更新libclamav库1.0.0版本

This commit is contained in:
2023-01-14 18:28:39 +08:00
parent b879ee0b2e
commit 45fe15f472
8531 changed files with 1222046 additions and 177272 deletions

View File

@@ -0,0 +1,319 @@
// SPDX-License-Identifier: Apache-2.0
extern crate glob;
use std::cell::RefCell;
use std::collections::HashMap;
use std::env;
use std::path::{Path, PathBuf};
use std::process::Command;
use glob::{MatchOptions, Pattern};
//================================================
// Commands
//================================================
thread_local! {
/// The errors encountered by the build script while executing commands.
static COMMAND_ERRORS: RefCell<HashMap<String, Vec<String>>> = RefCell::default();
}
/// Adds an error encountered by the build script while executing a command.
fn add_command_error(name: &str, path: &str, arguments: &[&str], message: String) {
COMMAND_ERRORS.with(|e| {
e.borrow_mut()
.entry(name.into())
.or_insert_with(Vec::new)
.push(format!(
"couldn't execute `{} {}` (path={}) ({})",
name,
arguments.join(" "),
path,
message,
))
});
}
/// A struct that prints the errors encountered by the build script while
/// executing commands when dropped (unless explictly discarded).
///
/// This is handy because we only want to print these errors when the build
/// script fails to link to an instance of `libclang`. For example, if
/// `llvm-config` couldn't be executed but an instance of `libclang` was found
/// anyway we don't want to pollute the build output with irrelevant errors.
#[derive(Default)]
pub struct CommandErrorPrinter {
discard: bool,
}
impl CommandErrorPrinter {
pub fn discard(mut self) {
self.discard = true;
}
}
impl Drop for CommandErrorPrinter {
fn drop(&mut self) {
if self.discard {
return;
}
let errors = COMMAND_ERRORS.with(|e| e.borrow().clone());
if let Some(errors) = errors.get("llvm-config") {
println!(
"cargo:warning=could not execute `llvm-config` one or more \
times, if the LLVM_CONFIG_PATH environment variable is set to \
a full path to valid `llvm-config` executable it will be used \
to try to find an instance of `libclang` on your system: {}",
errors
.iter()
.map(|e| format!("\"{}\"", e))
.collect::<Vec<_>>()
.join("\n "),
)
}
if let Some(errors) = errors.get("xcode-select") {
println!(
"cargo:warning=could not execute `xcode-select` one or more \
times, if a valid instance of this executable is on your PATH \
it will be used to try to find an instance of `libclang` on \
your system: {}",
errors
.iter()
.map(|e| format!("\"{}\"", e))
.collect::<Vec<_>>()
.join("\n "),
)
}
}
}
/// Executes a command and returns the `stdout` output if the command was
/// successfully executed (errors are added to `COMMAND_ERRORS`).
fn run_command(name: &str, path: &str, arguments: &[&str]) -> Option<String> {
let output = match Command::new(path).args(arguments).output() {
Ok(output) => output,
Err(error) => {
let message = format!("error: {}", error);
add_command_error(name, path, arguments, message);
return None;
}
};
if output.status.success() {
Some(String::from_utf8_lossy(&output.stdout).into_owned())
} else {
let message = format!("exit code: {}", output.status);
add_command_error(name, path, arguments, message);
None
}
}
/// Executes the `llvm-config` command and returns the `stdout` output if the
/// command was successfully executed (errors are added to `COMMAND_ERRORS`).
pub fn run_llvm_config(arguments: &[&str]) -> Option<String> {
let path = env::var("LLVM_CONFIG_PATH").unwrap_or_else(|_| "llvm-config".into());
run_command("llvm-config", &path, arguments)
}
/// Executes the `xcode-select` command and returns the `stdout` output if the
/// command was successfully executed (errors are added to `COMMAND_ERRORS`).
pub fn run_xcode_select(arguments: &[&str]) -> Option<String> {
run_command("xcode-select", "xcode-select", arguments)
}
//================================================
// Search Directories
//================================================
/// `libclang` directory patterns for Haiku.
const DIRECTORIES_HAIKU: &[&str] = &[
"/boot/system/lib",
"/boot/system/develop/lib",
"/boot/system/non-packaged/lib",
"/boot/system/non-packaged/develop/lib",
"/boot/home/config/non-packaged/lib",
"/boot/home/config/non-packaged/develop/lib",
];
/// `libclang` directory patterns for Linux (and FreeBSD).
const DIRECTORIES_LINUX: &[&str] = &[
"/usr/lib*",
"/usr/lib*/*",
"/usr/lib*/*/*",
"/usr/local/lib*",
"/usr/local/lib*/*",
"/usr/local/lib*/*/*",
"/usr/local/llvm*/lib*",
];
/// `libclang` directory patterns for macOS.
const DIRECTORIES_MACOS: &[&str] = &[
"/usr/local/opt/llvm*/lib",
"/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/lib",
"/Library/Developer/CommandLineTools/usr/lib",
"/usr/local/opt/llvm*/lib/llvm*/lib",
];
/// `libclang` directory patterns for Windows.
const DIRECTORIES_WINDOWS: &[&str] = &[
"C:\\LLVM\\lib",
"C:\\Program Files*\\LLVM\\lib",
"C:\\MSYS*\\MinGW*\\lib",
// LLVM + Clang can be installed as a component of Visual Studio.
// https://github.com/KyleMayes/clang-sys/issues/121
"C:\\Program Files*\\Microsoft Visual Studio\\*\\BuildTools\\VC\\Tools\\Llvm\\**\\bin",
// LLVM + Clang can be installed using Scoop (https://scoop.sh).
// Other Windows package managers install LLVM + Clang to previously listed
// system-wide directories.
"C:\\Users\\*\\scoop\\apps\\llvm\\current\\bin",
];
/// `libclang` directory patterns for illumos
const DIRECTORIES_ILLUMOS: &[&str] = &[
"/opt/ooce/clang-*/lib",
"/opt/ooce/llvm-*/lib",
];
//================================================
// Searching
//================================================
/// Finds the files in a directory that match one or more filename glob patterns
/// and returns the paths to and filenames of those files.
fn search_directory(directory: &Path, filenames: &[String]) -> Vec<(PathBuf, String)> {
// Escape the specified directory in case it contains characters that have
// special meaning in glob patterns (e.g., `[` or `]`).
let directory = Pattern::escape(directory.to_str().unwrap());
let directory = Path::new(&directory);
// Join the escaped directory to the filename glob patterns to obtain
// complete glob patterns for the files being searched for.
let paths = filenames
.iter()
.map(|f| directory.join(f).to_str().unwrap().to_owned());
// Prevent wildcards from matching path separators to ensure that the search
// is limited to the specified directory.
let mut options = MatchOptions::new();
options.require_literal_separator = true;
paths
.map(|p| glob::glob_with(&p, options))
.filter_map(Result::ok)
.flatten()
.filter_map(|p| {
let path = p.ok()?;
let filename = path.file_name()?.to_str().unwrap();
// The `libclang_shared` library has been renamed to `libclang-cpp`
// in Clang 10. This can cause instances of this library (e.g.,
// `libclang-cpp.so.10`) to be matched by patterns looking for
// instances of `libclang`.
if filename.contains("-cpp.") {
return None;
}
Some((directory.to_owned(), filename.into()))
})
.collect::<Vec<_>>()
}
/// Finds the files in a directory (and any relevant sibling directories) that
/// match one or more filename glob patterns and returns the paths to and
/// filenames of those files.
fn search_directories(directory: &Path, filenames: &[String]) -> Vec<(PathBuf, String)> {
let mut results = search_directory(directory, filenames);
// On Windows, `libclang.dll` is usually found in the LLVM `bin` directory
// while `libclang.lib` is usually found in the LLVM `lib` directory. To
// keep things consistent with other platforms, only LLVM `lib` directories
// are included in the backup search directory globs so we need to search
// the LLVM `bin` directory here.
if cfg!(target_os = "windows") && directory.ends_with("lib") {
let sibling = directory.parent().unwrap().join("bin");
results.extend(search_directory(&sibling, filenames).into_iter());
}
results
}
/// Finds the `libclang` static or dynamic libraries matching one or more
/// filename glob patterns and returns the paths to and filenames of those files.
pub fn search_libclang_directories(filenames: &[String], variable: &str) -> Vec<(PathBuf, String)> {
// Search only the path indicated by the relevant environment variable
// (e.g., `LIBCLANG_PATH`) if it is set.
if let Ok(path) = env::var(variable).map(|d| Path::new(&d).to_path_buf()) {
// Check if the path is a matching file.
if let Some(parent) = path.parent() {
let filename = path.file_name().unwrap().to_str().unwrap();
let libraries = search_directories(parent, filenames);
if libraries.iter().any(|(_, f)| f == filename) {
return vec![(parent.into(), filename.into())];
}
}
// Check if the path is directory containing a matching file.
return search_directories(&path, filenames);
}
let mut found = vec![];
// Search the `bin` and `lib` directories in the directory returned by
// `llvm-config --prefix`.
if let Some(output) = run_llvm_config(&["--prefix"]) {
let directory = Path::new(output.lines().next().unwrap()).to_path_buf();
found.extend(search_directories(&directory.join("bin"), filenames));
found.extend(search_directories(&directory.join("lib"), filenames));
found.extend(search_directories(&directory.join("lib64"), filenames));
}
// Search the toolchain directory in the directory returned by
// `xcode-select --print-path`.
if cfg!(target_os = "macos") {
if let Some(output) = run_xcode_select(&["--print-path"]) {
let directory = Path::new(output.lines().next().unwrap()).to_path_buf();
let directory = directory.join("Toolchains/XcodeDefault.xctoolchain/usr/lib");
found.extend(search_directories(&directory, filenames));
}
}
// Search the directories in the `LD_LIBRARY_PATH` environment variable.
if let Ok(path) = env::var("LD_LIBRARY_PATH") {
for directory in env::split_paths(&path) {
found.extend(search_directories(&directory, filenames));
}
}
// Determine the `libclang` directory patterns.
let directories = if cfg!(target_os = "haiku") {
DIRECTORIES_HAIKU
} else if cfg!(any(target_os = "linux", target_os = "freebsd")) {
DIRECTORIES_LINUX
} else if cfg!(target_os = "macos") {
DIRECTORIES_MACOS
} else if cfg!(target_os = "windows") {
DIRECTORIES_WINDOWS
} else if cfg!(target_os = "illumos") {
DIRECTORIES_ILLUMOS
} else {
&[]
};
// Search the directories provided by the `libclang` directory patterns.
let mut options = MatchOptions::new();
options.case_sensitive = false;
options.require_literal_separator = true;
for directory in directories.iter().rev() {
if let Ok(directories) = glob::glob_with(directory, options) {
for directory in directories.filter_map(Result::ok).filter(|p| p.is_dir()) {
found.extend(search_directories(&directory, filenames));
}
}
}
found
}

View File

@@ -0,0 +1,262 @@
// SPDX-License-Identifier: Apache-2.0
use std::env;
use std::fs::File;
use std::io::{self, Error, ErrorKind, Read, Seek, SeekFrom};
use std::path::{Path, PathBuf};
use super::common;
//================================================
// Validation
//================================================
/// Extracts the ELF class from the ELF header in a shared library.
fn parse_elf_header(path: &Path) -> io::Result<u8> {
let mut file = File::open(path)?;
let mut buffer = [0; 5];
file.read_exact(&mut buffer)?;
if buffer[..4] == [127, 69, 76, 70] {
Ok(buffer[4])
} else {
Err(Error::new(ErrorKind::InvalidData, "invalid ELF header"))
}
}
/// Extracts the magic number from the PE header in a shared library.
fn parse_pe_header(path: &Path) -> io::Result<u16> {
let mut file = File::open(path)?;
// Extract the header offset.
let mut buffer = [0; 4];
let start = SeekFrom::Start(0x3C);
file.seek(start)?;
file.read_exact(&mut buffer)?;
let offset = i32::from_le_bytes(buffer);
// Check the validity of the header.
file.seek(SeekFrom::Start(offset as u64))?;
file.read_exact(&mut buffer)?;
if buffer != [80, 69, 0, 0] {
return Err(Error::new(ErrorKind::InvalidData, "invalid PE header"));
}
// Extract the magic number.
let mut buffer = [0; 2];
file.seek(SeekFrom::Current(20))?;
file.read_exact(&mut buffer)?;
Ok(u16::from_le_bytes(buffer))
}
/// Checks that a `libclang` shared library matches the target platform.
fn validate_library(path: &Path) -> Result<(), String> {
if cfg!(any(target_os = "linux", target_os = "freebsd")) {
let class = parse_elf_header(path).map_err(|e| e.to_string())?;
if cfg!(target_pointer_width = "32") && class != 1 {
return Err("invalid ELF class (64-bit)".into());
}
if cfg!(target_pointer_width = "64") && class != 2 {
return Err("invalid ELF class (32-bit)".into());
}
Ok(())
} else if cfg!(target_os = "windows") {
let magic = parse_pe_header(path).map_err(|e| e.to_string())?;
if cfg!(target_pointer_width = "32") && magic != 267 {
return Err("invalid DLL (64-bit)".into());
}
if cfg!(target_pointer_width = "64") && magic != 523 {
return Err("invalid DLL (32-bit)".into());
}
Ok(())
} else {
Ok(())
}
}
//================================================
// Searching
//================================================
/// Extracts the version components in a `libclang` shared library filename.
fn parse_version(filename: &str) -> Vec<u32> {
let version = if let Some(version) = filename.strip_prefix("libclang.so.") {
version
} else if filename.starts_with("libclang-") {
&filename[9..filename.len() - 3]
} else {
return vec![];
};
version.split('.').map(|s| s.parse().unwrap_or(0)).collect()
}
/// Finds `libclang` shared libraries and returns the paths to, filenames of,
/// and versions of those shared libraries.
fn search_libclang_directories(runtime: bool) -> Result<Vec<(PathBuf, String, Vec<u32>)>, String> {
let mut files = vec![format!(
"{}clang{}",
env::consts::DLL_PREFIX,
env::consts::DLL_SUFFIX
)];
if cfg!(target_os = "linux") {
// Some Linux distributions don't create a `libclang.so` symlink, so we
// need to look for versioned files (e.g., `libclang-3.9.so`).
files.push("libclang-*.so".into());
// Some Linux distributions don't create a `libclang.so` symlink and
// don't have versioned files as described above, so we need to look for
// suffix versioned files (e.g., `libclang.so.1`). However, `ld` cannot
// link to these files, so this will only be included when linking at
// runtime.
if runtime {
files.push("libclang.so.*".into());
files.push("libclang-*.so.*".into());
}
}
if cfg!(any(
target_os = "freebsd",
target_os = "haiku",
target_os = "netbsd",
target_os = "openbsd",
)) {
// Some BSD distributions don't create a `libclang.so` symlink either,
// but use a different naming scheme for versioned files (e.g.,
// `libclang.so.7.0`).
files.push("libclang.so.*".into());
}
if cfg!(target_os = "windows") {
// The official LLVM build uses `libclang.dll` on Windows instead of
// `clang.dll`. However, unofficial builds such as MinGW use `clang.dll`.
files.push("libclang.dll".into());
}
// Find and validate `libclang` shared libraries and collect the versions.
let mut valid = vec![];
let mut invalid = vec![];
for (directory, filename) in common::search_libclang_directories(&files, "LIBCLANG_PATH") {
let path = directory.join(&filename);
match validate_library(&path) {
Ok(()) => {
let version = parse_version(&filename);
valid.push((directory, filename, version))
}
Err(message) => invalid.push(format!("({}: {})", path.display(), message)),
}
}
if !valid.is_empty() {
return Ok(valid);
}
let message = format!(
"couldn't find any valid shared libraries matching: [{}], set the \
`LIBCLANG_PATH` environment variable to a path where one of these files \
can be found (invalid: [{}])",
files
.iter()
.map(|f| format!("'{}'", f))
.collect::<Vec<_>>()
.join(", "),
invalid.join(", "),
);
Err(message)
}
/// Finds the "best" `libclang` shared library and returns the directory and
/// filename of that library.
pub fn find(runtime: bool) -> Result<(PathBuf, String), String> {
search_libclang_directories(runtime)?
.iter()
// We want to find the `libclang` shared library with the highest
// version number, hence `max_by_key` below.
//
// However, in the case where there are multiple such `libclang` shared
// libraries, we want to use the order in which they appeared in the
// list returned by `search_libclang_directories` as a tiebreaker since
// that function returns `libclang` shared libraries in descending order
// of preference by how they were found.
//
// `max_by_key`, perhaps surprisingly, returns the *last* item with the
// maximum key rather than the first which results in the opposite of
// the tiebreaking behavior we want. This is easily fixed by reversing
// the list first.
.rev()
.max_by_key(|f| &f.2)
.cloned()
.map(|(path, filename, _)| (path, filename))
.ok_or_else(|| "unreachable".into())
}
//================================================
// Linking
//================================================
/// Finds and links to a `libclang` shared library.
#[cfg(not(feature = "runtime"))]
pub fn link() {
let cep = common::CommandErrorPrinter::default();
use std::fs;
let (directory, filename) = find(false).unwrap();
println!("cargo:rustc-link-search={}", directory.display());
if cfg!(all(target_os = "windows", target_env = "msvc")) {
// Find the `libclang` stub static library required for the MSVC
// toolchain.
let lib = if !directory.ends_with("bin") {
directory
} else {
directory.parent().unwrap().join("lib")
};
if lib.join("libclang.lib").exists() {
println!("cargo:rustc-link-search={}", lib.display());
} else if lib.join("libclang.dll.a").exists() {
// MSYS and MinGW use `libclang.dll.a` instead of `libclang.lib`.
// It is linkable with the MSVC linker, but Rust doesn't recognize
// the `.a` suffix, so we need to copy it with a different name.
//
// FIXME: Maybe we can just hardlink or symlink it?
let out = env::var("OUT_DIR").unwrap();
fs::copy(
lib.join("libclang.dll.a"),
Path::new(&out).join("libclang.lib"),
)
.unwrap();
println!("cargo:rustc-link-search=native={}", out);
} else {
panic!(
"using '{}', so 'libclang.lib' or 'libclang.dll.a' must be \
available in {}",
filename,
lib.display(),
);
}
println!("cargo:rustc-link-lib=dylib=libclang");
} else {
let name = filename.trim_start_matches("lib");
// Strip extensions and trailing version numbers (e.g., the `.so.7.0` in
// `libclang.so.7.0`).
let name = match name.find(".dylib").or_else(|| name.find(".so")) {
Some(index) => &name[0..index],
None => name,
};
println!("cargo:rustc-link-lib=dylib={}", name);
}
cep.discard();
}

View File

@@ -0,0 +1,140 @@
// SPDX-License-Identifier: Apache-2.0
extern crate glob;
use std::path::{Path, PathBuf};
use glob::Pattern;
use common;
//================================================
// Searching
//================================================
/// Clang static libraries required to link to `libclang` 3.5 and later.
const CLANG_LIBRARIES: &[&str] = &[
"clang",
"clangAST",
"clangAnalysis",
"clangBasic",
"clangDriver",
"clangEdit",
"clangFrontend",
"clangIndex",
"clangLex",
"clangParse",
"clangRewrite",
"clangSema",
"clangSerialization",
];
/// Gets the name of an LLVM or Clang static library from a path.
fn get_library_name(path: &Path) -> Option<String> {
path.file_stem().map(|p| {
let string = p.to_string_lossy();
if let Some(name) = string.strip_prefix("lib") {
name.to_owned()
} else {
string.to_string()
}
})
}
/// Gets the LLVM static libraries required to link to `libclang`.
fn get_llvm_libraries() -> Vec<String> {
common::run_llvm_config(&["--libs"])
.unwrap()
.split_whitespace()
.filter_map(|p| {
// Depending on the version of `llvm-config` in use, listed
// libraries may be in one of two forms, a full path to the library
// or simply prefixed with `-l`.
if let Some(path) = p.strip_prefix("-l") {
Some(path.into())
} else {
get_library_name(Path::new(p))
}
})
.collect()
}
/// Gets the Clang static libraries required to link to `libclang`.
fn get_clang_libraries<P: AsRef<Path>>(directory: P) -> Vec<String> {
// Escape the directory in case it contains characters that have special
// meaning in glob patterns (e.g., `[` or `]`).
let directory = Pattern::escape(directory.as_ref().to_str().unwrap());
let directory = Path::new(&directory);
let pattern = directory.join("libclang*.a").to_str().unwrap().to_owned();
if let Ok(libraries) = glob::glob(&pattern) {
libraries
.filter_map(|l| l.ok().and_then(|l| get_library_name(&l)))
.collect()
} else {
CLANG_LIBRARIES.iter().map(|l| (*l).to_string()).collect()
}
}
/// Finds a directory containing LLVM and Clang static libraries and returns the
/// path to that directory.
fn find() -> PathBuf {
let name = if cfg!(target_os = "windows") {
"libclang.lib"
} else {
"libclang.a"
};
let files = common::search_libclang_directories(&[name.into()], "LIBCLANG_STATIC_PATH");
if let Some((directory, _)) = files.into_iter().next() {
directory
} else {
panic!("could not find any static libraries");
}
}
//================================================
// Linking
//================================================
/// Finds and links to `libclang` static libraries.
pub fn link() {
let cep = common::CommandErrorPrinter::default();
let directory = find();
// Specify required Clang static libraries.
println!("cargo:rustc-link-search=native={}", directory.display());
for library in get_clang_libraries(directory) {
println!("cargo:rustc-link-lib=static={}", library);
}
// Determine the shared mode used by LLVM.
let mode = common::run_llvm_config(&["--shared-mode"]).map(|m| m.trim().to_owned());
let prefix = if mode.map_or(false, |m| m == "static") {
"static="
} else {
""
};
// Specify required LLVM static libraries.
println!(
"cargo:rustc-link-search=native={}",
common::run_llvm_config(&["--libdir"]).unwrap().trim_end()
);
for library in get_llvm_libraries() {
println!("cargo:rustc-link-lib={}{}", prefix, library);
}
// Specify required system libraries.
// MSVC doesn't need this, as it tracks dependencies inside `.lib` files.
if cfg!(target_os = "freebsd") {
println!("cargo:rustc-flags=-l ffi -l ncursesw -l c++ -l z");
} else if cfg!(any(target_os = "haiku", target_os = "linux")) {
println!("cargo:rustc-flags=-l ffi -l ncursesw -l stdc++ -l z");
} else if cfg!(target_os = "macos") {
println!("cargo:rustc-flags=-l ffi -l ncurses -l c++ -l z");
}
cep.discard();
}