rust/compiler/rustc_metadata/src/locator.rs

1122 lines
46 KiB
Rust
Raw Normal View History

//! Finds crate binaries and loads their metadata
//!
//! Might I be the first to welcome you to a world of platform differences,
2014-08-02 01:40:21 +02:00
//! version requirements, dependency graphs, conflicting desires, and fun! This
//! is the major guts (along with metadata::creader) of the compiler for loading
//! crates and resolving dependencies. Let's take a tour!
//!
//! # The problem
//!
//! Each invocation of the compiler is immediately concerned with one primary
//! problem, to connect a set of crates to resolved crates on the filesystem.
//! Concretely speaking, the compiler follows roughly these steps to get here:
//!
//! 1. Discover a set of `extern crate` statements.
//! 2. Transform these directives into crate names. If the directive does not
//! have an explicit name, then the identifier is the name.
//! 3. For each of these crate names, find a corresponding crate on the
//! filesystem.
//!
//! Sounds easy, right? Let's walk into some of the nuances.
//!
//! ## Transitive Dependencies
//!
//! Let's say we've got three crates: A, B, and C. A depends on B, and B depends
//! on C. When we're compiling A, we primarily need to find and locate B, but we
//! also end up needing to find and locate C as well.
//!
//! The reason for this is that any of B's types could be composed of C's types,
//! any function in B could return a type from C, etc. To be able to guarantee
2019-02-08 14:53:55 +01:00
//! that we can always type-check/translate any function, we have to have
//! complete knowledge of the whole ecosystem, not just our immediate
//! dependencies.
//!
//! So now as part of the "find a corresponding crate on the filesystem" step
//! above, this involves also finding all crates for *all upstream
//! dependencies*. This includes all dependencies transitively.
//!
//! ## Rlibs and Dylibs
//!
//! The compiler has two forms of intermediate dependencies. These are dubbed
//! rlibs and dylibs for the static and dynamic variants, respectively. An rlib
//! is a rustc-defined file format (currently just an ar archive) while a dylib
//! is a platform-defined dynamic library. Each library has a metadata somewhere
//! inside of it.
//!
2016-11-10 22:00:33 +01:00
//! A third kind of dependency is an rmeta file. These are metadata files and do
//! not contain any code, etc. To a first approximation, these are treated in the
2016-11-10 03:57:30 +01:00
//! same way as rlibs. Where there is both an rlib and an rmeta file, the rlib
//! gets priority (even if the rmeta file is newer). An rmeta file is only
//! useful for checking a downstream crate, attempting to link one will cause an
//! error.
//!
//! When translating a crate name to a crate on the filesystem, we all of a
//! sudden need to take into account both rlibs and dylibs! Linkage later on may
//! use either one of these files, as each has their pros/cons. The job of crate
//! loading is to discover what's possible by finding all candidates.
//!
//! Most parts of this loading systems keep the dylib/rlib as just separate
//! variables.
//!
//! ## Where to look?
//!
//! We can't exactly scan your whole hard drive when looking for dependencies,
//! so we need to places to look. Currently the compiler will implicitly add the
//! target lib search path ($prefix/lib/rustlib/$target/lib) to any compilation,
//! and otherwise all -L flags are added to the search paths.
//!
//! ## What criterion to select on?
//!
//! This a pretty tricky area of loading crates. Given a file, how do we know
//! whether it's the right crate? Currently, the rules look along these lines:
//!
//! 1. Does the filename match an rlib/dylib pattern? That is to say, does the
//! filename have the right prefix/suffix?
//! 2. Does the filename have the right prefix for the crate name being queried?
//! This is filtering for files like `libfoo*.rlib` and such. If the crate
//! we're looking for was originally compiled with -C extra-filename, the
//! extra filename will be included in this prefix to reduce reading
//! metadata from crates that would otherwise share our prefix.
//! 3. Is the file an actual rust library? This is done by loading the metadata
//! from the library and making sure it's actually there.
//! 4. Does the name in the metadata agree with the name of the library?
//! 5. Does the target in the metadata agree with the current target?
//! 6. Does the SVH match? (more on this later)
//!
2014-08-02 01:40:21 +02:00
//! If the file answers `yes` to all these questions, then the file is
//! considered as being *candidate* for being accepted. It is illegal to have
//! more than two candidates as the compiler has no method by which to resolve
//! this conflict. Additionally, rlib/dylib candidates are considered
//! separately.
//!
//! After all this has happened, we have 1 or two files as candidates. These
//! represent the rlib/dylib file found for a library, and they're returned as
//! being found.
//!
//! ### What about versions?
//!
//! A lot of effort has been put forth to remove versioning from the compiler.
//! There have been forays in the past to have versioning baked in, but it was
//! largely always deemed insufficient to the point that it was recognized that
//! it's probably something the compiler shouldn't do anyway due to its
//! complicated nature and the state of the half-baked solutions.
//!
//! With a departure from versioning, the primary criterion for loading crates
//! is just the name of a crate. If we stopped here, it would imply that you
//! could never link two crates of the same name from different sources
//! together, which is clearly a bad state to be in.
//!
//! To resolve this problem, we come to the next section!
//!
//! # Expert Mode
//!
//! A number of flags have been added to the compiler to solve the "version
//! problem" in the previous section, as well as generally enabling more
//! powerful usage of the crate loading system of the compiler. The goal of
//! these flags and options are to enable third-party tools to drive the
//! compiler with prior knowledge about how the world should look.
//!
//! ## The `--extern` flag
//!
//! The compiler accepts a flag of this form a number of times:
//!
//! ```text
//! --extern crate-name=path/to/the/crate.rlib
//! ```
//!
//! This flag is basically the following letter to the compiler:
//!
//! > Dear rustc,
//! >
//! > When you are attempting to load the immediate dependency `crate-name`, I
2015-04-15 09:07:25 +02:00
//! > would like you to assume that the library is located at
//! > `path/to/the/crate.rlib`, and look nowhere else. Also, please do not
//! > assume that the path I specified has the name `crate-name`.
//!
//! This flag basically overrides most matching logic except for validating that
//! the file is indeed a rust library. The same `crate-name` can be specified
//! twice to specify the rlib/dylib pair.
//!
//! ## Enabling "multiple versions"
//!
//! This basically boils down to the ability to specify arbitrary packages to
//! the compiler. For example, if crate A wanted to use Bv1 and Bv2, then it
//! would look something like:
//!
//! ```compile_fail,E0463
//! extern crate b1;
//! extern crate b2;
//!
//! fn main() {}
//! ```
//!
//! and the compiler would be invoked as:
//!
//! ```text
//! rustc a.rs --extern b1=path/to/libb1.rlib --extern b2=path/to/libb2.rlib
//! ```
//!
//! In this scenario there are two crates named `b` and the compiler must be
//! manually driven to be informed where each crate is.
//!
//! ## Frobbing symbols
//!
//! One of the immediate problems with linking the same library together twice
//! in the same problem is dealing with duplicate symbols. The primary way to
//! deal with this in rustc is to add hashes to the end of each symbol.
//!
//! In order to force hashes to change between versions of a library, if
//! desired, the compiler exposes an option `-C metadata=foo`, which is used to
//! initially seed each symbol hash. The string `foo` is prepended to each
//! string-to-hash to ensure that symbols change over time.
//!
//! ## Loading transitive dependencies
//!
//! Dealing with same-named-but-distinct crates is not just a local problem, but
2014-08-02 01:40:21 +02:00
//! one that also needs to be dealt with for transitive dependencies. Note that
//! in the letter above `--extern` flags only apply to the *local* set of
//! dependencies, not the upstream transitive dependencies. Consider this
//! dependency graph:
//!
//! ```text
//! A.1 A.2
//! | |
//! | |
//! B C
//! \ /
//! \ /
//! D
//! ```
//!
//! In this scenario, when we compile `D`, we need to be able to distinctly
//! resolve `A.1` and `A.2`, but an `--extern` flag cannot apply to these
//! transitive dependencies.
//!
//! Note that the key idea here is that `B` and `C` are both *already compiled*.
//! That is, they have already resolved their dependencies. Due to unrelated
//! technical reasons, when a library is compiled, it is only compatible with
//! the *exact same* version of the upstream libraries it was compiled against.
//! We use the "Strict Version Hash" to identify the exact copy of an upstream
//! library.
//!
//! With this knowledge, we know that `B` and `C` will depend on `A` with
//! different SVH values, so we crawl the normal `-L` paths looking for
//! `liba*.rlib` and filter based on the contained SVH.
//!
//! In the end, this ends up not needing `--extern` to specify upstream
//! transitive dependencies.
//!
//! # Wrapping up
//!
//! That's the general overview of loading crates in the compiler, but it's by
//! no means all of the necessary details. Take a look at the rest of
2016-10-20 06:31:14 +02:00
//! metadata::locator or metadata::creader for all the juicy details!
2019-02-08 12:50:17 +01:00
use crate::creader::Library;
2019-12-22 23:42:04 +01:00
use crate::rmeta::{rustc_version, MetadataBlob, METADATA_HEADER};
2019-12-24 05:02:53 +01:00
use rustc_data_structures::fx::{FxHashMap, FxHashSet};
use rustc_data_structures::owning_ref::OwningRef;
2019-12-22 23:42:04 +01:00
use rustc_data_structures::svh::Svh;
use rustc_data_structures::sync::MetadataRef;
use rustc_errors::struct_span_err;
2020-03-29 17:19:48 +02:00
use rustc_middle::middle::cstore::{CrateSource, MetadataLoader};
use rustc_session::config::{self, CrateType};
use rustc_session::filesearch::{FileDoesntMatch, FileMatches, FileSearch};
use rustc_session::search_paths::PathKind;
use rustc_session::utils::CanonicalizedPath;
use rustc_session::{CrateDisambiguator, Session};
2020-01-01 19:30:57 +01:00
use rustc_span::symbol::{sym, Symbol};
use rustc_span::Span;
2019-12-22 23:42:04 +01:00
use rustc_target::spec::{Target, TargetTriple};
use snap::read::FrameDecoder;
use std::io::{Read, Result as IoResult, Write};
use std::ops::Deref;
use std::path::{Path, PathBuf};
use std::{cmp, fmt, fs};
2020-08-14 08:05:01 +02:00
use tracing::{debug, info, warn};
#[derive(Clone)]
crate struct CrateLocator<'a> {
// Immutable per-session configuration.
sess: &'a Session,
metadata_loader: &'a dyn MetadataLoader,
// Immutable per-search configuration.
crate_name: Symbol,
exact_paths: Vec<CanonicalizedPath>,
2019-11-24 13:29:35 +01:00
pub hash: Option<Svh>,
pub host_hash: Option<Svh>,
extra_filename: Option<&'a str>,
pub target: &'a Target,
pub triple: TargetTriple,
pub filesearch: FileSearch<'a>,
root: Option<&'a CratePaths>,
pub is_proc_macro: Option<bool>,
// Mutable in-progress state or output.
rejected_via_hash: Vec<CrateMismatch>,
rejected_via_triple: Vec<CrateMismatch>,
rejected_via_kind: Vec<CrateMismatch>,
rejected_via_version: Vec<CrateMismatch>,
rejected_via_filename: Vec<CrateMismatch>,
rustc: Optimize reading metadata by 4x We were previously reading metadata via `ar p`, but as learned from rustdoc awhile back, spawning a process to do something is pretty slow. Turns out LLVM has an Archive class to read archives, but it cannot write archives. This commits adds bindings to the read-only version of the LLVM archive class (with a new type that only has a read() method), and then it uses this class when reading the metadata out of rlibs. When you put this in tandem of not compressing the metadata, reading the metadata is 4x faster than it used to be The timings I got for reading metadata from the respective libraries was: libstd-04ff901e-0.9-pre.dylib => 100ms libstd-04ff901e-0.9-pre.rlib => 23ms librustuv-7945354c-0.9-pre.dylib => 4ms librustuv-7945354c-0.9-pre.rlib => 1ms librustc-5b94a16f-0.9-pre.dylib => 87ms librustc-5b94a16f-0.9-pre.rlib => 35ms libextra-a6ebb16f-0.9-pre.dylib => 63ms libextra-a6ebb16f-0.9-pre.rlib => 15ms libsyntax-2e4c0458-0.9-pre.dylib => 86ms libsyntax-2e4c0458-0.9-pre.rlib => 22ms In order to always take advantage of these faster metadata read-times, I sort the files in filesearch based on whether they have an rlib extension or not (prefer all rlib files first). Overall, this halved the compile time for a `fn main() {}` crate from 0.185s to 0.095s on my system (when preferring dynamic linking). Reading metadata is still the slowest pass of the compiler at 0.035s, but it's getting pretty close to linking at 0.021s! The next best optimization is to just not copy the metadata from LLVM because that's the most expensive part of reading metadata right now.
2013-12-17 05:58:21 +01:00
}
#[derive(Clone)]
crate struct CratePaths {
name: Symbol,
source: CrateSource,
}
impl CratePaths {
crate fn new(name: Symbol, source: CrateSource) -> CratePaths {
CratePaths { name, source }
}
}
#[derive(Copy, Clone, PartialEq)]
crate enum CrateFlavor {
Rlib,
2016-11-10 03:57:30 +01:00
Rmeta,
Dylib,
}
impl fmt::Display for CrateFlavor {
2019-02-08 12:50:17 +01:00
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
f.write_str(match *self {
CrateFlavor::Rlib => "rlib",
2016-11-10 03:57:30 +01:00
CrateFlavor::Rmeta => "rmeta",
CrateFlavor::Dylib => "dylib",
})
}
}
impl<'a> CrateLocator<'a> {
crate fn new(
sess: &'a Session,
metadata_loader: &'a dyn MetadataLoader,
crate_name: Symbol,
2019-11-24 13:29:35 +01:00
hash: Option<Svh>,
host_hash: Option<Svh>,
extra_filename: Option<&'a str>,
is_host: bool,
path_kind: PathKind,
root: Option<&'a CratePaths>,
is_proc_macro: Option<bool>,
) -> CrateLocator<'a> {
CrateLocator {
sess,
metadata_loader,
crate_name,
exact_paths: if hash.is_none() {
2019-12-22 23:42:04 +01:00
sess.opts
.externs
.get(&crate_name.as_str())
.into_iter()
2019-12-05 23:43:53 +01:00
.filter_map(|entry| entry.files())
.flatten()
.cloned()
2019-12-22 23:42:04 +01:00
.collect()
} else {
// SVH being specified means this is a transitive dependency,
// so `--extern` options do not apply.
Vec::new()
},
hash,
host_hash,
extra_filename,
target: if is_host { &sess.host } else { &sess.target },
triple: if is_host {
TargetTriple::from_triple(config::host_triple())
} else {
sess.opts.target_triple.clone()
},
filesearch: if is_host {
sess.host_filesearch(path_kind)
} else {
sess.target_filesearch(path_kind)
},
root,
is_proc_macro,
rejected_via_hash: Vec::new(),
rejected_via_triple: Vec::new(),
rejected_via_kind: Vec::new(),
rejected_via_version: Vec::new(),
rejected_via_filename: Vec::new(),
}
}
crate fn reset(&mut self) {
self.rejected_via_hash.clear();
self.rejected_via_triple.clear();
self.rejected_via_kind.clear();
self.rejected_via_version.clear();
self.rejected_via_filename.clear();
}
crate fn maybe_load_library_crate(&mut self) -> Result<Option<Library>, CrateError> {
if !self.exact_paths.is_empty() {
return self.find_commandline_library();
}
let mut seen_paths = FxHashSet::default();
if let Some(extra_filename) = self.extra_filename {
if let library @ Some(_) = self.find_library_crate(extra_filename, &mut seen_paths)? {
return Ok(library);
}
}
self.find_library_crate("", &mut seen_paths)
Add generation of static libraries to rustc This commit implements the support necessary for generating both intermediate and result static rust libraries. This is an implementation of my thoughts in https://mail.mozilla.org/pipermail/rust-dev/2013-November/006686.html. When compiling a library, we still retain the "lib" option, although now there are "rlib", "staticlib", and "dylib" as options for crate_type (and these are stackable). The idea of "lib" is to generate the "compiler default" instead of having too choose (although all are interchangeable). For now I have left the "complier default" to be a dynamic library for size reasons. Of the rust libraries, lib{std,extra,rustuv} will bootstrap with an rlib/dylib pair, but lib{rustc,syntax,rustdoc,rustpkg} will only be built as a dynamic object. I chose this for size reasons, but also because you're probably not going to be embedding the rustc compiler anywhere any time soon. Other than the options outlined above, there are a few defaults/preferences that are now opinionated in the compiler: * If both a .dylib and .rlib are found for a rust library, the compiler will prefer the .rlib variant. This is overridable via the -Z prefer-dynamic option * If generating a "lib", the compiler will generate a dynamic library. This is overridable by explicitly saying what flavor you'd like (rlib, staticlib, dylib). * If no options are passed to the command line, and no crate_type is found in the destination crate, then an executable is generated With this change, you can successfully build a rust program with 0 dynamic dependencies on rust libraries. There is still a dynamic dependency on librustrt, but I plan on removing that in a subsequent commit. This change includes no tests just yet. Our current testing infrastructure/harnesses aren't very amenable to doing flavorful things with linking, so I'm planning on adding a new mode of testing which I believe belongs as a separate commit. Closes #552
2013-11-15 23:03:29 +01:00
}
2019-12-22 23:42:04 +01:00
fn find_library_crate(
&mut self,
extra_prefix: &str,
seen_paths: &mut FxHashSet<PathBuf>,
) -> Result<Option<Library>, CrateError> {
Add generation of static libraries to rustc This commit implements the support necessary for generating both intermediate and result static rust libraries. This is an implementation of my thoughts in https://mail.mozilla.org/pipermail/rust-dev/2013-November/006686.html. When compiling a library, we still retain the "lib" option, although now there are "rlib", "staticlib", and "dylib" as options for crate_type (and these are stackable). The idea of "lib" is to generate the "compiler default" instead of having too choose (although all are interchangeable). For now I have left the "complier default" to be a dynamic library for size reasons. Of the rust libraries, lib{std,extra,rustuv} will bootstrap with an rlib/dylib pair, but lib{rustc,syntax,rustdoc,rustpkg} will only be built as a dynamic object. I chose this for size reasons, but also because you're probably not going to be embedding the rustc compiler anywhere any time soon. Other than the options outlined above, there are a few defaults/preferences that are now opinionated in the compiler: * If both a .dylib and .rlib are found for a rust library, the compiler will prefer the .rlib variant. This is overridable via the -Z prefer-dynamic option * If generating a "lib", the compiler will generate a dynamic library. This is overridable by explicitly saying what flavor you'd like (rlib, staticlib, dylib). * If no options are passed to the command line, and no crate_type is found in the destination crate, then an executable is generated With this change, you can successfully build a rust program with 0 dynamic dependencies on rust libraries. There is still a dynamic dependency on librustrt, but I plan on removing that in a subsequent commit. This change includes no tests just yet. Our current testing infrastructure/harnesses aren't very amenable to doing flavorful things with linking, so I'm planning on adding a new mode of testing which I believe belongs as a separate commit. Closes #552
2013-11-15 23:03:29 +01:00
// want: crate_name.dir_part() + prefix + crate_name.file_part + "-"
let dylib_prefix = format!("{}{}{}", self.target.dll_prefix, self.crate_name, extra_prefix);
let rlib_prefix = format!("lib{}{}", self.crate_name, extra_prefix);
let staticlib_prefix =
format!("{}{}{}", self.target.staticlib_prefix, self.crate_name, extra_prefix);
2019-12-22 23:42:04 +01:00
let mut candidates: FxHashMap<_, (FxHashMap<_, _>, FxHashMap<_, _>, FxHashMap<_, _>)> =
Default::default();
let mut staticlibs = vec![];
Re-work loading crates with nicer errors This commit rewrites crate loading internally in attempt to look at less metadata and provide nicer errors. The loading is now split up into a few stages: 1. Collect a mapping of (hash => ~[Path]) for a set of candidate libraries for a given search. The hash is the hash in the filename and the Path is the location of the library in question. All candidates are filtered based on their prefix/suffix (dylib/rlib appropriate) and then the hash/version are split up and are compared (if necessary). This means that if you're looking for an exact hash of library you don't have to open up the metadata of all libraries named the same, but also in your path. 2. Once this mapping is constructed, each (hash, ~[Path]) pair is filtered down to just a Path. This is necessary because the same rlib could show up twice in the path in multiple locations. Right now the filenames are based on just the crate id, so this could be indicative of multiple version of a crate during one crate_id lifetime in the path. If multiple duplicate crates are found, an error is generated. 3. Now that we have a mapping of (hash => Path), we error on multiple versions saying that multiple versions were found. Only if there's one (hash => Path) pair do we actually return that Path and its metadata. With this restructuring, it restructures code so errors which were assertions previously are now first-class errors. Additionally, this should read much less metadata with lots of crates of the same name or same version in a path. Closes #11908
2014-02-10 21:50:53 +01:00
// First, find all possible candidate rlibs and dylibs purely based on
// the name of the files themselves. We're trying to match against an
// exact crate name and a possibly an exact hash.
Re-work loading crates with nicer errors This commit rewrites crate loading internally in attempt to look at less metadata and provide nicer errors. The loading is now split up into a few stages: 1. Collect a mapping of (hash => ~[Path]) for a set of candidate libraries for a given search. The hash is the hash in the filename and the Path is the location of the library in question. All candidates are filtered based on their prefix/suffix (dylib/rlib appropriate) and then the hash/version are split up and are compared (if necessary). This means that if you're looking for an exact hash of library you don't have to open up the metadata of all libraries named the same, but also in your path. 2. Once this mapping is constructed, each (hash, ~[Path]) pair is filtered down to just a Path. This is necessary because the same rlib could show up twice in the path in multiple locations. Right now the filenames are based on just the crate id, so this could be indicative of multiple version of a crate during one crate_id lifetime in the path. If multiple duplicate crates are found, an error is generated. 3. Now that we have a mapping of (hash => Path), we error on multiple versions saying that multiple versions were found. Only if there's one (hash => Path) pair do we actually return that Path and its metadata. With this restructuring, it restructures code so errors which were assertions previously are now first-class errors. Additionally, this should read much less metadata with lots of crates of the same name or same version in a path. Closes #11908
2014-02-10 21:50:53 +01:00
//
// During this step, we can filter all found libraries based on the
// name and id found in the crate id (we ignore the path portion for
// filename matching), as well as the exact hash (if specified). If we
// end up having many candidates, we must look at the metadata to
// perform exact matches against hashes/crate ids. Note that opening up
// the metadata is where we do an exact match against the full contents
// of the crate id (path/name/id).
//
// The goal of this step is to look at as little metadata as possible.
self.filesearch.search(|spf, kind| {
let file = match &spf.file_name_str {
Re-work loading crates with nicer errors This commit rewrites crate loading internally in attempt to look at less metadata and provide nicer errors. The loading is now split up into a few stages: 1. Collect a mapping of (hash => ~[Path]) for a set of candidate libraries for a given search. The hash is the hash in the filename and the Path is the location of the library in question. All candidates are filtered based on their prefix/suffix (dylib/rlib appropriate) and then the hash/version are split up and are compared (if necessary). This means that if you're looking for an exact hash of library you don't have to open up the metadata of all libraries named the same, but also in your path. 2. Once this mapping is constructed, each (hash, ~[Path]) pair is filtered down to just a Path. This is necessary because the same rlib could show up twice in the path in multiple locations. Right now the filenames are based on just the crate id, so this could be indicative of multiple version of a crate during one crate_id lifetime in the path. If multiple duplicate crates are found, an error is generated. 3. Now that we have a mapping of (hash => Path), we error on multiple versions saying that multiple versions were found. Only if there's one (hash => Path) pair do we actually return that Path and its metadata. With this restructuring, it restructures code so errors which were assertions previously are now first-class errors. Additionally, this should read much less metadata with lots of crates of the same name or same version in a path. Closes #11908
2014-02-10 21:50:53 +01:00
None => return FileDoesntMatch,
Some(file) => file,
};
2019-12-22 23:42:04 +01:00
let (hash, found_kind) = if file.starts_with(&rlib_prefix) && file.ends_with(".rlib") {
(&file[(rlib_prefix.len())..(file.len() - ".rlib".len())], CrateFlavor::Rlib)
} else if file.starts_with(&rlib_prefix) && file.ends_with(".rmeta") {
(&file[(rlib_prefix.len())..(file.len() - ".rmeta".len())], CrateFlavor::Rmeta)
} else if file.starts_with(&dylib_prefix) && file.ends_with(&self.target.dll_suffix) {
(
&file[(dylib_prefix.len())..(file.len() - self.target.dll_suffix.len())],
CrateFlavor::Dylib,
)
2019-12-22 23:42:04 +01:00
} else {
if file.starts_with(&staticlib_prefix)
&& file.ends_with(&self.target.staticlib_suffix)
{
staticlibs
.push(CrateMismatch { path: spf.path.clone(), got: "static".to_string() });
2019-12-22 23:42:04 +01:00
}
return FileDoesntMatch;
};
info!("lib candidate: {}", spf.path.display());
2014-09-18 23:05:52 +02:00
let (rlibs, rmetas, dylibs) = candidates.entry(hash.to_string()).or_default();
let path = fs::canonicalize(&spf.path).unwrap_or_else(|_| spf.path.clone());
if seen_paths.contains(&path) {
return FileDoesntMatch;
};
seen_paths.insert(path.clone());
match found_kind {
CrateFlavor::Rlib => rlibs.insert(path, kind),
CrateFlavor::Rmeta => rmetas.insert(path, kind),
CrateFlavor::Dylib => dylibs.insert(path, kind),
};
FileMatches
2013-11-29 03:03:38 +01:00
});
self.rejected_via_kind.extend(staticlibs);
Add generation of static libraries to rustc This commit implements the support necessary for generating both intermediate and result static rust libraries. This is an implementation of my thoughts in https://mail.mozilla.org/pipermail/rust-dev/2013-November/006686.html. When compiling a library, we still retain the "lib" option, although now there are "rlib", "staticlib", and "dylib" as options for crate_type (and these are stackable). The idea of "lib" is to generate the "compiler default" instead of having too choose (although all are interchangeable). For now I have left the "complier default" to be a dynamic library for size reasons. Of the rust libraries, lib{std,extra,rustuv} will bootstrap with an rlib/dylib pair, but lib{rustc,syntax,rustdoc,rustpkg} will only be built as a dynamic object. I chose this for size reasons, but also because you're probably not going to be embedding the rustc compiler anywhere any time soon. Other than the options outlined above, there are a few defaults/preferences that are now opinionated in the compiler: * If both a .dylib and .rlib are found for a rust library, the compiler will prefer the .rlib variant. This is overridable via the -Z prefer-dynamic option * If generating a "lib", the compiler will generate a dynamic library. This is overridable by explicitly saying what flavor you'd like (rlib, staticlib, dylib). * If no options are passed to the command line, and no crate_type is found in the destination crate, then an executable is generated With this change, you can successfully build a rust program with 0 dynamic dependencies on rust libraries. There is still a dynamic dependency on librustrt, but I plan on removing that in a subsequent commit. This change includes no tests just yet. Our current testing infrastructure/harnesses aren't very amenable to doing flavorful things with linking, so I'm planning on adding a new mode of testing which I believe belongs as a separate commit. Closes #552
2013-11-15 23:03:29 +01:00
Re-work loading crates with nicer errors This commit rewrites crate loading internally in attempt to look at less metadata and provide nicer errors. The loading is now split up into a few stages: 1. Collect a mapping of (hash => ~[Path]) for a set of candidate libraries for a given search. The hash is the hash in the filename and the Path is the location of the library in question. All candidates are filtered based on their prefix/suffix (dylib/rlib appropriate) and then the hash/version are split up and are compared (if necessary). This means that if you're looking for an exact hash of library you don't have to open up the metadata of all libraries named the same, but also in your path. 2. Once this mapping is constructed, each (hash, ~[Path]) pair is filtered down to just a Path. This is necessary because the same rlib could show up twice in the path in multiple locations. Right now the filenames are based on just the crate id, so this could be indicative of multiple version of a crate during one crate_id lifetime in the path. If multiple duplicate crates are found, an error is generated. 3. Now that we have a mapping of (hash => Path), we error on multiple versions saying that multiple versions were found. Only if there's one (hash => Path) pair do we actually return that Path and its metadata. With this restructuring, it restructures code so errors which were assertions previously are now first-class errors. Additionally, this should read much less metadata with lots of crates of the same name or same version in a path. Closes #11908
2014-02-10 21:50:53 +01:00
// We have now collected all known libraries into a set of candidates
// keyed of the filename hash listed. For each filename, we also have a
// list of rlibs/dylibs that apply. Here, we map each of these lists
// (per hash), to a Library candidate for returning.
//
// A Library candidate is created if the metadata for the set of
// libraries corresponds to the crate id and hash criteria that this
2014-04-21 06:49:39 +02:00
// search is being performed for.
let mut libraries = FxHashMap::default();
2016-11-10 03:57:30 +01:00
for (_hash, (rlibs, rmetas, dylibs)) in candidates {
if let Some((svh, lib)) = self.extract_lib(rlibs, rmetas, dylibs)? {
libraries.insert(svh, lib);
Re-work loading crates with nicer errors This commit rewrites crate loading internally in attempt to look at less metadata and provide nicer errors. The loading is now split up into a few stages: 1. Collect a mapping of (hash => ~[Path]) for a set of candidate libraries for a given search. The hash is the hash in the filename and the Path is the location of the library in question. All candidates are filtered based on their prefix/suffix (dylib/rlib appropriate) and then the hash/version are split up and are compared (if necessary). This means that if you're looking for an exact hash of library you don't have to open up the metadata of all libraries named the same, but also in your path. 2. Once this mapping is constructed, each (hash, ~[Path]) pair is filtered down to just a Path. This is necessary because the same rlib could show up twice in the path in multiple locations. Right now the filenames are based on just the crate id, so this could be indicative of multiple version of a crate during one crate_id lifetime in the path. If multiple duplicate crates are found, an error is generated. 3. Now that we have a mapping of (hash => Path), we error on multiple versions saying that multiple versions were found. Only if there's one (hash => Path) pair do we actually return that Path and its metadata. With this restructuring, it restructures code so errors which were assertions previously are now first-class errors. Additionally, this should read much less metadata with lots of crates of the same name or same version in a path. Closes #11908
2014-02-10 21:50:53 +01:00
}
}
// Having now translated all relevant found hashes into libraries, see
// what we've got and figure out if we found multiple candidates for
// libraries or not.
match libraries.len() {
0 => Ok(None),
1 => Ok(Some(libraries.into_iter().next().unwrap().1)),
_ => Err(CrateError::MultipleMatchingCrates(self.crate_name, libraries)),
Add generation of static libraries to rustc This commit implements the support necessary for generating both intermediate and result static rust libraries. This is an implementation of my thoughts in https://mail.mozilla.org/pipermail/rust-dev/2013-November/006686.html. When compiling a library, we still retain the "lib" option, although now there are "rlib", "staticlib", and "dylib" as options for crate_type (and these are stackable). The idea of "lib" is to generate the "compiler default" instead of having too choose (although all are interchangeable). For now I have left the "complier default" to be a dynamic library for size reasons. Of the rust libraries, lib{std,extra,rustuv} will bootstrap with an rlib/dylib pair, but lib{rustc,syntax,rustdoc,rustpkg} will only be built as a dynamic object. I chose this for size reasons, but also because you're probably not going to be embedding the rustc compiler anywhere any time soon. Other than the options outlined above, there are a few defaults/preferences that are now opinionated in the compiler: * If both a .dylib and .rlib are found for a rust library, the compiler will prefer the .rlib variant. This is overridable via the -Z prefer-dynamic option * If generating a "lib", the compiler will generate a dynamic library. This is overridable by explicitly saying what flavor you'd like (rlib, staticlib, dylib). * If no options are passed to the command line, and no crate_type is found in the destination crate, then an executable is generated With this change, you can successfully build a rust program with 0 dynamic dependencies on rust libraries. There is still a dynamic dependency on librustrt, but I plan on removing that in a subsequent commit. This change includes no tests just yet. Our current testing infrastructure/harnesses aren't very amenable to doing flavorful things with linking, so I'm planning on adding a new mode of testing which I believe belongs as a separate commit. Closes #552
2013-11-15 23:03:29 +01:00
}
}
fn extract_lib(
&mut self,
rlibs: FxHashMap<PathBuf, PathKind>,
rmetas: FxHashMap<PathBuf, PathKind>,
dylibs: FxHashMap<PathBuf, PathKind>,
) -> Result<Option<(Svh, Library)>, CrateError> {
let mut slot = None;
// Order here matters, rmeta should come first. See comment in
// `extract_one` below.
let source = CrateSource {
rmeta: self.extract_one(rmetas, CrateFlavor::Rmeta, &mut slot)?,
rlib: self.extract_one(rlibs, CrateFlavor::Rlib, &mut slot)?,
dylib: self.extract_one(dylibs, CrateFlavor::Dylib, &mut slot)?,
};
Ok(slot.map(|(svh, metadata)| (svh, Library { source, metadata })))
}
fn needs_crate_flavor(&self, flavor: CrateFlavor) -> bool {
if flavor == CrateFlavor::Dylib && self.is_proc_macro == Some(true) {
return true;
}
// The all loop is because `--crate-type=rlib --crate-type=rlib` is
// legal and produces both inside this type.
2020-05-16 06:44:28 +02:00
let is_rlib = self.sess.crate_types().iter().all(|c| *c == CrateType::Rlib);
let needs_object_code = self.sess.opts.output_types.should_codegen();
// If we're producing an rlib, then we don't need object code.
// Or, if we're not producing object code, then we don't need it either
// (e.g., if we're a cdylib but emitting just metadata).
if is_rlib || !needs_object_code {
flavor == CrateFlavor::Rmeta
} else {
// we need all flavors (perhaps not true, but what we do for now)
true
}
}
Re-work loading crates with nicer errors This commit rewrites crate loading internally in attempt to look at less metadata and provide nicer errors. The loading is now split up into a few stages: 1. Collect a mapping of (hash => ~[Path]) for a set of candidate libraries for a given search. The hash is the hash in the filename and the Path is the location of the library in question. All candidates are filtered based on their prefix/suffix (dylib/rlib appropriate) and then the hash/version are split up and are compared (if necessary). This means that if you're looking for an exact hash of library you don't have to open up the metadata of all libraries named the same, but also in your path. 2. Once this mapping is constructed, each (hash, ~[Path]) pair is filtered down to just a Path. This is necessary because the same rlib could show up twice in the path in multiple locations. Right now the filenames are based on just the crate id, so this could be indicative of multiple version of a crate during one crate_id lifetime in the path. If multiple duplicate crates are found, an error is generated. 3. Now that we have a mapping of (hash => Path), we error on multiple versions saying that multiple versions were found. Only if there's one (hash => Path) pair do we actually return that Path and its metadata. With this restructuring, it restructures code so errors which were assertions previously are now first-class errors. Additionally, this should read much less metadata with lots of crates of the same name or same version in a path. Closes #11908
2014-02-10 21:50:53 +01:00
// Attempts to extract *one* library from the set `m`. If the set has no
// elements, `None` is returned. If the set has more than one element, then
// the errors and notes are emitted about the set of libraries.
//
// With only one library in the set, this function will extract it, and then
// read the metadata from it if `*slot` is `None`. If the metadata couldn't
// be read, it is assumed that the file isn't a valid rust library (no
// errors are emitted).
2019-12-22 23:42:04 +01:00
fn extract_one(
&mut self,
m: FxHashMap<PathBuf, PathKind>,
flavor: CrateFlavor,
slot: &mut Option<(Svh, MetadataBlob)>,
) -> Result<Option<(PathBuf, PathKind)>, CrateError> {
// If we are producing an rlib, and we've already loaded metadata, then
// we should not attempt to discover further crate sources (unless we're
// locating a proc macro; exact logic is in needs_crate_flavor). This means
// that under -Zbinary-dep-depinfo we will not emit a dependency edge on
// the *unused* rlib, and by returning `None` here immediately we
// guarantee that we do indeed not use it.
//
// See also #68149 which provides more detail on why emitting the
// dependency on the rlib is a bad thing.
//
2020-03-06 12:13:55 +01:00
// We currently do not verify that these other sources are even in sync,
// and this is arguably a bug (see #10786), but because reading metadata
// is quite slow (especially from dylibs) we currently do not read it
// from the other crate sources.
if slot.is_some() {
if m.is_empty() || !self.needs_crate_flavor(flavor) {
return Ok(None);
} else if m.len() == 1 {
return Ok(Some(m.into_iter().next().unwrap()));
}
}
let mut ret: Option<(PathBuf, PathKind)> = None;
let mut err_data: Option<Vec<PathBuf>> = None;
for (lib, kind) in m {
info!("{} reading metadata from: {}", flavor, lib.display());
let (hash, metadata) =
match get_metadata_section(self.target, flavor, &lib, self.metadata_loader) {
Ok(blob) => {
if let Some(h) = self.crate_matches(&blob, &lib) {
(h, blob)
} else {
info!("metadata mismatch");
continue;
}
}
Err(err) => {
2018-08-18 13:54:14 +02:00
warn!("no metadata found: {}", err);
continue;
Re-work loading crates with nicer errors This commit rewrites crate loading internally in attempt to look at less metadata and provide nicer errors. The loading is now split up into a few stages: 1. Collect a mapping of (hash => ~[Path]) for a set of candidate libraries for a given search. The hash is the hash in the filename and the Path is the location of the library in question. All candidates are filtered based on their prefix/suffix (dylib/rlib appropriate) and then the hash/version are split up and are compared (if necessary). This means that if you're looking for an exact hash of library you don't have to open up the metadata of all libraries named the same, but also in your path. 2. Once this mapping is constructed, each (hash, ~[Path]) pair is filtered down to just a Path. This is necessary because the same rlib could show up twice in the path in multiple locations. Right now the filenames are based on just the crate id, so this could be indicative of multiple version of a crate during one crate_id lifetime in the path. If multiple duplicate crates are found, an error is generated. 3. Now that we have a mapping of (hash => Path), we error on multiple versions saying that multiple versions were found. Only if there's one (hash => Path) pair do we actually return that Path and its metadata. With this restructuring, it restructures code so errors which were assertions previously are now first-class errors. Additionally, this should read much less metadata with lots of crates of the same name or same version in a path. Closes #11908
2014-02-10 21:50:53 +01:00
}
};
// If we see multiple hashes, emit an error about duplicate candidates.
if slot.as_ref().map_or(false, |s| s.0 != hash) {
if let Some(candidates) = err_data {
return Err(CrateError::MultipleCandidates(
self.crate_name,
flavor,
candidates,
));
2015-12-20 22:00:43 +01:00
}
err_data = Some(vec![ret.as_ref().unwrap().0.clone()]);
*slot = None;
}
if let Some(candidates) = &mut err_data {
candidates.push(lib);
continue;
Add generation of static libraries to rustc This commit implements the support necessary for generating both intermediate and result static rust libraries. This is an implementation of my thoughts in https://mail.mozilla.org/pipermail/rust-dev/2013-November/006686.html. When compiling a library, we still retain the "lib" option, although now there are "rlib", "staticlib", and "dylib" as options for crate_type (and these are stackable). The idea of "lib" is to generate the "compiler default" instead of having too choose (although all are interchangeable). For now I have left the "complier default" to be a dynamic library for size reasons. Of the rust libraries, lib{std,extra,rustuv} will bootstrap with an rlib/dylib pair, but lib{rustc,syntax,rustdoc,rustpkg} will only be built as a dynamic object. I chose this for size reasons, but also because you're probably not going to be embedding the rustc compiler anywhere any time soon. Other than the options outlined above, there are a few defaults/preferences that are now opinionated in the compiler: * If both a .dylib and .rlib are found for a rust library, the compiler will prefer the .rlib variant. This is overridable via the -Z prefer-dynamic option * If generating a "lib", the compiler will generate a dynamic library. This is overridable by explicitly saying what flavor you'd like (rlib, staticlib, dylib). * If no options are passed to the command line, and no crate_type is found in the destination crate, then an executable is generated With this change, you can successfully build a rust program with 0 dynamic dependencies on rust libraries. There is still a dynamic dependency on librustrt, but I plan on removing that in a subsequent commit. This change includes no tests just yet. Our current testing infrastructure/harnesses aren't very amenable to doing flavorful things with linking, so I'm planning on adding a new mode of testing which I believe belongs as a separate commit. Closes #552
2013-11-15 23:03:29 +01:00
}
// Ok so at this point we've determined that `(lib, kind)` above is
// a candidate crate to load, and that `slot` is either none (this
// is the first crate of its kind) or if some the previous path has
// the exact same hash (e.g., it's the exact same crate).
//
// In principle these two candidate crates are exactly the same so
// we can choose either of them to link. As a stupidly gross hack,
// however, we favor crate in the sysroot.
//
// You can find more info in rust-lang/rust#39518 and various linked
// issues, but the general gist is that during testing libstd the
// compilers has two candidates to choose from: one in the sysroot
// and one in the deps folder. These two crates are the exact same
// crate but if the compiler chooses the one in the deps folder
// it'll cause spurious errors on Windows.
//
// As a result, we favor the sysroot crate here. Note that the
// candidates are all canonicalized, so we canonicalize the sysroot
// as well.
if let Some((prev, _)) = &ret {
let sysroot = &self.sess.sysroot;
2019-12-22 23:42:04 +01:00
let sysroot = sysroot.canonicalize().unwrap_or_else(|_| sysroot.to_path_buf());
if prev.starts_with(&sysroot) {
2019-12-22 23:42:04 +01:00
continue;
}
}
*slot = Some((hash, metadata));
rustc: Fix a leak in dependency= paths With the addition of separate search paths to the compiler, it was intended that applications such as Cargo could require a `--extern` flag per `extern crate` directive in the source. The system can currently be subverted, however, due to the `existing_match()` logic in the crate loader. When loading crates we first attempt to match an `extern crate` directive against all previously loaded crates to avoid reading metadata twice. This "hit the cache if possible" step was erroneously leaking crates across the search path boundaries, however. For example: extern crate b; extern crate a; If `b` depends on `a`, then it will load crate `a` when the `extern crate b` directive is being processed. When the compiler reaches `extern crate a` it will use the previously loaded version no matter what. If the compiler was not invoked with `-L crate=path/to/a`, it will still succeed. This behavior is allowing `extern crate` declarations in Cargo without a corresponding declaration in the manifest of a dependency, which is considered a bug. This commit fixes this problem by keeping track of the origin search path for a crate. Crates loaded from the dependency search path are not candidates for crates which are loaded from the crate search path. As a result of this fix, this is a likely a breaking change for a number of Cargo packages. If the compiler starts informing that a crate can no longer be found, it likely means that the dependency was forgotten in your Cargo.toml. [breaking-change]
2015-01-06 17:46:07 +01:00
ret = Some((lib, kind));
Add generation of static libraries to rustc This commit implements the support necessary for generating both intermediate and result static rust libraries. This is an implementation of my thoughts in https://mail.mozilla.org/pipermail/rust-dev/2013-November/006686.html. When compiling a library, we still retain the "lib" option, although now there are "rlib", "staticlib", and "dylib" as options for crate_type (and these are stackable). The idea of "lib" is to generate the "compiler default" instead of having too choose (although all are interchangeable). For now I have left the "complier default" to be a dynamic library for size reasons. Of the rust libraries, lib{std,extra,rustuv} will bootstrap with an rlib/dylib pair, but lib{rustc,syntax,rustdoc,rustpkg} will only be built as a dynamic object. I chose this for size reasons, but also because you're probably not going to be embedding the rustc compiler anywhere any time soon. Other than the options outlined above, there are a few defaults/preferences that are now opinionated in the compiler: * If both a .dylib and .rlib are found for a rust library, the compiler will prefer the .rlib variant. This is overridable via the -Z prefer-dynamic option * If generating a "lib", the compiler will generate a dynamic library. This is overridable by explicitly saying what flavor you'd like (rlib, staticlib, dylib). * If no options are passed to the command line, and no crate_type is found in the destination crate, then an executable is generated With this change, you can successfully build a rust program with 0 dynamic dependencies on rust libraries. There is still a dynamic dependency on librustrt, but I plan on removing that in a subsequent commit. This change includes no tests just yet. Our current testing infrastructure/harnesses aren't very amenable to doing flavorful things with linking, so I'm planning on adding a new mode of testing which I believe belongs as a separate commit. Closes #552
2013-11-15 23:03:29 +01:00
}
2015-12-20 22:00:43 +01:00
if let Some(candidates) = err_data {
Err(CrateError::MultipleCandidates(self.crate_name, flavor, candidates))
2015-12-20 22:00:43 +01:00
} else {
Ok(ret)
2015-12-20 22:00:43 +01:00
}
Add generation of static libraries to rustc This commit implements the support necessary for generating both intermediate and result static rust libraries. This is an implementation of my thoughts in https://mail.mozilla.org/pipermail/rust-dev/2013-November/006686.html. When compiling a library, we still retain the "lib" option, although now there are "rlib", "staticlib", and "dylib" as options for crate_type (and these are stackable). The idea of "lib" is to generate the "compiler default" instead of having too choose (although all are interchangeable). For now I have left the "complier default" to be a dynamic library for size reasons. Of the rust libraries, lib{std,extra,rustuv} will bootstrap with an rlib/dylib pair, but lib{rustc,syntax,rustdoc,rustpkg} will only be built as a dynamic object. I chose this for size reasons, but also because you're probably not going to be embedding the rustc compiler anywhere any time soon. Other than the options outlined above, there are a few defaults/preferences that are now opinionated in the compiler: * If both a .dylib and .rlib are found for a rust library, the compiler will prefer the .rlib variant. This is overridable via the -Z prefer-dynamic option * If generating a "lib", the compiler will generate a dynamic library. This is overridable by explicitly saying what flavor you'd like (rlib, staticlib, dylib). * If no options are passed to the command line, and no crate_type is found in the destination crate, then an executable is generated With this change, you can successfully build a rust program with 0 dynamic dependencies on rust libraries. There is still a dynamic dependency on librustrt, but I plan on removing that in a subsequent commit. This change includes no tests just yet. Our current testing infrastructure/harnesses aren't very amenable to doing flavorful things with linking, so I'm planning on adding a new mode of testing which I believe belongs as a separate commit. Closes #552
2013-11-15 23:03:29 +01:00
}
fn crate_matches(&mut self, metadata: &MetadataBlob, libpath: &Path) -> Option<Svh> {
let rustc_version = rustc_version();
let found_version = metadata.get_rustc_version();
if found_version != rustc_version {
2019-12-22 23:42:04 +01:00
info!("Rejecting via version: expected {} got {}", rustc_version, found_version);
self.rejected_via_version
.push(CrateMismatch { path: libpath.to_path_buf(), got: found_version });
return None;
}
let root = metadata.get_root();
if let Some(expected_is_proc_macro) = self.is_proc_macro {
let is_proc_macro = root.is_proc_macro_crate();
if is_proc_macro != expected_is_proc_macro {
2019-12-22 23:42:04 +01:00
info!(
"Rejecting via proc macro: expected {} got {}",
expected_is_proc_macro, is_proc_macro
);
return None;
}
}
2020-10-27 01:02:06 +01:00
if self.exact_paths.is_empty() && self.crate_name != root.name() {
info!("Rejecting via crate name");
return None;
}
if root.triple() != &self.triple {
2019-12-22 23:42:04 +01:00
info!("Rejecting via crate triple: expected {} got {}", self.triple, root.triple());
self.rejected_via_triple.push(CrateMismatch {
path: libpath.to_path_buf(),
got: root.triple().to_string(),
});
return None;
}
let hash = root.hash();
2019-11-24 13:29:35 +01:00
if let Some(expected_hash) = self.hash {
if hash != expected_hash {
info!("Rejecting via hash: expected {} got {}", expected_hash, hash);
2019-12-22 23:42:04 +01:00
self.rejected_via_hash
.push(CrateMismatch { path: libpath.to_path_buf(), got: hash.to_string() });
return None;
}
}
Some(hash)
}
fn find_commandline_library(&mut self) -> Result<Option<Library>, CrateError> {
// First, filter out all libraries that look suspicious. We only accept
// files which actually exist that have the correct naming scheme for
// rlibs/dylibs.
let mut rlibs = FxHashMap::default();
let mut rmetas = FxHashMap::default();
let mut dylibs = FxHashMap::default();
for loc in &self.exact_paths {
if !loc.canonicalized().exists() {
return Err(CrateError::ExternLocationNotExist(
self.crate_name,
loc.original().clone(),
));
}
let file = match loc.original().file_name().and_then(|s| s.to_str()) {
Some(file) => file,
None => {
return Err(CrateError::ExternLocationNotFile(
self.crate_name,
loc.original().clone(),
));
}
};
if file.starts_with("lib") && (file.ends_with(".rlib") || file.ends_with(".rmeta"))
|| file.starts_with(&self.target.dll_prefix)
&& file.ends_with(&self.target.dll_suffix)
{
// Make sure there's at most one rlib and at most one dylib.
// Note to take care and match against the non-canonicalized name:
// some systems save build artifacts into content-addressed stores
// that do not preserve extensions, and then link to them using
// e.g. symbolic links. If we canonicalize too early, we resolve
// the symlink, the file type is lost and we might treat rlibs and
// rmetas as dylibs.
let loc_canon = loc.canonicalized().clone();
let loc = loc.original();
if loc.file_name().unwrap().to_str().unwrap().ends_with(".rlib") {
rlibs.insert(loc_canon, PathKind::ExternFlag);
2016-11-10 03:57:30 +01:00
} else if loc.file_name().unwrap().to_str().unwrap().ends_with(".rmeta") {
rmetas.insert(loc_canon, PathKind::ExternFlag);
} else {
dylibs.insert(loc_canon, PathKind::ExternFlag);
}
} else {
self.rejected_via_filename
.push(CrateMismatch { path: loc.original().clone(), got: String::new() });
}
}
// Extract the dylib/rlib/rmeta triple.
Ok(self.extract_lib(rlibs, rmetas, dylibs)?.map(|(_, lib)| lib))
}
crate fn into_error(self) -> CrateError {
CrateError::LocatorCombined(CombinedLocatorError {
crate_name: self.crate_name,
root: self.root.cloned(),
triple: self.triple,
dll_prefix: self.target.dll_prefix.clone(),
dll_suffix: self.target.dll_suffix.clone(),
rejected_via_hash: self.rejected_via_hash,
rejected_via_triple: self.rejected_via_triple,
rejected_via_kind: self.rejected_via_kind,
rejected_via_version: self.rejected_via_version,
rejected_via_filename: self.rejected_via_filename,
})
}
}
/// A trivial wrapper for `Mmap` that implements `StableDeref`.
struct StableDerefMmap(memmap2::Mmap);
impl Deref for StableDerefMmap {
type Target = [u8];
fn deref(&self) -> &[u8] {
self.0.deref()
}
}
unsafe impl stable_deref_trait::StableDeref for StableDerefMmap {}
fn get_metadata_section(
2019-12-22 23:42:04 +01:00
target: &Target,
flavor: CrateFlavor,
filename: &Path,
loader: &dyn MetadataLoader,
) -> Result<MetadataBlob, String> {
if !filename.exists() {
return Err(format!("no such file: '{}'", filename.display()));
}
2018-03-03 06:17:06 +01:00
let raw_bytes: MetadataRef = match flavor {
CrateFlavor::Rlib => loader.get_rlib_metadata(target, filename)?,
CrateFlavor::Dylib => {
let buf = loader.get_dylib_metadata(target, filename)?;
// The header is uncompressed
let header_len = METADATA_HEADER.len();
debug!("checking {} bytes of metadata-version stamp", header_len);
let header = &buf[..cmp::min(header_len, buf.len())];
if header != METADATA_HEADER {
2019-12-22 23:42:04 +01:00
return Err(format!(
"incompatible metadata version found: '{}'",
filename.display()
));
rustc: Optimize reading metadata by 4x We were previously reading metadata via `ar p`, but as learned from rustdoc awhile back, spawning a process to do something is pretty slow. Turns out LLVM has an Archive class to read archives, but it cannot write archives. This commits adds bindings to the read-only version of the LLVM archive class (with a new type that only has a read() method), and then it uses this class when reading the metadata out of rlibs. When you put this in tandem of not compressing the metadata, reading the metadata is 4x faster than it used to be The timings I got for reading metadata from the respective libraries was: libstd-04ff901e-0.9-pre.dylib => 100ms libstd-04ff901e-0.9-pre.rlib => 23ms librustuv-7945354c-0.9-pre.dylib => 4ms librustuv-7945354c-0.9-pre.rlib => 1ms librustc-5b94a16f-0.9-pre.dylib => 87ms librustc-5b94a16f-0.9-pre.rlib => 35ms libextra-a6ebb16f-0.9-pre.dylib => 63ms libextra-a6ebb16f-0.9-pre.rlib => 15ms libsyntax-2e4c0458-0.9-pre.dylib => 86ms libsyntax-2e4c0458-0.9-pre.rlib => 22ms In order to always take advantage of these faster metadata read-times, I sort the files in filesearch based on whether they have an rlib extension or not (prefer all rlib files first). Overall, this halved the compile time for a `fn main() {}` crate from 0.185s to 0.095s on my system (when preferring dynamic linking). Reading metadata is still the slowest pass of the compiler at 0.035s, but it's getting pretty close to linking at 0.021s! The next best optimization is to just not copy the metadata from LLVM because that's the most expensive part of reading metadata right now.
2013-12-17 05:58:21 +01:00
}
2013-06-20 07:52:02 +02:00
// Header is okay -> inflate the actual metadata
let compressed_bytes = &buf[header_len..];
debug!("inflating {} bytes of compressed metadata", compressed_bytes.len());
let mut inflated = Vec::new();
match FrameDecoder::new(compressed_bytes).read_to_end(&mut inflated) {
2019-12-22 23:42:04 +01:00
Ok(_) => rustc_erase_owner!(OwningRef::new(inflated).map_owner_box()),
Err(_) => {
return Err(format!("failed to decompress metadata: {}", filename.display()));
}
}
}
CrateFlavor::Rmeta => {
// mmap the file, because only a small fraction of it is read.
2019-12-22 23:42:04 +01:00
let file = std::fs::File::open(filename)
.map_err(|_| format!("failed to open rmeta metadata: '{}'", filename.display()))?;
let mmap = unsafe { memmap2::Mmap::map(&file) };
2019-12-22 23:42:04 +01:00
let mmap = mmap
.map_err(|_| format!("failed to mmap rmeta metadata: '{}'", filename.display()))?;
rustc_erase_owner!(OwningRef::new(StableDerefMmap(mmap)).map_owner_box())
}
};
let blob = MetadataBlob::new(raw_bytes);
if blob.is_compatible() {
Ok(blob)
} else {
Err(format!("incompatible metadata version found: '{}'", filename.display()))
}
}
/// Look for a plugin registrar. Returns its library path and crate disambiguator.
pub fn find_plugin_registrar(
sess: &Session,
metadata_loader: &dyn MetadataLoader,
span: Span,
name: Symbol,
) -> (PathBuf, CrateDisambiguator) {
match find_plugin_registrar_impl(sess, metadata_loader, name) {
Ok(res) => res,
Err(err) => err.report(sess, span),
}
}
fn find_plugin_registrar_impl<'a>(
sess: &'a Session,
metadata_loader: &dyn MetadataLoader,
name: Symbol,
) -> Result<(PathBuf, CrateDisambiguator), CrateError> {
info!("find plugin registrar `{}`", name);
let mut locator = CrateLocator::new(
sess,
metadata_loader,
name,
None, // hash
None, // host_hash
None, // extra_filename
true, // is_host
PathKind::Crate,
None, // root
None, // is_proc_macro
);
match locator.maybe_load_library_crate()? {
Some(library) => match library.source.dylib {
Some(dylib) => Ok((dylib.0, library.metadata.get_root().disambiguator())),
None => Err(CrateError::NonDylibPlugin(name)),
},
None => Err(locator.into_error()),
}
}
2019-02-08 14:53:55 +01:00
/// A diagnostic function for dumping crate metadata to an output stream.
2019-12-22 23:42:04 +01:00
pub fn list_file_metadata(
target: &Target,
path: &Path,
metadata_loader: &dyn MetadataLoader,
out: &mut dyn Write,
) -> IoResult<()> {
let filename = path.file_name().unwrap().to_str().unwrap();
let flavor = if filename.ends_with(".rlib") {
CrateFlavor::Rlib
2016-11-10 03:57:30 +01:00
} else if filename.ends_with(".rmeta") {
CrateFlavor::Rmeta
} else {
CrateFlavor::Dylib
};
match get_metadata_section(target, flavor, path, metadata_loader) {
Ok(metadata) => metadata.list_crate_metadata(out),
Err(msg) => write!(out, "{}\n", msg),
}
}
// ------------------------------------------ Error reporting -------------------------------------
#[derive(Clone)]
struct CrateMismatch {
path: PathBuf,
got: String,
}
/// Candidate rejection reasons collected during crate search.
/// If no candidate is accepted, then these reasons are presented to the user,
/// otherwise they are ignored.
crate struct CombinedLocatorError {
crate_name: Symbol,
root: Option<CratePaths>,
triple: TargetTriple,
dll_prefix: String,
dll_suffix: String,
rejected_via_hash: Vec<CrateMismatch>,
rejected_via_triple: Vec<CrateMismatch>,
rejected_via_kind: Vec<CrateMismatch>,
rejected_via_version: Vec<CrateMismatch>,
rejected_via_filename: Vec<CrateMismatch>,
}
crate enum CrateError {
NonAsciiName(Symbol),
ExternLocationNotExist(Symbol, PathBuf),
ExternLocationNotFile(Symbol, PathBuf),
MultipleCandidates(Symbol, CrateFlavor, Vec<PathBuf>),
MultipleMatchingCrates(Symbol, FxHashMap<Svh, Library>),
SymbolConflictsCurrent(Symbol),
SymbolConflictsOthers(Symbol),
StableCrateIdCollision(Symbol, Symbol),
DlOpen(String),
DlSym(String),
LocatorCombined(CombinedLocatorError),
NonDylibPlugin(Symbol),
}
impl CrateError {
crate fn report(self, sess: &Session, span: Span) -> ! {
let mut err = match self {
CrateError::NonAsciiName(crate_name) => sess.struct_span_err(
span,
&format!("cannot load a crate with a non-ascii name `{}`", crate_name),
),
CrateError::ExternLocationNotExist(crate_name, loc) => sess.struct_span_err(
span,
&format!("extern location for {} does not exist: {}", crate_name, loc.display()),
),
CrateError::ExternLocationNotFile(crate_name, loc) => sess.struct_span_err(
span,
&format!("extern location for {} is not a file: {}", crate_name, loc.display()),
),
CrateError::MultipleCandidates(crate_name, flavor, candidates) => {
let mut err = struct_span_err!(
sess,
span,
E0465,
"multiple {} candidates for `{}` found",
flavor,
crate_name,
);
for (i, candidate) in candidates.iter().enumerate() {
err.span_note(span, &format!("candidate #{}: {}", i + 1, candidate.display()));
}
err
}
CrateError::MultipleMatchingCrates(crate_name, libraries) => {
let mut err = struct_span_err!(
sess,
span,
E0464,
"multiple matching crates for `{}`",
crate_name
);
let candidates = libraries
.iter()
.filter_map(|(_, lib)| {
let crate_name = &lib.metadata.get_root().name().as_str();
match (&lib.source.dylib, &lib.source.rlib) {
(Some((pd, _)), Some((pr, _))) => Some(format!(
"\ncrate `{}`: {}\n{:>padding$}",
crate_name,
pd.display(),
pr.display(),
padding = 8 + crate_name.len()
)),
(Some((p, _)), None) | (None, Some((p, _))) => {
Some(format!("\ncrate `{}`: {}", crate_name, p.display()))
}
(None, None) => None,
}
})
.collect::<String>();
err.note(&format!("candidates:{}", candidates));
err
}
CrateError::SymbolConflictsCurrent(root_name) => struct_span_err!(
sess,
span,
E0519,
"the current crate is indistinguishable from one of its dependencies: it has the \
same crate-name `{}` and was compiled with the same `-C metadata` arguments. \
This will result in symbol conflicts between the two.",
root_name,
),
CrateError::SymbolConflictsOthers(root_name) => struct_span_err!(
sess,
span,
E0523,
"found two different crates with name `{}` that are not distinguished by differing \
`-C metadata`. This will result in symbol conflicts between the two.",
root_name,
),
CrateError::StableCrateIdCollision(crate_name0, crate_name1) => {
let msg = format!(
"found crates (`{}` and `{}`) with colliding StableCrateId values.",
crate_name0, crate_name1
);
sess.struct_span_err(span, &msg)
}
CrateError::DlOpen(s) | CrateError::DlSym(s) => sess.struct_span_err(span, &s),
CrateError::LocatorCombined(locator) => {
let crate_name = locator.crate_name;
let add = match &locator.root {
None => String::new(),
Some(r) => format!(" which `{}` depends on", r.name),
};
let mut msg = "the following crate versions were found:".to_string();
let mut err = if !locator.rejected_via_hash.is_empty() {
let mut err = struct_span_err!(
sess,
span,
E0460,
"found possibly newer version of crate `{}`{}",
crate_name,
add,
);
err.note("perhaps that crate needs to be recompiled?");
let mismatches = locator.rejected_via_hash.iter();
for CrateMismatch { path, .. } in mismatches {
msg.push_str(&format!("\ncrate `{}`: {}", crate_name, path.display()));
}
if let Some(r) = locator.root {
for path in r.source.paths() {
msg.push_str(&format!("\ncrate `{}`: {}", r.name, path.display()));
}
}
err.note(&msg);
err
} else if !locator.rejected_via_triple.is_empty() {
let mut err = struct_span_err!(
sess,
span,
E0461,
"couldn't find crate `{}` with expected target triple {}{}",
crate_name,
locator.triple,
add,
);
let mismatches = locator.rejected_via_triple.iter();
for CrateMismatch { path, got } in mismatches {
msg.push_str(&format!(
"\ncrate `{}`, target triple {}: {}",
crate_name,
got,
path.display(),
));
}
err.note(&msg);
err
} else if !locator.rejected_via_kind.is_empty() {
let mut err = struct_span_err!(
sess,
span,
E0462,
"found staticlib `{}` instead of rlib or dylib{}",
crate_name,
add,
);
err.help("please recompile that crate using --crate-type lib");
let mismatches = locator.rejected_via_kind.iter();
for CrateMismatch { path, .. } in mismatches {
msg.push_str(&format!("\ncrate `{}`: {}", crate_name, path.display()));
}
err.note(&msg);
err
} else if !locator.rejected_via_version.is_empty() {
let mut err = struct_span_err!(
sess,
span,
E0514,
"found crate `{}` compiled by an incompatible version of rustc{}",
crate_name,
add,
);
err.help(&format!(
"please recompile that crate using this compiler ({})",
rustc_version(),
));
let mismatches = locator.rejected_via_version.iter();
for CrateMismatch { path, got } in mismatches {
msg.push_str(&format!(
"\ncrate `{}` compiled by {}: {}",
crate_name,
got,
path.display(),
));
}
err.note(&msg);
err
} else {
let mut err = struct_span_err!(
sess,
span,
E0463,
"can't find crate for `{}`{}",
crate_name,
add,
);
if (crate_name == sym::std || crate_name == sym::core)
&& locator.triple != TargetTriple::from_triple(config::host_triple())
{
err.note(&format!("the `{}` target may not be installed", locator.triple));
} else if crate_name == sym::profiler_builtins {
err.note(&"the compiler may have been built without the profiler runtime");
}
err.span_label(span, "can't find crate");
err
};
if !locator.rejected_via_filename.is_empty() {
let mismatches = locator.rejected_via_filename.iter();
for CrateMismatch { path, .. } in mismatches {
err.note(&format!(
"extern location for {} is of an unknown type: {}",
crate_name,
path.display(),
))
.help(&format!(
"file name should be lib*.rlib or {}*.{}",
locator.dll_prefix, locator.dll_suffix
));
}
}
err
}
CrateError::NonDylibPlugin(crate_name) => struct_span_err!(
sess,
span,
E0457,
"plugin `{}` only found in rlib format, but must be available in dylib format",
crate_name,
),
};
err.emit();
sess.abort_if_errors();
unreachable!();
}
}