really enjoying Rust. Noticed a few typos so I searched around for a few more--here's some fixes.
Ran `make check` and got `summary of 24 test runs: 4868 passed; 0 failed; 330 ignored`.
Thanks!
Sean
At the moment this only includes type checking and there is no code generation support yet. I wanted to get the design reviewed first.
From discussion with @graydon at #5841, re-implemented as `#[simd]` attribute on structs.
Progressing towards #3499.
r? @ILyoan
This pulls all the logic for discovering the crate entry point into a new pass (out of resolve and typeck), then changes it so that main is only looked for at the crate level (`#[main]` can still be used anywhere).
I don't understand the special android logic here and worry that I may have broken it.
Also fixed the docstring on `TC_ONCE_CLOSURE` (was accidentally the same as `TC_MUTABLE`) and shifted the `TC_EMPTY_ENUM` bit left by one since whatever previously used that bit has been removed.
I just had `git apply` fix most of them and then did a quick skim over the diff to fix a few cases where it did the wrong thing (mostly replacing tabs with 4 spaces, when someone's editor had them at 8 spaces).
In rustpkg, pass around sysroot; in rustpkg tests, set the sysroot
manually so that tests can find libcore and such.
With bonus metadata::filesearch refactoring to avoid copies.
r? @pcwalton
Sorry this is so big, and sorry the first commit is just titled 'wip'.
Some interesting bits
* [LocalServices](f9069baa70) - This is the set of runtime capabilities that *all* Rust code should expect access to, including the local heap, GC, logging, unwinding.
* [impl Reader, etc. for Option](5fbb0949a5) - Constructors like `File::open` return Option<FileStream>. This lets you write I/O code without ever unwrapping an option.
This series adds a lot of [documentation](https://github.com/brson/rust/blob/io/src/libcore/rt/io/mod.rs#L11) to `core::rt::io`.
Recent demoding makes the visitor glue leak. It hasn't shown up in tests
because the box annihilator deletes the leaked boxes. This affects the
new scheduler though which does not yet have a box annihilator.
I don't think there's any great way to test this besides setting up
a task that doesn't run the box annihilator and I don't know that that's
a capability we want tasks to have.
Lots of linking arguments need to be passed as -Wl,--foo so giving the
comma meaning at the rustc layer makes those flags impossible to pass.
Multiple arguments can now be passed from a shell by quoting the
argument: --link-args='-lfoo -Wl,--as-needed'.
I don't know how one would write a separate test for this sort of thing. Building the compiler, and `make check` worked, which should mean I didn't screw anything.
Lots of linking arguments need to be passed as -Wl,--foo so giving the
comma meaning at the rustc layer makes those flags impossible to pass.
Multiple arguments can now be passed from a shell by quoting the
argument: --link-args='-lfoo -Wl,--as-needed'.
I've added trt_field_vtable, trt_field_box, and trt_field_tydesc, and
inserted them in place of the "magic numbers" used to access trait
object fields through GEPi().
The drop block has been deprecated for quite some time. This patch series removes support for parsing it and all the related machinery that made drop work.
As a side feature of all this, I also added the ability to annote fields in structs. This allows comments to be properly associated with an individual field. However, I didn't update `rustdoc` to integrate these comment blocks into the documentation it generates.
After much discussion on IRC and #4819, we have decided to revert to the old naming of the `/` operator. This does not change its behavior. In making this change, we also have had to rename some of the methods in the `Integer` trait. Here is a list of the methods that have changed:
- `Quot::quot` -> `Div::div`
- `Rem::rem` - stays the same
- `Integer::quot_rem` -> `Integer::div_rem`
- `Integer::div` -> `Integer::div_floor`
- `Integer::modulo` -> `Integer::mod_floor`
- `Integer::div_mod` -> `Integer::div_mod_floor`
Adds two extra flags: `--linker` which takes extra flags to pass to the linker, can be used multiple times and `--print-link-args` which prints out linker arguments. Currently `--print-link-args` needs execution to get past translation to get the `LinkMeta` data.
I haven't done tests or updated any extra documentation yet, so this pull request is currently here for review.
Cases like `Either<@int,()>` have a null case with at most one value but
a nonzero number of fields; if we misreport this, then bad things can
happen inside of, for example, pattern matching.
Closes#6117.
r? @pcwalton
A month's worth of parser cleanup here. Much of this is new comments and renaming. A number of these commits also remove unneeded code. Probably the biggest refactor here is splitting "parse_item_or_view_item" into two functions; it turns out that the only overlap between items in foreign modules and items in regular modules was macros, so this refactor should make things substantially easier for future maintenance.
