Use arena allocation instead of reference counting for `Module`s to fix memory leaks from `Rc` cycles.
A module references its module children and its import resolutions, and an import resolution references the module defining the imported name, so there is a cycle whenever a module imports something from an ancestor module.
For example,
```rust
mod foo { // `foo` references `bar`.
fn baz() {}
mod bar { // `bar` references the import.
use foo::baz; // The import references `foo`.
}
}
```
I also re-enabled the use of `#[thread_local]` on AArch64. It was originally disabled in the PR that introduced AArch64 (#19790), but the reasons for this were not explained. `#[thread_local]` seems to work fine in my tests on AArch64, so I don't think this should be an issue.
cc @alexcrichton @akiss77
This mixes in additional information into the hash that is
passed to -C extra-filename. It can be used to further distinguish
the standard libraries if they must be installed next to each
other.
Closes#29559
Frankly, I'm not sure if this solves a real problem. It's meant to help with side-by-side and overlapping installations where there are two sets of libs in /usr, but there are other potential issues there as well, including that some of our artifacts don't use this extra-filename munging, and it's not something our installers can support at all.
cc @jauhien Do you still think this helps the Gentoo case?
Since `darwin` is really `apple-darwin`, the valgrind-rpass tests were not actually being run with valgrind on mac before. Also, the `HOST` check was completely wrong.
r? @alexcrichton
This mixes in additional information into the hash that is
passed to -C extra-filename. It can be used to further distinguish
the standard libraries if they must be installed next to each
other.
Closes#29559
It's been awhile since we last updated jemalloc, and there's likely some bugs
that have been fixed since the last version we're using, so let's try to update
again.
It's been awhile since we last updated jemalloc, and there's likely some bugs
that have been fixed since the last version we're using, so let's try to update
again.
this PR reverts previous ones, that tried to make `cc` to found `estdc++` in `/usr/local/lib`. It causes more trouble than it resolvs things: rustc become unbuildable if another version already exists in `/usr/local` (for example, `libstd-xxxx.so` is found in `/usr/local/lib` and in builddir).
so this PR tries another way to achieve build, but using the good linker for building. By default, rustc use `cc` for linking. But under OpenBSD, `cc` is gcc 4.2.1 from base, whereas we build with gcc 4.9 from ports. By linking using the compiler found at compile-time, we ensure that the compiler will found his own stdc++ library without trouble.
r? @alexcrichton
By default, rustc use `cc` as linker. Under OpenBSD, `cc` is gcc version 4.2.1.
So use the compiler found at configure-time for linking: it will be gcc 4.9.
It permits to resolv problem of finding -lestdc++ or -lgcc. For base gcc (4.2), there are in not standard path, whereas for ports gcc (4.9) there are in standard path.
It looks like #27937 accidentally switched the llvmdeps file from the target to
the host by accident, so be sure to use the right llvmdeps file which is built
for the target when building rustc_llvm
This handles cases when the LLVM used isn't configured will the 'usual' targets. Also, cases where LLVM is shared are also handled (ie with `LD_LIBRARY_PATH` etc).
This handles cases when the LLVM used isn't configured will the 'usual'
targets. Also, cases where LLVM is shared are also handled (ie with
`LD_LIBRARY_PATH` etc).
With this commit, metadata encoding and decoding can make use of thread-local encoding and decoding contexts. These allow implementers of `serialize::Encodable` and `Decodable` to access information and
datastructures that would otherwise not be available to them. For example, we can automatically translate def-id and span information during decoding because the decoding context knows which crate the data is decoded from. Or it allows to make `ty::Ty` decodable because the context has access to the `ty::ctxt` that is needed for creating `ty::Ty` instances.
Some notes:
- `tls::with_encoding_context()` and `tls::with_decoding_context()` (as opposed to their unsafe versions) try to prevent the TLS data getting out-of-sync by making sure that the encoder/decoder passed in is actually the same as the one stored in the context. This should prevent accidentally reading from the wrong decoder.
- There are no real tests in this PR. I had a unit tests for some of the core aspects of the TLS implementation but it was kind of brittle, a lot of code for mocking `ty::ctxt`, `crate_metadata`, etc and did actually test not so much. The code will soon be tested by the first incremental compilation auto-tests that rely on MIR being properly serialized. However, if people think that some tests should be added before this can land, I'll try to provide some that make sense.
r? @nikomatsakis
With this commit, metadata encoding and decoding can make use of
thread-local encoding and decoding contexts. These allow implementers
of serialize::Encodable and Decodable to access information and
datastructures that would otherwise not be available to them. For
example, we can automatically translate def-id and span information
during decoding because the decoding context knows which crate the
data is decoded from. Or it allows to make ty::Ty decodable because
the context has access to the ty::ctxt that is needed for creating
ty::Ty instances.
The `rsbegin.o` and `rsend.o` build products should not be generated
on non WinGnu platforms.
This is another path to resolving #30063 for non win-gnu targets.
(And it won't require a snapshot, unlike PR #30208.)
r? @alexcrichton
The `rsbegin.o` and `rsend.o` build products should not be generated
on non WinGnu platforms.
This is another path to resolving #30063 for non win-gnu targets.
(And it won't require a snapshot, unlike PR #30208.)
In #29932, I moved the location of TRPL, but I missed making the changes
in mk/tests.mk. This led to #30088 landing with a broken example.
As such, #30113 will need to land before this.
Previously the file was not regenrated upon modification of `src/rustllvm` or others.
Now it will be rebuilt if `src/llvm` or `src/rustllvm` is touched.
Also added *.rs rule to 'clean' rule so that it is removed upon 'make
clean'.
Fixes#28614.
Debian wants to build all binaries with particular hardening flags.
The Rust makefiles are inconsistent in which architectures they
correctly include CFLAGS/etc from the enivoronment (see mk/cfg/*).
This patch adds LDFLAGS, and then unconditionally prepends
CFLAGS/LDFLAGS/etc to the build commands.
The book was located under 'src/doc/trpl' because originally, it was
going to be hosted under that URL. Late in the game, before 1.0, we
decided that /book was a better one, so we changed the output, but
not the input. This causes confusion for no good reason. So we'll change
the source directory to look like the output directory, like for every
other thing in src/doc.
Rather than modifying the installer to disable directory rewriting,
this patch modifies the directory structure passed to the installer so
that the rewriting gives the correct results. This means that if a
non-standard --libdir is passed to configure then the same --libdir
option (relative to the --prefix) must be passed to the install
script. In the `make install` case this is handled automatically.
Binary distributions are generally generated using the default
--libdir and then have paths optionally rewritten by the installer,
which should continue to work.
This has the advantage of not complicating the installer interface
intended for end-user use.
Fixes#29561
On distros that use i486 or i586 in their CHOST, Rust will fail to build
because it is not handling i486 or i586 like i686 is handled. This
changes the match to do work for all instances of i?86 instead of just
i686. The Yocto Project still uses i586 as a target.
Signed-off-by: Doug Goldstein <cardoe@cardoe.com>
under openbsd, the library path of libstdc++ need to be explicit (due
to the fact the default linker `cc` is gcc-4.2, and not gcc-4.9).
but when a recent LLVM is installed, rustc compilation pikes the bad
LLVM version (which live in /usr/local/lib, which is same directory of
libestdc++.so for gcc-4.9).
this patch move the libstdc++ path from RUST_FLAGS_<target> to special
variable, and use it *after* LLVM_LIBDIR_RUSTFLAGS_<target> in
arguments.
r? @alexcrichton
Rather than modifying the installer to disable directory rewriting,
this patch modifies the directory structure passed to the installer so
that the rewriting gives the correct results. This means that if a
non-standard --libdir is passed to configure then the same --libdir
option (relative to the --prefix) must be passed to the install
script. In the `make install` case this is handled automatically.
Binary distributions are generally generated using the default
--libdir and then have paths optionally rewritten by the installer,
which should continue to work.
This has the advantage of not complicating the installer interface
intended for end-user use.
Fixes#29561
On distros that use i486 or i586 in their CHOST, Rust will fail to build
because it is not handling i486 or i586 like i686 is handled. This
changes the match to do work for all instances of i?86 instead of just
i686. The Yocto Project still uses i586 as a target.
Signed-off-by: Doug Goldstein <cardoe@cardoe.com>
I noticed the nomicon was not listed!
I also removed links to racer and rustfmt since they were not *doc-specific* links, just links to tools, as well as pointed the cargo link directly at the docs.
Removed all the community stuff. There are lots of other places to find this now, including the website.
With pending website changes this page will continue to be pared back, reflecting only what's in-tree, not general Rust docs.
r? @steveklabnik
This should get `--libdir` working as well as it was a couple of weeks ago. (That is, it still rewrites paths incorrectly but it no longer fails during `make install`.)
Fixesgentoo/gentoo-rust#28 and gentoo/gentoo-rust#29.
This is to handle the case where CFG_LIBDIR is not a direct child of
CFG_PREFIX (in other words, where CFG_LIBDIR_RELATIVE has more than
one component).
Emacs warns that makefile lines that start with spaces followed by
tabs are "suspicious". These were harmless since they were
continuation lines, but getting rid of the warning is nice and this
version looks better.
The important one is $(MAKE). make handles recipes containing the
literal string "$(MAKE)" specially, so it is important to make sure it
isn't evaluated until recipe invocation time.
under openbsd, the library path of libstdc++ need to be explicit (due
to the fact the default linker `cc` is gcc-4.2, and not gcc-4.9).
but when a recent LLVM is installed, rustc compilation pikes the bad
LLVM version (which live in /usr/local/lib, which is same directory of
libestdc++.so for gcc-4.9).
this patch move the libstdc++ path from RUST_FLAGS_<target> to special
variable, and use it *after* LLVM_LIBDIR_RUSTFLAGS_<target> in
arguments.
Quite a bit of cruft in the valgrind suppressions. I started from a clean slate and found a few unique failures; this commit also moves the tests "fixed" by these suppressions into run-pass-valgrind.
* Delete `sys::unix::{c, sync}` as these are now all folded into libc itself
* Update all references to use `libc` as a result.
* Update all references to the new flat namespace.
* Moves all windows bindings into sys::c
According to a recent [discussion on IRC](https://botbot.me/mozilla/rust-tools/2015-10-27/?msg=52887517&page=2), there's no good reason for Windows builds to store target libraries under `bin`, when on every other platform they are under `lib`.
This might be a [breaking-change] for some users. I am pretty sure VisualRust has that path hard-coded somewhere.
r? @brson
Note: for now, this change only affects `-windows-gnu` builds.
So why was this `libgcc` dylib dependency needed in the first place?
The stack unwinder needs to know about locations of unwind tables of all the modules loaded in the current process. The easiest portable way of achieving this is to have each module register itself with the unwinder when loaded into the process. All modules compiled by GCC do this by calling the __register_frame_info() in their startup code (that's `crtbegin.o` and `crtend.o`, which are automatically linked into any gcc output).
Another important piece is that there should be only one copy of the unwinder (and thus unwind tables registry) in the process. This pretty much means that the unwinder must be in a shared library (unless everything is statically linked).
Now, Rust compiler tries very hard to make sure that any given Rust crate appears in the final output just once. So if we link the unwinder statically to one of Rust's crates, everything should be fine.
Unfortunately, GCC startup objects are built under assumption that `libgcc` is the one true place for the unwind info registry, so I couldn't find any better way than to replace them. So out go `crtbegin`/`crtend`, in come `rsbegin`/`rsend`!
A side benefit of this change is that rustc is now more in control of the command line that goes to the linker, so we could stop using `gcc` as the linker driver and just invoke `ld` directly.
Previously the file was not regenrated upon modification of src/rustllvm or others.
Now it will be rebuilt if `src/llvm` or `src/rustllvm` is touched.
Also added *.rs rule to 'clean' rule so that it is removed upon 'make
clean'.
It looks like the target libs aren't actually the same across hosts so instead
of always packaging the target libs from CFG_BUILD take the target libs from the
host if we have them and then only failing that do we take them from CFG_BUILD.
Closes#29228
This commit splits out the standard library from the current 'rustc' package
into a new 'rust-std' package. This is the basis for the work on easily
packaging compilers that can cross-compile to new targets.
Since it isn't possible to disable linkage of just GCC startup objects, we now need logic for finding libc installation directory and copying the required startup files (e.g. crt2.o) to rustlib directory.
Bonus change: use the `-nodefaultlibs` flag on Windows, thus paving the way to direct linker invocation.
