rust/src/librustc/metadata/creader.rs

451 lines
16 KiB
Rust
Raw Normal View History

// Copyright 2012-2014 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
#![allow(non_camel_case_types)]
//! Validates all used crates and extern libraries and loads their metadata
use back::link;
use back::svh::Svh;
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 16:03:29 -06:00
use driver::session::Session;
use driver::{driver, config};
use metadata::cstore;
use metadata::cstore::CStore;
use metadata::decoder;
use metadata::loader;
use metadata::loader::CratePaths;
2014-03-27 12:28:38 -05:00
use std::rc::Rc;
use collections::HashMap;
use syntax::ast;
use syntax::abi;
2012-09-04 13:54:36 -05:00
use syntax::attr;
use syntax::attr::AttrMetaMethods;
use syntax::codemap::{Span};
use syntax::diagnostic::SpanHandler;
2013-12-25 12:10:33 -06:00
use syntax::ext::base::{CrateLoader, MacroCrate};
use syntax::parse::token::InternedString;
use syntax::parse::token;
2013-12-28 11:16:48 -06:00
use syntax::crateid::CrateId;
use syntax::visit;
struct Env<'a> {
sess: &'a Session,
next_crate_num: ast::CrateNum,
}
// Traverses an AST, reading all the information about use'd crates and extern
2011-07-07 23:03:09 -05:00
// libraries necessary for later resolving, typechecking, linking, etc.
2014-03-05 08:36:01 -06:00
pub fn read_crates(sess: &Session,
krate: &ast::Crate) {
2013-12-21 18:46:55 -06:00
let mut e = Env {
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 16:03:29 -06:00
sess: sess,
next_crate_num: sess.cstore.next_crate_num(),
};
visit_crate(&e, krate);
2014-03-05 08:36:01 -06:00
visit::walk_crate(&mut e, krate, ());
dump_crates(&sess.cstore);
warn_if_multiple_versions(sess.diagnostic(), &sess.cstore)
2013-12-21 18:46:55 -06:00
}
2014-03-05 08:36:01 -06:00
impl<'a> visit::Visitor<()> for Env<'a> {
fn visit_view_item(&mut self, a: &ast::ViewItem, _: ()) {
2014-03-05 08:36:01 -06:00
visit_view_item(self, a);
visit::walk_view_item(self, a, ());
}
fn visit_item(&mut self, a: &ast::Item, _: ()) {
2014-03-05 08:36:01 -06:00
visit_item(self, a);
visit::walk_item(self, a, ());
}
}
fn dump_crates(cstore: &CStore) {
debug!("resolved crates:");
cstore.iter_crate_data(|_, data| {
debug!("crate_id: {}", data.crate_id());
debug!(" cnum: {}", data.cnum);
debug!(" hash: {}", data.hash());
})
2012-04-09 17:06:38 -05:00
}
fn warn_if_multiple_versions(diag: &SpanHandler, cstore: &CStore) {
let mut map = HashMap::new();
cstore.iter_crate_data(|cnum, data| {
let crateid = data.crate_id();
let key = (crateid.name.clone(), crateid.path.clone());
map.find_or_insert_with(key, |_| Vec::new()).push(cnum);
});
for ((name, _), dupes) in map.move_iter() {
if dupes.len() == 1 { continue }
diag.handler().warn(
format!("using multiple versions of crate `{}`", name));
for dupe in dupes.move_iter() {
let data = cstore.get_crate_data(dupe);
diag.span_note(data.span, "used here");
loader::note_crateid_attr(diag, &data.crate_id());
}
}
2011-07-07 23:03:09 -05:00
}
fn visit_crate(e: &Env, c: &ast::Crate) {
for a in c.attrs.iter().filter(|m| m.name().equiv(&("link_args"))) {
match a.value_str() {
2014-03-09 08:20:44 -05:00
Some(ref linkarg) => e.sess.cstore.add_used_link_args(linkarg.get()),
None => { /* fallthrough */ }
}
}
}
fn should_link(i: &ast::ViewItem) -> bool {
i.attrs.iter().all(|attr| {
2014-01-21 12:08:10 -06:00
attr.name().get() != "phase" ||
2013-12-25 12:10:33 -06:00
attr.meta_item_list().map_or(false, |phases| {
attr::contains_name(phases.as_slice(), "link")
2013-12-25 12:10:33 -06:00
})
})
}
2013-12-25 12:10:33 -06:00
fn visit_view_item(e: &mut Env, i: &ast::ViewItem) {
if !should_link(i) {
2013-12-25 12:10:33 -06:00
return;
}
match extract_crate_info(e, i) {
2013-12-25 12:10:33 -06:00
Some(info) => {
let (cnum, _, _) = resolve_crate(e,
&None,
info.ident.as_slice(),
&info.crate_id,
None,
i.span);
2013-12-25 12:10:33 -06:00
e.sess.cstore.add_extern_mod_stmt_cnum(info.id, cnum);
}
None => ()
}
}
struct CrateInfo {
ident: StrBuf,
crate_id: CrateId,
2013-12-25 12:10:33 -06:00
id: ast::NodeId,
should_link: bool,
2013-12-25 12:10:33 -06:00
}
fn extract_crate_info(e: &Env, i: &ast::ViewItem) -> Option<CrateInfo> {
match i.node {
ast::ViewItemExternCrate(ident, ref path_opt, id) => {
let ident = token::get_ident(ident);
debug!("resolving extern crate stmt. ident: {:?} path_opt: {:?}",
ident, path_opt);
let crate_id = match *path_opt {
Some((ref path_str, _)) => {
let crateid: Option<CrateId> = from_str(path_str.get());
2013-12-25 12:10:33 -06:00
match crateid {
None => {
e.sess.span_err(i.span, "malformed crate id");
return None
2013-12-25 12:10:33 -06:00
}
Some(id) => id
2013-12-25 12:10:33 -06:00
}
}
None => from_str(ident.get().to_str()).unwrap()
2013-12-25 12:10:33 -06:00
};
Some(CrateInfo {
ident: ident.get().to_strbuf(),
crate_id: crate_id,
id: id,
should_link: should_link(i),
2013-12-25 12:10:33 -06:00
})
}
_ => None
}
2011-07-07 23:37:56 -05:00
}
fn visit_item(e: &Env, i: &ast::Item) {
match i.node {
ast::ItemForeignMod(ref fm) => {
if fm.abi == abi::Rust || fm.abi == abi::RustIntrinsic {
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 16:03:29 -06:00
return;
}
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 16:03:29 -06:00
// First, add all of the custom link_args attributes
let link_args = i.attrs.iter()
.filter_map(|at| if at.name().equiv(&("link_args")) {
Some(at)
} else {
None
})
.collect::<Vec<&ast::Attribute>>();
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 16:03:29 -06:00
for m in link_args.iter() {
match m.value_str() {
2014-03-09 08:20:44 -05:00
Some(linkarg) => e.sess.cstore.add_used_link_args(linkarg.get()),
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 16:03:29 -06:00
None => { /* fallthrough */ }
}
}
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 16:03:29 -06:00
// Next, process all of the #[link(..)]-style arguments
let link_args = i.attrs.iter()
.filter_map(|at| if at.name().equiv(&("link")) {
Some(at)
} else {
None
})
.collect::<Vec<&ast::Attribute>>();
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 16:03:29 -06:00
for m in link_args.iter() {
match m.meta_item_list() {
Some(items) => {
2013-11-28 20:03:38 -06:00
let kind = items.iter().find(|k| {
k.name().equiv(&("kind"))
2013-11-28 20:03:38 -06:00
}).and_then(|a| a.value_str());
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 16:03:29 -06:00
let kind = match kind {
Some(k) => {
if k.equiv(&("static")) {
cstore::NativeStatic
} else if e.sess.targ_cfg.os == abi::OsMacos &&
k.equiv(&("framework")) {
cstore::NativeFramework
} else if k.equiv(&("framework")) {
e.sess.span_err(m.span,
"native frameworks are only available \
on OSX targets");
cstore::NativeUnknown
} else {
e.sess.span_err(m.span,
format!("unknown kind: `{}`", k));
cstore::NativeUnknown
}
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 16:03:29 -06:00
}
None => cstore::NativeUnknown
};
2013-11-28 20:03:38 -06:00
let n = items.iter().find(|n| {
n.name().equiv(&("name"))
2013-11-28 20:03:38 -06:00
}).and_then(|a| a.value_str());
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 16:03:29 -06:00
let n = match n {
Some(n) => n,
None => {
e.sess.span_err(m.span,
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 16:03:29 -06:00
"#[link(...)] specified without \
`name = \"foo\"`");
InternedString::new("foo")
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 16:03:29 -06:00
}
};
if n.get().is_empty() {
e.sess.span_err(m.span, "#[link(name = \"\")] given with empty name");
} else {
e.sess
.cstore
.add_used_library(n.get().to_strbuf(), 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 16:03:29 -06:00
}
None => {}
2013-02-16 11:48:28 -06:00
}
2011-07-27 07:19:39 -05:00
}
2011-07-07 23:37:56 -05:00
}
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 16:03:29 -06:00
_ => { }
2011-07-07 23:37:56 -05:00
}
}
fn existing_match(e: &Env, crate_id: &CrateId,
hash: Option<&Svh>) -> Option<ast::CrateNum> {
let mut ret = None;
e.sess.cstore.iter_crate_data(|cnum, data| {
let other_id = data.crate_id();
if crate_id.matches(&other_id) {
let other_hash = data.hash();
match hash {
Some(hash) if *hash != other_hash => {}
Some(..) | None => { ret = Some(cnum); }
}
2012-06-04 10:03:14 -05:00
}
});
return ret;
}
fn register_crate<'a>(e: &mut Env,
root: &Option<CratePaths>,
ident: &str,
crate_id: &CrateId,
span: Span,
lib: loader::Library)
-> (ast::CrateNum, Rc<cstore::crate_metadata>,
cstore::CrateSource) {
// Claim this crate number and cache it
let cnum = e.next_crate_num;
e.next_crate_num += 1;
// Stash paths for top-most crate locally if necessary.
let crate_paths = if root.is_none() {
Some(CratePaths {
ident: ident.to_strbuf(),
dylib: lib.dylib.clone(),
rlib: lib.rlib.clone(),
})
} else {
None
};
// Maintain a reference to the top most crate.
let root = if root.is_some() { root } else { &crate_paths };
let cnum_map = resolve_crate_deps(e, root, lib.metadata.as_slice(), span);
let loader::Library{ dylib, rlib, metadata } = lib;
let cmeta = Rc::new( cstore::crate_metadata {
name: crate_id.name.to_strbuf(),
data: metadata,
cnum_map: cnum_map,
cnum: cnum,
span: span,
});
let source = cstore::CrateSource {
dylib: dylib,
rlib: rlib,
cnum: cnum,
};
e.sess.cstore.set_crate_data(cnum, cmeta.clone());
e.sess.cstore.add_used_crate_source(source.clone());
(cnum, cmeta, source)
}
fn resolve_crate<'a>(e: &mut Env,
root: &Option<CratePaths>,
ident: &str,
crate_id: &CrateId,
hash: Option<&Svh>,
span: Span)
2014-04-17 07:06:25 -05:00
-> (ast::CrateNum, Rc<cstore::crate_metadata>,
cstore::CrateSource) {
match existing_match(e, crate_id, hash) {
None => {
let id_hash = link::crate_id_hash(crate_id);
let mut load_ctxt = loader::Context {
sess: e.sess,
span: span,
ident: ident,
crate_id: crate_id,
id_hash: id_hash.as_slice(),
hash: hash.map(|a| &*a),
filesearch: e.sess.target_filesearch(),
os: config::cfg_os_to_meta_os(e.sess.targ_cfg.os),
triple: e.sess.targ_cfg.target_strs.target_triple.as_slice(),
root: root,
rejected_via_hash: vec!(),
rejected_via_triple: vec!(),
};
let library = load_ctxt.load_library_crate();
register_crate(e, root, ident, crate_id, span, library)
}
Some(cnum) => (cnum,
e.sess.cstore.get_crate_data(cnum),
e.sess.cstore.get_used_crate_source(cnum).unwrap())
}
}
// Go through the crate metadata and load any crates that it references
fn resolve_crate_deps(e: &mut Env,
root: &Option<CratePaths>,
cdata: &[u8], span : Span)
-> cstore::cnum_map {
debug!("resolving deps of external crate");
// The map from crate numbers in the crate we're resolving to local crate
// numbers
2014-04-22 11:06:43 -05:00
decoder::get_crate_deps(cdata).iter().map(|dep| {
debug!("resolving dep crate {} hash: `{}`", dep.crate_id, dep.hash);
let (local_cnum, _, _) = resolve_crate(e, root,
dep.crate_id.name.as_slice(),
&dep.crate_id,
Some(&dep.hash),
span);
2014-04-22 11:06:43 -05:00
(dep.cnum, local_cnum)
}).collect()
}
2013-12-25 12:10:33 -06:00
2014-03-05 08:36:01 -06:00
pub struct Loader<'a> {
env: Env<'a>,
2013-12-25 12:10:33 -06:00
}
2014-03-05 08:36:01 -06:00
impl<'a> Loader<'a> {
pub fn new(sess: &'a Session) -> Loader<'a> {
2013-12-25 12:10:33 -06:00
Loader {
env: Env {
sess: sess,
next_crate_num: sess.cstore.next_crate_num(),
2013-12-25 12:10:33 -06:00
}
}
}
}
2014-03-05 08:36:01 -06:00
impl<'a> CrateLoader for Loader<'a> {
fn load_crate(&mut self, krate: &ast::ViewItem) -> MacroCrate {
let info = extract_crate_info(&self.env, krate).unwrap();
let target_triple = self.env.sess.targ_cfg.target_strs.target_triple.as_slice();
let is_cross = target_triple != driver::host_triple();
let mut should_link = info.should_link && !is_cross;
let id_hash = link::crate_id_hash(&info.crate_id);
let os = config::get_os(driver::host_triple()).unwrap();
let mut load_ctxt = loader::Context {
sess: self.env.sess,
span: krate.span,
ident: info.ident.as_slice(),
crate_id: &info.crate_id,
id_hash: id_hash.as_slice(),
hash: None,
filesearch: self.env.sess.host_filesearch(),
triple: driver::host_triple(),
os: config::cfg_os_to_meta_os(os),
root: &None,
rejected_via_hash: vec!(),
rejected_via_triple: vec!(),
};
let library = match load_ctxt.maybe_load_library_crate() {
Some (l) => l,
None if is_cross => {
// try loading from target crates (only valid if there are
// no syntax extensions)
load_ctxt.triple = target_triple;
load_ctxt.os = config::cfg_os_to_meta_os(self.env.sess.targ_cfg.os);
load_ctxt.filesearch = self.env.sess.target_filesearch();
let lib = load_ctxt.load_library_crate();
if decoder::get_macro_registrar_fn(lib.metadata.as_slice()).is_some() {
let message = format!("crate `{}` contains a macro_registrar fn but \
only a version for triple `{}` could be found (need {})",
info.ident, target_triple, driver::host_triple());
self.env.sess.span_err(krate.span, message);
// need to abort now because the syntax expansion
// code will shortly attempt to load and execute
// code from the found library.
self.env.sess.abort_if_errors();
}
should_link = info.should_link;
lib
}
None => { load_ctxt.report_load_errs(); unreachable!() },
};
let macros = decoder::get_exported_macros(library.metadata.as_slice());
let registrar = decoder::get_macro_registrar_fn(library.metadata.as_slice()).map(|id| {
decoder::get_symbol(library.metadata.as_slice(), id).to_strbuf()
});
let mc = MacroCrate {
lib: library.dylib.clone(),
macros: macros.move_iter().map(|x| x.to_strbuf()).collect(),
registrar_symbol: registrar,
};
if should_link {
// register crate now to avoid double-reading metadata
register_crate(&mut self.env, &None, info.ident.as_slice(),
&info.crate_id, krate.span, library);
2013-12-25 12:10:33 -06:00
}
mc
2013-12-25 12:10:33 -06:00
}
}