rust/src/librustc/middle/astencode.rs

2114 lines
80 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)]
// FIXME: remove this after snapshot, and Results are handled
#![allow(unused_must_use)]
use metadata::common as c;
use metadata::cstore as cstore;
use driver::session::Session;
use metadata::decoder;
use middle::def;
use metadata::encoder as e;
use middle::freevars::{CaptureMode, freevar_entry};
use middle::freevars;
use middle::region;
use metadata::tydecode;
use metadata::tydecode::{DefIdSource, NominalType, TypeWithId, TypeParameter};
use metadata::tydecode::{RegionParameter};
use metadata::tyencode;
use middle::subst;
use middle::subst::VecPerParamSpace;
use middle::typeck::{MethodCall, MethodCallee, MethodOrigin};
use middle::{ty, typeck};
use util::ppaux::ty_to_string;
2014-01-06 14:00:46 +02:00
use syntax::{ast, ast_map, ast_util, codemap, fold};
use syntax::ast_util::PostExpansionMethod;
use syntax::codemap::Span;
use syntax::fold::Folder;
2013-05-14 17:27:27 -07:00
use syntax::parse::token;
2014-09-07 20:09:06 +03:00
use syntax::ptr::P;
use syntax;
2014-02-26 12:58:41 -05:00
use libc;
2013-11-10 22:46:32 -08:00
use std::io::Seek;
core: Remove the cast module This commit revisits the `cast` module in libcore and libstd, and scrutinizes all functions inside of it. The result was to remove the `cast` module entirely, folding all functionality into the `mem` module. Specifically, this is the fate of each function in the `cast` module. * transmute - This function was moved to `mem`, but it is now marked as #[unstable]. This is due to planned changes to the `transmute` function and how it can be invoked (see the #[unstable] comment). For more information, see RFC 5 and #12898 * transmute_copy - This function was moved to `mem`, with clarification that is is not an error to invoke it with T/U that are different sizes, but rather that it is strongly discouraged. This function is now #[stable] * forget - This function was moved to `mem` and marked #[stable] * bump_box_refcount - This function was removed due to the deprecation of managed boxes as well as its questionable utility. * transmute_mut - This function was previously deprecated, and removed as part of this commit. * transmute_mut_unsafe - This function doesn't serve much of a purpose when it can be achieved with an `as` in safe code, so it was removed. * transmute_lifetime - This function was removed because it is likely a strong indication that code is incorrect in the first place. * transmute_mut_lifetime - This function was removed for the same reasons as `transmute_lifetime` * copy_lifetime - This function was moved to `mem`, but it is marked `#[unstable]` now due to the likelihood of being removed in the future if it is found to not be very useful. * copy_mut_lifetime - This function was also moved to `mem`, but had the same treatment as `copy_lifetime`. * copy_lifetime_vec - This function was removed because it is not used today, and its existence is not necessary with DST (copy_lifetime will suffice). In summary, the cast module was stripped down to these functions, and then the functions were moved to the `mem` module. transmute - #[unstable] transmute_copy - #[stable] forget - #[stable] copy_lifetime - #[unstable] copy_mut_lifetime - #[unstable] [breaking-change]
2014-05-09 10:34:51 -07:00
use std::mem;
use std::gc::GC;
use std::rc::Rc;
2013-10-21 17:11:42 -07:00
2014-07-29 18:27:28 -07:00
use rbml::io::SeekableMemWriter;
use rbml::{reader, writer};
use rbml;
use serialize;
use serialize::{Decodable, Decoder, DecoderHelpers, Encodable};
use serialize::{EncoderHelpers};
#[cfg(test)] use syntax::parse;
#[cfg(test)] use syntax::print::pprust;
2014-09-07 20:09:06 +03:00
struct DecodeContext<'a, 'b, 'tcx: 'a> {
tcx: &'a ty::ctxt<'tcx>,
2014-09-07 20:09:06 +03:00
cdata: &'b cstore::crate_metadata,
from_id_range: ast_util::IdRange,
to_id_range: ast_util::IdRange
}
trait tr {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> Self;
}
trait tr_intern {
2014-09-07 20:09:06 +03:00
fn tr_intern(&self, dcx: &DecodeContext) -> ast::DefId;
}
remove seek from std::io::MemWriter, add SeekableMemWriter to librustc Not all users of MemWriter need to seek, but having MemWriter seekable adds between 3-29% in overhead in certain circumstances. This fixes that performance gap by making a non-seekable MemWriter, and creating a new SeekableMemWriter for those circumstances when that functionality is actually needed. ``` test io::mem::test::bench_buf_reader ... bench: 682 ns/iter (+/- 85) test io::mem::test::bench_buf_writer ... bench: 580 ns/iter (+/- 57) test io::mem::test::bench_mem_reader ... bench: 793 ns/iter (+/- 99) test io::mem::test::bench_mem_writer_001_0000 ... bench: 48 ns/iter (+/- 27) test io::mem::test::bench_mem_writer_001_0010 ... bench: 65 ns/iter (+/- 27) = 153 MB/s test io::mem::test::bench_mem_writer_001_0100 ... bench: 132 ns/iter (+/- 12) = 757 MB/s test io::mem::test::bench_mem_writer_001_1000 ... bench: 802 ns/iter (+/- 151) = 1246 MB/s test io::mem::test::bench_mem_writer_100_0000 ... bench: 481 ns/iter (+/- 28) test io::mem::test::bench_mem_writer_100_0010 ... bench: 1957 ns/iter (+/- 126) = 510 MB/s test io::mem::test::bench_mem_writer_100_0100 ... bench: 8222 ns/iter (+/- 434) = 1216 MB/s test io::mem::test::bench_mem_writer_100_1000 ... bench: 82496 ns/iter (+/- 11191) = 1212 MB/s test io::mem::test::bench_seekable_mem_writer_001_0000 ... bench: 48 ns/iter (+/- 2) test io::mem::test::bench_seekable_mem_writer_001_0010 ... bench: 64 ns/iter (+/- 2) = 156 MB/s test io::mem::test::bench_seekable_mem_writer_001_0100 ... bench: 129 ns/iter (+/- 7) = 775 MB/s test io::mem::test::bench_seekable_mem_writer_001_1000 ... bench: 801 ns/iter (+/- 159) = 1248 MB/s test io::mem::test::bench_seekable_mem_writer_100_0000 ... bench: 711 ns/iter (+/- 51) test io::mem::test::bench_seekable_mem_writer_100_0010 ... bench: 2532 ns/iter (+/- 227) = 394 MB/s test io::mem::test::bench_seekable_mem_writer_100_0100 ... bench: 8962 ns/iter (+/- 947) = 1115 MB/s test io::mem::test::bench_seekable_mem_writer_100_1000 ... bench: 85086 ns/iter (+/- 11555) = 1175 MB/s ``` [breaking-change]
2014-07-29 16:31:39 -07:00
pub type Encoder<'a> = writer::Encoder<'a, SeekableMemWriter>;
// ______________________________________________________________________
// Top-level methods.
pub fn encode_inlined_item(ecx: &e::EncodeContext,
rbml_w: &mut Encoder,
ii: e::InlinedItemRef) {
let id = match ii {
e::IIItemRef(i) => i.id,
e::IIForeignRef(i) => i.id,
2014-09-07 20:09:06 +03:00
e::IITraitItemRef(_, &ast::ProvidedMethod(ref m)) => m.id,
e::IITraitItemRef(_, &ast::RequiredMethod(ref m)) => m.id,
e::IITraitItemRef(_, &ast::TypeTraitItem(ref ti)) => ti.id,
e::IIImplItemRef(_, &ast::MethodImplItem(ref m)) => m.id,
e::IIImplItemRef(_, &ast::TypeImplItem(ref ti)) => ti.id,
2014-01-06 14:00:46 +02:00
};
debug!("> Encoding inlined item: {} ({})",
ecx.tcx.map.path_to_string(id),
rbml_w.writer.tell());
2014-09-07 20:09:06 +03:00
// Folding could be avoided with a smarter encoder.
2014-01-06 14:00:46 +02:00
let ii = simplify_ast(ii);
let id_range = ast_util::compute_id_range_for_inlined_item(&ii);
rbml_w.start_tag(c::tag_ast as uint);
id_range.encode(rbml_w);
2014-09-07 20:09:06 +03:00
encode_ast(rbml_w, &ii);
encode_side_tables_for_ii(ecx, rbml_w, &ii);
rbml_w.end_tag();
debug!("< Encoded inlined fn: {} ({})",
ecx.tcx.map.path_to_string(id),
rbml_w.writer.tell());
}
2014-09-07 20:09:06 +03:00
impl<'a, 'b, 'c, 'tcx> ast_map::FoldOps for &'a DecodeContext<'b, 'c, 'tcx> {
fn new_id(&self, id: ast::NodeId) -> ast::NodeId {
if id == ast::DUMMY_NODE_ID {
// Used by ast_map to map the NodeInlinedParent.
self.tcx.sess.next_node_id()
} else {
self.tr_id(id)
}
}
fn new_def_id(&self, def_id: ast::DefId) -> ast::DefId {
self.tr_def_id(def_id)
}
fn new_span(&self, span: Span) -> Span {
self.tr_span(span)
}
}
pub fn decode_inlined_item<'tcx>(cdata: &cstore::crate_metadata,
tcx: &ty::ctxt<'tcx>,
path: Vec<ast_map::PathElem>,
par_doc: rbml::Doc)
-> Result<&'tcx ast::InlinedItem, Vec<ast_map::PathElem>> {
2012-08-06 12:34:08 -07:00
match par_doc.opt_child(c::tag_ast) {
None => Err(path),
2012-08-20 12:23:37 -07:00
Some(ast_doc) => {
let mut path_as_str = None;
debug!("> Decoding inlined fn: {}::?",
{
// Do an Option dance to use the path after it is moved below.
let s = ast_map::path_to_string(ast_map::Values(path.iter()));
path_as_str = Some(s);
path_as_str.as_ref().map(|x| x.as_slice())
});
let mut ast_dsr = reader::Decoder::new(ast_doc);
2014-03-29 01:05:46 +01:00
let from_id_range = Decodable::decode(&mut ast_dsr).unwrap();
2014-09-07 20:09:06 +03:00
let to_id_range = reserve_id_range(&tcx.sess, from_id_range);
let dcx = &DecodeContext {
cdata: cdata,
tcx: tcx,
from_id_range: from_id_range,
to_id_range: to_id_range
};
let raw_ii = decode_ast(ast_doc);
2014-09-07 20:09:06 +03:00
let ii = ast_map::map_decoded_item(&dcx.tcx.map, path, raw_ii, dcx);
let ident = match *ii {
ast::IIItem(ref i) => i.ident,
ast::IIForeign(ref i) => i.ident,
ast::IITraitItem(_, ref ti) => {
match *ti {
ast::ProvidedMethod(ref m) => m.pe_ident(),
ast::RequiredMethod(ref ty_m) => ty_m.ident,
ast::TypeTraitItem(ref ti) => ti.ident,
2014-09-07 20:09:06 +03:00
}
},
ast::IIImplItem(_, ref m) => {
match *m {
ast::MethodImplItem(ref m) => m.pe_ident(),
ast::TypeImplItem(ref ti) => ti.ident,
}
}
2014-01-06 14:00:46 +02:00
};
debug!("Fn named: {}", token::get_ident(ident));
debug!("< Decoded inlined fn: {}::{}",
path_as_str.unwrap(),
token::get_ident(ident));
2014-09-07 20:09:06 +03:00
region::resolve_inlined_item(&tcx.sess, &tcx.region_maps, ii);
decode_side_tables(dcx, ast_doc);
match *ii {
ast::IIItem(ref i) => {
debug!(">>> DECODED ITEM >>>\n{}\n<<< DECODED ITEM <<<",
2014-09-07 20:09:06 +03:00
syntax::print::pprust::item_to_string(&**i));
}
2012-08-03 19:59:04 -07:00
_ => { }
}
Ok(ii)
}
}
}
// ______________________________________________________________________
// Enumerating the IDs which appear in an AST
2014-03-05 16:36:01 +02:00
fn reserve_id_range(sess: &Session,
from_id_range: ast_util::IdRange) -> ast_util::IdRange {
// Handle the case of an empty range:
2013-03-15 15:24:24 -04:00
if from_id_range.empty() { return from_id_range; }
let cnt = from_id_range.max - from_id_range.min;
let to_id_min = sess.reserve_node_ids(cnt);
let to_id_max = to_id_min + cnt;
ast_util::IdRange { min: to_id_min, max: to_id_max }
}
2014-09-07 20:09:06 +03:00
impl<'a, 'b, 'tcx> DecodeContext<'a, 'b, 'tcx> {
pub fn tr_id(&self, id: ast::NodeId) -> ast::NodeId {
/*!
* Translates an internal id, meaning a node id that is known
* to refer to some part of the item currently being inlined,
* such as a local variable or argument. All naked node-ids
* that appear in types have this property, since if something
* might refer to an external item we would use a def-id to
* allow for the possibility that the item resides in another
* crate.
*/
// from_id_range should be non-empty
2013-03-15 15:24:24 -04:00
assert!(!self.from_id_range.empty());
(id - self.from_id_range.min + self.to_id_range.min)
}
pub fn tr_def_id(&self, did: ast::DefId) -> ast::DefId {
/*!
* Translates an EXTERNAL def-id, converting the crate number
* from the one used in the encoded data to the current crate
* numbers.. By external, I mean that it be translated to a
* reference to the item in its original crate, as opposed to
* being translated to a reference to the inlined version of
* the item. This is typically, but not always, what you
* want, because most def-ids refer to external things like
* types or other fns that may or may not be inlined. Note
* that even when the inlined function is referencing itself
* recursively, we would want `tr_def_id` for that
* reference--- conceptually the function calls the original,
* non-inlined version, and trans deals with linking that
* recursive call to the inlined copy.
*
* However, there are a *few* cases where def-ids are used but
* we know that the thing being referenced is in fact *internal*
* to the item being inlined. In those cases, you should use
* `tr_intern_def_id()` below.
*/
2014-09-07 20:09:06 +03:00
decoder::translate_def_id(self.cdata, did)
}
pub fn tr_intern_def_id(&self, did: ast::DefId) -> ast::DefId {
/*!
* Translates an INTERNAL def-id, meaning a def-id that is
* known to refer to some part of the item currently being
* inlined. In that case, we want to convert the def-id to
* refer to the current crate and to the new, inlined node-id.
*/
assert_eq!(did.krate, ast::LOCAL_CRATE);
ast::DefId { krate: ast::LOCAL_CRATE, node: self.tr_id(did.node) }
}
pub fn tr_span(&self, _span: Span) -> Span {
codemap::DUMMY_SP // FIXME (#1972): handle span properly
}
}
impl tr_intern for ast::DefId {
2014-09-07 20:09:06 +03:00
fn tr_intern(&self, dcx: &DecodeContext) -> ast::DefId {
dcx.tr_intern_def_id(*self)
}
}
impl tr for ast::DefId {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> ast::DefId {
dcx.tr_def_id(*self)
}
}
impl tr for Option<ast::DefId> {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> Option<ast::DefId> {
self.map(|d| dcx.tr_def_id(d))
}
}
impl tr for Span {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> Span {
dcx.tr_span(*self)
}
}
trait def_id_encoder_helpers {
fn emit_def_id(&mut self, did: ast::DefId);
}
impl<S:serialize::Encoder<E>, E> def_id_encoder_helpers for S {
fn emit_def_id(&mut self, did: ast::DefId) {
did.encode(self).ok().unwrap()
}
}
trait def_id_decoder_helpers {
2014-09-07 20:09:06 +03:00
fn read_def_id(&mut self, dcx: &DecodeContext) -> ast::DefId;
fn read_def_id_nodcx(&mut self,
2014-04-17 15:06:25 +03:00
cdata: &cstore::crate_metadata) -> ast::DefId;
}
impl<D:serialize::Decoder<E>, E> def_id_decoder_helpers for D {
2014-09-07 20:09:06 +03:00
fn read_def_id(&mut self, dcx: &DecodeContext) -> ast::DefId {
let did: ast::DefId = Decodable::decode(self).ok().unwrap();
2014-09-07 20:09:06 +03:00
did.tr(dcx)
}
2014-09-07 20:09:06 +03:00
fn read_def_id_nodcx(&mut self,
2014-04-17 15:06:25 +03:00
cdata: &cstore::crate_metadata) -> ast::DefId {
let did: ast::DefId = Decodable::decode(self).ok().unwrap();
decoder::translate_def_id(cdata, did)
}
}
// ______________________________________________________________________
// Encoding and decoding the AST itself
//
// The hard work is done by an autogenerated module astencode_gen. To
// regenerate astencode_gen, run src/etc/gen-astencode. It will
// replace astencode_gen with a dummy file and regenerate its
// contents. If you get compile errors, the dummy file
// remains---resolve the errors and then rerun astencode_gen.
// Annoying, I know, but hopefully only temporary.
//
// When decoding, we have to renumber the AST so that the node ids that
// appear within are disjoint from the node ids in our existing ASTs.
// We also have to adjust the spans: for now we just insert a dummy span,
// but eventually we should add entries to the local codemap as required.
2014-09-07 20:09:06 +03:00
fn encode_ast(rbml_w: &mut Encoder, item: &ast::InlinedItem) {
rbml_w.start_tag(c::tag_tree as uint);
item.encode(rbml_w);
rbml_w.end_tag();
}
2014-01-06 14:00:46 +02:00
struct NestedItemsDropper;
impl Folder for NestedItemsDropper {
2014-09-07 20:09:06 +03:00
fn fold_block(&mut self, blk: P<ast::Block>) -> P<ast::Block> {
blk.and_then(|ast::Block {id, stmts, expr, rules, span, ..}| {
2014-09-14 20:27:36 -07:00
let stmts_sans_items = stmts.into_iter().filter_map(|stmt| {
2014-09-07 20:09:06 +03:00
let use_stmt = match stmt.node {
ast::StmtExpr(_, _) | ast::StmtSemi(_, _) => true,
ast::StmtDecl(ref decl, _) => {
match decl.node {
ast::DeclLocal(_) => true,
ast::DeclItem(_) => false,
}
}
2014-09-07 20:09:06 +03:00
ast::StmtMac(..) => fail!("unexpanded macro in astencode")
};
if use_stmt {
Some(stmt)
} else {
None
}
2014-09-07 20:09:06 +03:00
}).collect();
let blk_sans_items = P(ast::Block {
view_items: Vec::new(), // I don't know if we need the view_items
// here, but it doesn't break tests!
stmts: stmts_sans_items,
expr: expr,
id: id,
rules: rules,
span: span,
});
fold::noop_fold_block(blk_sans_items, self)
})
}
}
// Produces a simplified copy of the AST which does not include things
// that we do not need to or do not want to export. For example, we
// do not include any nested items: if these nested items are to be
// inlined, their AST will be exported separately (this only makes
// sense because, in Rust, nested items are independent except for
// their visibility).
//
// As it happens, trans relies on the fact that we do not export
// nested items, as otherwise it would get confused when translating
// inlined items.
fn simplify_ast(ii: e::InlinedItemRef) -> ast::InlinedItem {
2014-01-06 14:00:46 +02:00
let mut fld = NestedItemsDropper;
2014-01-06 14:00:46 +02:00
match ii {
// HACK we're not dropping items.
e::IIItemRef(i) => {
2014-09-07 20:09:06 +03:00
ast::IIItem(fold::noop_fold_item(P(i.clone()), &mut fld)
.expect_one("expected one item"))
}
2014-09-07 20:09:06 +03:00
e::IITraitItemRef(d, ti) => {
ast::IITraitItem(d, match *ti {
ast::ProvidedMethod(ref m) => {
ast::ProvidedMethod(
fold::noop_fold_method(m.clone(), &mut fld)
.expect_one("noop_fold_method must produce \
exactly one method"))
}
2014-09-07 20:09:06 +03:00
ast::RequiredMethod(ref ty_m) => {
ast::RequiredMethod(
fold::noop_fold_type_method(ty_m.clone(), &mut fld))
}
ast::TypeTraitItem(ref associated_type) => {
ast::TypeTraitItem(
P(fold::noop_fold_associated_type(
(**associated_type).clone(),
&mut fld)))
}
2014-09-07 20:09:06 +03:00
})
}
e::IIImplItemRef(d, m) => {
ast::IIImplItem(d, match *m {
ast::MethodImplItem(ref m) => {
ast::MethodImplItem(
fold::noop_fold_method(m.clone(), &mut fld)
.expect_one("noop_fold_method must produce \
exactly one method"))
}
ast::TypeImplItem(ref td) => {
ast::TypeImplItem(
P(fold::noop_fold_typedef((**td).clone(), &mut fld)))
}
})
}
e::IIForeignRef(i) => {
2014-09-07 20:09:06 +03:00
ast::IIForeign(fold::noop_fold_foreign_item(P(i.clone()), &mut fld))
}
}
}
fn decode_ast(par_doc: rbml::Doc) -> ast::InlinedItem {
let chi_doc = par_doc.get(c::tag_tree as uint);
let mut d = reader::Decoder::new(chi_doc);
2014-03-29 01:05:46 +01:00
Decodable::decode(&mut d).unwrap()
}
// ______________________________________________________________________
// Encoding and decoding of ast::def
2014-09-07 20:09:06 +03:00
fn decode_def(dcx: &DecodeContext, doc: rbml::Doc) -> def::Def {
let mut dsr = reader::Decoder::new(doc);
let def: def::Def = Decodable::decode(&mut dsr).unwrap();
2014-09-07 20:09:06 +03:00
def.tr(dcx)
}
impl tr for def::Def {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> def::Def {
2013-02-22 01:41:37 -05:00
match *self {
2014-09-07 20:09:06 +03:00
def::DefFn(did, p) => def::DefFn(did.tr(dcx), p),
def::DefStaticMethod(did, wrapped_did2, p) => {
2014-09-07 20:09:06 +03:00
def::DefStaticMethod(did.tr(dcx),
match wrapped_did2 {
def::FromTrait(did2) => {
2014-09-07 20:09:06 +03:00
def::FromTrait(did2.tr(dcx))
}
def::FromImpl(did2) => {
2014-09-07 20:09:06 +03:00
def::FromImpl(did2.tr(dcx))
}
},
p)
}
def::DefMethod(did0, did1) => {
2014-09-07 20:09:06 +03:00
def::DefMethod(did0.tr(dcx), did1.map(|did1| did1.tr(dcx)))
}
2014-09-07 20:09:06 +03:00
def::DefSelfTy(nid) => { def::DefSelfTy(dcx.tr_id(nid)) }
def::DefMod(did) => { def::DefMod(did.tr(dcx)) }
def::DefForeignMod(did) => { def::DefForeignMod(did.tr(dcx)) }
def::DefStatic(did, m) => { def::DefStatic(did.tr(dcx), m) }
def::DefArg(nid, b) => { def::DefArg(dcx.tr_id(nid), b) }
def::DefLocal(nid, b) => { def::DefLocal(dcx.tr_id(nid), b) }
def::DefVariant(e_did, v_did, is_s) => {
2014-09-07 20:09:06 +03:00
def::DefVariant(e_did.tr(dcx), v_did.tr(dcx), is_s)
},
2014-09-07 20:09:06 +03:00
def::DefTrait(did) => def::DefTrait(did.tr(dcx)),
def::DefTy(did, is_enum) => def::DefTy(did.tr(dcx), is_enum),
def::DefAssociatedTy(did) => def::DefAssociatedTy(did.tr(dcx)),
def::DefPrimTy(p) => def::DefPrimTy(p),
2014-09-07 20:09:06 +03:00
def::DefTyParam(s, did, v) => def::DefTyParam(s, did.tr(dcx), v),
def::DefBinding(nid, bm) => def::DefBinding(dcx.tr_id(nid), bm),
def::DefUse(did) => def::DefUse(did.tr(dcx)),
def::DefUpvar(nid1, def, nid2, nid3) => {
2014-09-07 20:09:06 +03:00
def::DefUpvar(dcx.tr_id(nid1),
box(GC) (*def).tr(dcx),
dcx.tr_id(nid2),
dcx.tr_id(nid3))
}
2014-09-07 20:09:06 +03:00
def::DefStruct(did) => def::DefStruct(did.tr(dcx)),
def::DefRegion(nid) => def::DefRegion(dcx.tr_id(nid)),
def::DefTyParamBinder(nid) => {
2014-09-07 20:09:06 +03:00
def::DefTyParamBinder(dcx.tr_id(nid))
}
2014-09-07 20:09:06 +03:00
def::DefLabel(nid) => def::DefLabel(dcx.tr_id(nid))
}
}
}
// ______________________________________________________________________
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
// Encoding and decoding of ancillary information
impl tr for ty::Region {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> ty::Region {
2013-02-22 01:41:37 -05:00
match *self {
ty::ReLateBound(id, br) => {
2014-09-07 20:09:06 +03:00
ty::ReLateBound(dcx.tr_id(id), br.tr(dcx))
}
ty::ReEarlyBound(id, space, index, ident) => {
2014-09-07 20:09:06 +03:00
ty::ReEarlyBound(dcx.tr_id(id), space, index, ident)
}
ty::ReScope(id) => {
2014-09-07 20:09:06 +03:00
ty::ReScope(dcx.tr_id(id))
}
ty::ReEmpty | ty::ReStatic | ty::ReInfer(..) => {
*self
}
2013-10-29 10:34:11 -04:00
ty::ReFree(ref fr) => {
2014-09-07 20:09:06 +03:00
ty::ReFree(ty::FreeRegion {scope_id: dcx.tr_id(fr.scope_id),
bound_region: fr.bound_region.tr(dcx)})
}
}
}
}
2013-10-29 10:34:11 -04:00
impl tr for ty::BoundRegion {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> ty::BoundRegion {
2013-02-22 01:41:37 -05:00
match *self {
2013-10-29 10:34:11 -04:00
ty::BrAnon(_) |
ty::BrFresh(_) => *self,
2014-09-07 20:09:06 +03:00
ty::BrNamed(id, ident) => ty::BrNamed(dcx.tr_def_id(id),
ident),
}
}
}
impl tr for ty::TraitStore {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> ty::TraitStore {
match *self {
ty::RegionTraitStore(r, m) => {
2014-09-07 20:09:06 +03:00
ty::RegionTraitStore(r.tr(dcx), m)
}
ty::UniqTraitStore => ty::UniqTraitStore
}
}
}
// ______________________________________________________________________
// Encoding and decoding of freevar information
fn encode_freevar_entry(rbml_w: &mut Encoder, fv: &freevar_entry) {
(*fv).encode(rbml_w).unwrap();
}
fn encode_capture_mode(rbml_w: &mut Encoder, cm: CaptureMode) {
cm.encode(rbml_w).unwrap();
}
trait rbml_decoder_helper {
2014-09-07 20:09:06 +03:00
fn read_freevar_entry(&mut self, dcx: &DecodeContext)
-> freevar_entry;
fn read_capture_mode(&mut self) -> CaptureMode;
}
impl<'a> rbml_decoder_helper for reader::Decoder<'a> {
2014-09-07 20:09:06 +03:00
fn read_freevar_entry(&mut self, dcx: &DecodeContext)
-> freevar_entry {
2014-03-29 01:05:46 +01:00
let fv: freevar_entry = Decodable::decode(self).unwrap();
2014-09-07 20:09:06 +03:00
fv.tr(dcx)
}
fn read_capture_mode(&mut self) -> CaptureMode {
let cm: CaptureMode = Decodable::decode(self).unwrap();
cm
}
}
impl tr for freevar_entry {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> freevar_entry {
freevar_entry {
2014-09-07 20:09:06 +03:00
def: self.def.tr(dcx),
span: self.span.tr(dcx),
}
}
}
impl tr for ty::UpvarBorrow {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> ty::UpvarBorrow {
ty::UpvarBorrow {
kind: self.kind,
2014-09-07 20:09:06 +03:00
region: self.region.tr(dcx)
}
}
}
// ______________________________________________________________________
// Encoding and decoding of MethodCallee
trait read_method_callee_helper {
2014-09-07 20:09:06 +03:00
fn read_method_callee(&mut self, dcx: &DecodeContext)
-> (typeck::ExprAdjustment, MethodCallee);
}
fn encode_method_callee(ecx: &e::EncodeContext,
rbml_w: &mut Encoder,
adjustment: typeck::ExprAdjustment,
method: &MethodCallee) {
use serialize::Encoder;
rbml_w.emit_struct("MethodCallee", 4, |rbml_w| {
rbml_w.emit_struct_field("adjustment", 0u, |rbml_w| {
adjustment.encode(rbml_w)
});
rbml_w.emit_struct_field("origin", 1u, |rbml_w| {
Ok(rbml_w.emit_method_origin(ecx, &method.origin))
});
rbml_w.emit_struct_field("ty", 2u, |rbml_w| {
Ok(rbml_w.emit_ty(ecx, method.ty))
});
rbml_w.emit_struct_field("substs", 3u, |rbml_w| {
Ok(rbml_w.emit_substs(ecx, &method.substs))
})
2014-03-29 01:05:46 +01:00
}).unwrap();
}
impl<'a> read_method_callee_helper for reader::Decoder<'a> {
2014-09-07 20:09:06 +03:00
fn read_method_callee(&mut self, dcx: &DecodeContext)
-> (typeck::ExprAdjustment, MethodCallee) {
2014-03-29 01:05:46 +01:00
self.read_struct("MethodCallee", 4, |this| {
let adjustment = this.read_struct_field("adjustment", 0, |this| {
Decodable::decode(this)
2014-03-29 01:05:46 +01:00
}).unwrap();
Ok((adjustment, MethodCallee {
2014-03-29 01:05:46 +01:00
origin: this.read_struct_field("origin", 1, |this| {
Ok(this.read_method_origin(dcx))
2014-03-29 01:05:46 +01:00
}).unwrap(),
ty: this.read_struct_field("ty", 2, |this| {
2014-09-07 20:09:06 +03:00
Ok(this.read_ty(dcx))
2014-03-29 01:05:46 +01:00
}).unwrap(),
substs: this.read_struct_field("substs", 3, |this| {
2014-09-07 20:09:06 +03:00
Ok(this.read_substs(dcx))
2014-03-29 01:05:46 +01:00
}).unwrap()
}))
2014-03-29 01:05:46 +01:00
}).unwrap()
}
}
impl tr for MethodOrigin {
2014-09-07 20:09:06 +03:00
fn tr(&self, dcx: &DecodeContext) -> MethodOrigin {
2013-02-22 01:41:37 -05:00
match *self {
2014-09-07 20:09:06 +03:00
typeck::MethodStatic(did) => typeck::MethodStatic(did.tr(dcx)),
typeck::MethodStaticUnboxedClosure(did) => {
2014-09-07 20:09:06 +03:00
typeck::MethodStaticUnboxedClosure(did.tr(dcx))
}
typeck::MethodTypeParam(ref mp) => {
typeck::MethodTypeParam(
typeck::MethodParam {
// def-id is already translated when we read it out
trait_ref: mp.trait_ref.clone(),
method_num: mp.method_num,
}
)
}
typeck::MethodTraitObject(ref mo) => {
typeck::MethodTraitObject(
typeck::MethodObject {
trait_ref: mo.trait_ref.clone(),
.. *mo
}
)
}
}
}
}
// ______________________________________________________________________
// Encoding and decoding vtable_res
pub fn encode_vtable_res(ecx: &e::EncodeContext,
rbml_w: &mut Encoder,
dr: &typeck::vtable_res) {
// can't autogenerate this code because automatic code of
// ty::t doesn't work, and there is no way (atm) to have
// hand-written encoding routines combine with auto-generated
// ones. perhaps we should fix this.
encode_vec_per_param_space(
rbml_w, dr,
|rbml_w, param_tables| encode_vtable_param_res(ecx, rbml_w,
param_tables))
}
pub fn encode_vtable_param_res(ecx: &e::EncodeContext,
rbml_w: &mut Encoder,
2014-04-10 14:04:45 +03:00
param_tables: &typeck::vtable_param_res) {
rbml_w.emit_from_vec(param_tables.as_slice(), |rbml_w, vtable_origin| {
Ok(encode_vtable_origin(ecx, rbml_w, vtable_origin))
2014-03-29 01:05:46 +01:00
}).unwrap()
}
pub fn encode_unboxed_closure_kind(ebml_w: &mut Encoder,
kind: ty::UnboxedClosureKind) {
use serialize::Encoder;
ebml_w.emit_enum("UnboxedClosureKind", |ebml_w| {
match kind {
ty::FnUnboxedClosureKind => {
ebml_w.emit_enum_variant("FnUnboxedClosureKind", 0, 3, |_| {
Ok(())
})
}
ty::FnMutUnboxedClosureKind => {
ebml_w.emit_enum_variant("FnMutUnboxedClosureKind", 1, 3, |_| {
Ok(())
})
}
ty::FnOnceUnboxedClosureKind => {
ebml_w.emit_enum_variant("FnOnceUnboxedClosureKind",
2,
3,
|_| {
Ok(())
})
}
}
}).unwrap()
}
pub fn encode_vtable_origin(ecx: &e::EncodeContext,
rbml_w: &mut Encoder,
vtable_origin: &typeck::vtable_origin) {
use serialize::Encoder;
rbml_w.emit_enum("vtable_origin", |rbml_w| {
match *vtable_origin {
typeck::vtable_static(def_id, ref substs, ref vtable_res) => {
rbml_w.emit_enum_variant("vtable_static", 0u, 3u, |rbml_w| {
rbml_w.emit_enum_variant_arg(0u, |rbml_w| {
Ok(rbml_w.emit_def_id(def_id))
});
rbml_w.emit_enum_variant_arg(1u, |rbml_w| {
Ok(rbml_w.emit_substs(ecx, substs))
});
rbml_w.emit_enum_variant_arg(2u, |rbml_w| {
Ok(encode_vtable_res(ecx, rbml_w, vtable_res))
})
})
}
2012-08-03 19:59:04 -07:00
typeck::vtable_param(pn, bn) => {
rbml_w.emit_enum_variant("vtable_param", 1u, 3u, |rbml_w| {
rbml_w.emit_enum_variant_arg(0u, |rbml_w| {
pn.encode(rbml_w)
});
rbml_w.emit_enum_variant_arg(1u, |rbml_w| {
rbml_w.emit_uint(bn)
})
})
}
typeck::vtable_unboxed_closure(def_id) => {
rbml_w.emit_enum_variant("vtable_unboxed_closure",
2u,
1u,
|rbml_w| {
rbml_w.emit_enum_variant_arg(0u, |rbml_w| {
Ok(rbml_w.emit_def_id(def_id))
})
})
}
typeck::vtable_error => {
rbml_w.emit_enum_variant("vtable_error", 3u, 3u, |_rbml_w| {
Ok(())
})
}
}
2014-03-29 01:05:46 +01:00
}).unwrap()
}
pub trait vtable_decoder_helpers {
fn read_vec_per_param_space<T>(&mut self,
f: |&mut Self| -> T)
-> VecPerParamSpace<T>;
fn read_vtable_res_with_key(&mut self,
tcx: &ty::ctxt,
2014-04-17 15:06:25 +03:00
cdata: &cstore::crate_metadata)
-> (typeck::ExprAdjustment, typeck::vtable_res);
fn read_vtable_res(&mut self,
2014-04-17 15:06:25 +03:00
tcx: &ty::ctxt, cdata: &cstore::crate_metadata)
2013-02-22 01:41:37 -05:00
-> typeck::vtable_res;
fn read_vtable_param_res(&mut self,
2014-04-17 15:06:25 +03:00
tcx: &ty::ctxt, cdata: &cstore::crate_metadata)
-> typeck::vtable_param_res;
fn read_vtable_origin(&mut self,
2014-04-17 15:06:25 +03:00
tcx: &ty::ctxt, cdata: &cstore::crate_metadata)
-> typeck::vtable_origin;
}
impl<'a> vtable_decoder_helpers for reader::Decoder<'a> {
fn read_vec_per_param_space<T>(&mut self,
f: |&mut reader::Decoder<'a>| -> T)
-> VecPerParamSpace<T>
{
let types = self.read_to_vec(|this| Ok(f(this))).unwrap();
let selfs = self.read_to_vec(|this| Ok(f(this))).unwrap();
let fns = self.read_to_vec(|this| Ok(f(this))).unwrap();
VecPerParamSpace::new(types, selfs, fns)
}
fn read_vtable_res_with_key(&mut self,
tcx: &ty::ctxt,
2014-04-17 15:06:25 +03:00
cdata: &cstore::crate_metadata)
-> (typeck::ExprAdjustment, typeck::vtable_res) {
2014-03-29 01:05:46 +01:00
self.read_struct("VtableWithKey", 2, |this| {
let adjustment = this.read_struct_field("adjustment", 0, |this| {
Decodable::decode(this)
2014-03-29 01:05:46 +01:00
}).unwrap();
Ok((adjustment, this.read_struct_field("vtable_res", 1, |this| {
2014-03-29 01:05:46 +01:00
Ok(this.read_vtable_res(tcx, cdata))
}).unwrap()))
}).unwrap()
}
fn read_vtable_res(&mut self,
tcx: &ty::ctxt,
cdata: &cstore::crate_metadata)
-> typeck::vtable_res
{
self.read_vec_per_param_space(
|this| this.read_vtable_param_res(tcx, cdata))
}
fn read_vtable_param_res(&mut self,
2014-04-17 15:06:25 +03:00
tcx: &ty::ctxt, cdata: &cstore::crate_metadata)
-> typeck::vtable_param_res {
2014-04-10 14:04:45 +03:00
self.read_to_vec(|this| Ok(this.read_vtable_origin(tcx, cdata)))
2014-09-14 20:27:36 -07:00
.unwrap().into_iter().collect()
}
fn read_vtable_origin(&mut self,
2014-04-17 15:06:25 +03:00
tcx: &ty::ctxt, cdata: &cstore::crate_metadata)
-> typeck::vtable_origin {
2014-03-29 01:05:46 +01:00
self.read_enum("vtable_origin", |this| {
this.read_enum_variant(["vtable_static",
"vtable_param",
"vtable_error",
"vtable_unboxed_closure"],
|this, i| {
2014-03-29 01:05:46 +01:00
Ok(match i {
0 => {
typeck::vtable_static(
2014-03-29 01:05:46 +01:00
this.read_enum_variant_arg(0u, |this| {
2014-09-07 20:09:06 +03:00
Ok(this.read_def_id_nodcx(cdata))
2014-03-29 01:05:46 +01:00
}).unwrap(),
this.read_enum_variant_arg(1u, |this| {
2014-09-07 20:09:06 +03:00
Ok(this.read_substs_nodcx(tcx, cdata))
2014-03-29 01:05:46 +01:00
}).unwrap(),
this.read_enum_variant_arg(2u, |this| {
Ok(this.read_vtable_res(tcx, cdata))
}).unwrap()
)
}
2012-08-21 17:22:45 -07:00
1 => {
typeck::vtable_param(
2014-03-29 01:05:46 +01:00
this.read_enum_variant_arg(0u, |this| {
Decodable::decode(this)
2014-03-29 01:05:46 +01:00
}).unwrap(),
this.read_enum_variant_arg(1u, |this| {
this.read_uint()
2014-03-29 01:05:46 +01:00
}).unwrap()
)
}
2 => {
typeck::vtable_unboxed_closure(
this.read_enum_variant_arg(0u, |this| {
2014-09-07 20:09:06 +03:00
Ok(this.read_def_id_nodcx(cdata))
}).unwrap()
)
}
3 => {
typeck::vtable_error
}
_ => fail!("bad enum variant")
})
})
2014-03-29 01:05:46 +01:00
}).unwrap()
}
}
// ___________________________________________________________________________
//
fn encode_vec_per_param_space<T>(rbml_w: &mut Encoder,
v: &subst::VecPerParamSpace<T>,
f: |&mut Encoder, &T|) {
for &space in subst::ParamSpace::all().iter() {
rbml_w.emit_from_vec(v.get_slice(space),
|rbml_w, n| Ok(f(rbml_w, n))).unwrap();
}
}
// ______________________________________________________________________
// Encoding and decoding the side tables
trait get_ty_str_ctxt<'tcx> {
fn ty_str_ctxt<'a>(&'a self) -> tyencode::ctxt<'a, 'tcx>;
}
impl<'a, 'tcx> get_ty_str_ctxt<'tcx> for e::EncodeContext<'a, 'tcx> {
fn ty_str_ctxt<'a>(&'a self) -> tyencode::ctxt<'a, 'tcx> {
tyencode::ctxt {
diag: self.tcx.sess.diagnostic(),
ds: e::def_to_string,
tcx: self.tcx,
2014-04-22 19:06:43 +03:00
abbrevs: &self.type_abbrevs
}
}
}
trait rbml_writer_helpers {
fn emit_closure_type(&mut self,
ecx: &e::EncodeContext,
closure_type: &ty::ClosureTy);
fn emit_method_origin(&mut self,
ecx: &e::EncodeContext,
method_origin: &typeck::MethodOrigin);
fn emit_ty(&mut self, ecx: &e::EncodeContext, ty: ty::t);
fn emit_tys(&mut self, ecx: &e::EncodeContext, tys: &[ty::t]);
fn emit_type_param_def(&mut self,
ecx: &e::EncodeContext,
type_param_def: &ty::TypeParameterDef);
fn emit_trait_ref(&mut self, ecx: &e::EncodeContext, ty: &ty::TraitRef);
fn emit_polytype(&mut self,
ecx: &e::EncodeContext,
pty: ty::Polytype);
fn emit_substs(&mut self, ecx: &e::EncodeContext, substs: &subst::Substs);
fn emit_existential_bounds(&mut self, ecx: &e::EncodeContext, bounds: &ty::ExistentialBounds);
fn emit_builtin_bounds(&mut self, ecx: &e::EncodeContext, bounds: &ty::BuiltinBounds);
fn emit_auto_adjustment(&mut self, ecx: &e::EncodeContext, adj: &ty::AutoAdjustment);
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
fn emit_autoref(&mut self, ecx: &e::EncodeContext, autoref: &ty::AutoRef);
fn emit_auto_deref_ref(&mut self, ecx: &e::EncodeContext, auto_deref_ref: &ty::AutoDerefRef);
fn emit_unsize_kind(&mut self, ecx: &e::EncodeContext, uk: &ty::UnsizeKind);
}
impl<'a> rbml_writer_helpers for Encoder<'a> {
fn emit_closure_type(&mut self,
ecx: &e::EncodeContext,
closure_type: &ty::ClosureTy) {
self.emit_opaque(|this| {
Ok(e::write_closure_type(ecx, this, closure_type))
});
}
fn emit_method_origin(&mut self,
ecx: &e::EncodeContext,
method_origin: &typeck::MethodOrigin)
{
use serialize::Encoder;
self.emit_enum("MethodOrigin", |this| {
match *method_origin {
typeck::MethodStatic(def_id) => {
this.emit_enum_variant("MethodStatic", 0, 1, |this| {
Ok(this.emit_def_id(def_id))
})
}
typeck::MethodStaticUnboxedClosure(def_id) => {
this.emit_enum_variant("MethodStaticUnboxedClosure", 1, 1, |this| {
Ok(this.emit_def_id(def_id))
})
}
typeck::MethodTypeParam(ref p) => {
this.emit_enum_variant("MethodTypeParam", 2, 1, |this| {
this.emit_struct("MethodParam", 2, |this| {
try!(this.emit_struct_field("trait_ref", 0, |this| {
Ok(this.emit_trait_ref(ecx, &*p.trait_ref))
}));
try!(this.emit_struct_field("method_num", 0, |this| {
this.emit_uint(p.method_num)
}));
Ok(())
})
})
}
typeck::MethodTraitObject(ref o) => {
this.emit_enum_variant("MethodTraitObject", 3, 1, |this| {
this.emit_struct("MethodObject", 2, |this| {
try!(this.emit_struct_field("trait_ref", 0, |this| {
Ok(this.emit_trait_ref(ecx, &*o.trait_ref))
}));
try!(this.emit_struct_field("object_trait_id", 0, |this| {
Ok(this.emit_def_id(o.object_trait_id))
}));
try!(this.emit_struct_field("method_num", 0, |this| {
this.emit_uint(o.method_num)
}));
try!(this.emit_struct_field("real_index", 0, |this| {
this.emit_uint(o.real_index)
}));
Ok(())
})
})
}
}
});
}
fn emit_ty(&mut self, ecx: &e::EncodeContext, ty: ty::t) {
2014-03-29 01:05:46 +01:00
self.emit_opaque(|this| Ok(e::write_type(ecx, this, ty)));
}
fn emit_tys(&mut self, ecx: &e::EncodeContext, tys: &[ty::t]) {
2014-03-29 01:05:46 +01:00
self.emit_from_vec(tys, |this, ty| Ok(this.emit_ty(ecx, *ty)));
}
fn emit_trait_ref(&mut self,
ecx: &e::EncodeContext,
trait_ref: &ty::TraitRef) {
self.emit_opaque(|this| Ok(e::write_trait_ref(ecx, this, trait_ref)));
}
fn emit_type_param_def(&mut self,
ecx: &e::EncodeContext,
type_param_def: &ty::TypeParameterDef) {
self.emit_opaque(|this| {
2014-03-29 01:05:46 +01:00
Ok(tyencode::enc_type_param_def(this.writer,
&ecx.ty_str_ctxt(),
type_param_def))
});
}
fn emit_polytype(&mut self,
ecx: &e::EncodeContext,
pty: ty::Polytype) {
use serialize::Encoder;
self.emit_struct("Polytype", 2, |this| {
this.emit_struct_field("generics", 0, |this| {
this.emit_struct("Generics", 2, |this| {
this.emit_struct_field("types", 0, |this| {
Ok(encode_vec_per_param_space(
this, &pty.generics.types,
|this, def| this.emit_type_param_def(ecx, def)))
});
this.emit_struct_field("regions", 1, |this| {
Ok(encode_vec_per_param_space(
this, &pty.generics.regions,
|this, def| def.encode(this).unwrap()))
})
})
});
this.emit_struct_field("ty", 1, |this| {
Ok(this.emit_ty(ecx, pty.ty))
})
});
}
fn emit_existential_bounds(&mut self, ecx: &e::EncodeContext, bounds: &ty::ExistentialBounds) {
self.emit_opaque(|this| Ok(tyencode::enc_existential_bounds(this.writer,
&ecx.ty_str_ctxt(),
bounds)));
}
fn emit_builtin_bounds(&mut self, ecx: &e::EncodeContext, bounds: &ty::BuiltinBounds) {
self.emit_opaque(|this| Ok(tyencode::enc_builtin_bounds(this.writer,
&ecx.ty_str_ctxt(),
bounds)));
}
fn emit_substs(&mut self, ecx: &e::EncodeContext, substs: &subst::Substs) {
2014-03-29 01:05:46 +01:00
self.emit_opaque(|this| Ok(tyencode::enc_substs(this.writer,
&ecx.ty_str_ctxt(),
substs)));
}
fn emit_auto_adjustment(&mut self, ecx: &e::EncodeContext, adj: &ty::AutoAdjustment) {
use serialize::Encoder;
self.emit_enum("AutoAdjustment", |this| {
match *adj {
ty::AdjustAddEnv(store) => {
this.emit_enum_variant("AutoAddEnv", 0, 1, |this| {
this.emit_enum_variant_arg(0, |this| store.encode(this))
})
}
ty::AdjustDerefRef(ref auto_deref_ref) => {
this.emit_enum_variant("AutoDerefRef", 1, 1, |this| {
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
this.emit_enum_variant_arg(0,
|this| Ok(this.emit_auto_deref_ref(ecx, auto_deref_ref)))
})
}
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
}
});
}
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
fn emit_autoref(&mut self, ecx: &e::EncodeContext, autoref: &ty::AutoRef) {
2014-08-06 11:59:40 +02:00
use serialize::Encoder;
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
self.emit_enum("AutoRef", |this| {
match autoref {
&ty::AutoPtr(r, m, None) => {
this.emit_enum_variant("AutoPtr", 0, 3, |this| {
this.emit_enum_variant_arg(0, |this| r.encode(this));
this.emit_enum_variant_arg(1, |this| m.encode(this));
this.emit_enum_variant_arg(2,
|this| this.emit_option(|this| this.emit_option_none()))
})
}
&ty::AutoPtr(r, m, Some(box ref a)) => {
this.emit_enum_variant("AutoPtr", 0, 3, |this| {
this.emit_enum_variant_arg(0, |this| r.encode(this));
this.emit_enum_variant_arg(1, |this| m.encode(this));
this.emit_enum_variant_arg(2, |this| this.emit_option(
|this| this.emit_option_some(|this| Ok(this.emit_autoref(ecx, a)))))
})
}
&ty::AutoUnsize(ref uk) => {
this.emit_enum_variant("AutoUnsize", 1, 1, |this| {
this.emit_enum_variant_arg(0, |this| Ok(this.emit_unsize_kind(ecx, uk)))
})
}
&ty::AutoUnsizeUniq(ref uk) => {
this.emit_enum_variant("AutoUnsizeUniq", 2, 1, |this| {
this.emit_enum_variant_arg(0, |this| Ok(this.emit_unsize_kind(ecx, uk)))
})
}
&ty::AutoUnsafe(m, None) => {
this.emit_enum_variant("AutoUnsafe", 3, 2, |this| {
this.emit_enum_variant_arg(0, |this| m.encode(this));
this.emit_enum_variant_arg(1,
|this| this.emit_option(|this| this.emit_option_none()))
})
}
&ty::AutoUnsafe(m, Some(box ref a)) => {
this.emit_enum_variant("AutoUnsafe", 3, 2, |this| {
this.emit_enum_variant_arg(0, |this| m.encode(this));
this.emit_enum_variant_arg(1, |this| this.emit_option(
|this| this.emit_option_some(|this| Ok(this.emit_autoref(ecx, a)))))
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
})
}
}
});
}
fn emit_auto_deref_ref(&mut self, ecx: &e::EncodeContext, auto_deref_ref: &ty::AutoDerefRef) {
2014-08-06 11:59:40 +02:00
use serialize::Encoder;
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
self.emit_struct("AutoDerefRef", 2, |this| {
this.emit_struct_field("autoderefs", 0, |this| auto_deref_ref.autoderefs.encode(this));
this.emit_struct_field("autoref", 1, |this| {
this.emit_option(|this| {
match auto_deref_ref.autoref {
None => this.emit_option_none(),
Some(ref a) => this.emit_option_some(|this| Ok(this.emit_autoref(ecx, a))),
}
})
})
});
}
fn emit_unsize_kind(&mut self, ecx: &e::EncodeContext, uk: &ty::UnsizeKind) {
2014-08-06 11:59:40 +02:00
use serialize::Encoder;
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
self.emit_enum("UnsizeKind", |this| {
match *uk {
ty::UnsizeLength(len) => {
this.emit_enum_variant("UnsizeLength", 0, 1, |this| {
this.emit_enum_variant_arg(0, |this| len.encode(this))
})
}
ty::UnsizeStruct(box ref uk, idx) => {
this.emit_enum_variant("UnsizeStruct", 1, 2, |this| {
this.emit_enum_variant_arg(0, |this| Ok(this.emit_unsize_kind(ecx, uk)));
this.emit_enum_variant_arg(1, |this| idx.encode(this))
})
}
ty::UnsizeVtable(ty::TyTrait { def_id: def_id,
bounds: ref b,
substs: ref substs },
self_ty) => {
this.emit_enum_variant("UnsizeVtable", 2, 4, |this| {
this.emit_enum_variant_arg(
0, |this| Ok(this.emit_existential_bounds(ecx, b)));
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
this.emit_enum_variant_arg(1, |this| def_id.encode(this));
this.emit_enum_variant_arg(2, |this| Ok(this.emit_ty(ecx, self_ty)));
this.emit_enum_variant_arg(3, |this| Ok(this.emit_substs(ecx, substs)))
})
}
}
});
}
}
trait write_tag_and_id {
fn tag(&mut self, tag_id: c::astencode_tag, f: |&mut Self|);
fn id(&mut self, id: ast::NodeId);
}
impl<'a> write_tag_and_id for Encoder<'a> {
fn tag(&mut self,
tag_id: c::astencode_tag,
f: |&mut Encoder<'a>|) {
self.start_tag(tag_id as uint);
f(self);
self.end_tag();
}
fn id(&mut self, id: ast::NodeId) {
self.wr_tagged_u64(c::tag_table_id as uint, id as u64);
}
}
struct SideTableEncodingIdVisitor<'a,'b:'a> {
2014-06-25 12:47:34 -07:00
ecx_ptr: *const libc::c_void,
new_rbml_w: &'a mut Encoder<'b>,
}
impl<'a,'b> ast_util::IdVisitingOperation for
SideTableEncodingIdVisitor<'a,'b> {
fn visit_id(&self, id: ast::NodeId) {
// Note: this will cause a copy of rbml_w, which is bad as
// it is mutable. But I believe it's harmless since we generate
// balanced EBML.
//
// FIXME(pcwalton): Don't copy this way.
let mut new_rbml_w = unsafe {
self.new_rbml_w.unsafe_clone()
};
// See above
let ecx: &e::EncodeContext = unsafe {
core: Remove the cast module This commit revisits the `cast` module in libcore and libstd, and scrutinizes all functions inside of it. The result was to remove the `cast` module entirely, folding all functionality into the `mem` module. Specifically, this is the fate of each function in the `cast` module. * transmute - This function was moved to `mem`, but it is now marked as #[unstable]. This is due to planned changes to the `transmute` function and how it can be invoked (see the #[unstable] comment). For more information, see RFC 5 and #12898 * transmute_copy - This function was moved to `mem`, with clarification that is is not an error to invoke it with T/U that are different sizes, but rather that it is strongly discouraged. This function is now #[stable] * forget - This function was moved to `mem` and marked #[stable] * bump_box_refcount - This function was removed due to the deprecation of managed boxes as well as its questionable utility. * transmute_mut - This function was previously deprecated, and removed as part of this commit. * transmute_mut_unsafe - This function doesn't serve much of a purpose when it can be achieved with an `as` in safe code, so it was removed. * transmute_lifetime - This function was removed because it is likely a strong indication that code is incorrect in the first place. * transmute_mut_lifetime - This function was removed for the same reasons as `transmute_lifetime` * copy_lifetime - This function was moved to `mem`, but it is marked `#[unstable]` now due to the likelihood of being removed in the future if it is found to not be very useful. * copy_mut_lifetime - This function was also moved to `mem`, but had the same treatment as `copy_lifetime`. * copy_lifetime_vec - This function was removed because it is not used today, and its existence is not necessary with DST (copy_lifetime will suffice). In summary, the cast module was stripped down to these functions, and then the functions were moved to the `mem` module. transmute - #[unstable] transmute_copy - #[stable] forget - #[stable] copy_lifetime - #[unstable] copy_mut_lifetime - #[unstable] [breaking-change]
2014-05-09 10:34:51 -07:00
mem::transmute(self.ecx_ptr)
};
encode_side_tables_for_id(ecx, &mut new_rbml_w, id)
}
}
fn encode_side_tables_for_ii(ecx: &e::EncodeContext,
rbml_w: &mut Encoder,
ii: &ast::InlinedItem) {
rbml_w.start_tag(c::tag_table as uint);
let mut new_rbml_w = unsafe {
rbml_w.unsafe_clone()
};
// Because the ast visitor uses @IdVisitingOperation, I can't pass in
// ecx directly, but /I/ know that it'll be fine since the lifetime is
// tied to the CrateContext that lives throughout this entire section.
ast_util::visit_ids_for_inlined_item(ii, &SideTableEncodingIdVisitor {
ecx_ptr: unsafe {
core: Remove the cast module This commit revisits the `cast` module in libcore and libstd, and scrutinizes all functions inside of it. The result was to remove the `cast` module entirely, folding all functionality into the `mem` module. Specifically, this is the fate of each function in the `cast` module. * transmute - This function was moved to `mem`, but it is now marked as #[unstable]. This is due to planned changes to the `transmute` function and how it can be invoked (see the #[unstable] comment). For more information, see RFC 5 and #12898 * transmute_copy - This function was moved to `mem`, with clarification that is is not an error to invoke it with T/U that are different sizes, but rather that it is strongly discouraged. This function is now #[stable] * forget - This function was moved to `mem` and marked #[stable] * bump_box_refcount - This function was removed due to the deprecation of managed boxes as well as its questionable utility. * transmute_mut - This function was previously deprecated, and removed as part of this commit. * transmute_mut_unsafe - This function doesn't serve much of a purpose when it can be achieved with an `as` in safe code, so it was removed. * transmute_lifetime - This function was removed because it is likely a strong indication that code is incorrect in the first place. * transmute_mut_lifetime - This function was removed for the same reasons as `transmute_lifetime` * copy_lifetime - This function was moved to `mem`, but it is marked `#[unstable]` now due to the likelihood of being removed in the future if it is found to not be very useful. * copy_mut_lifetime - This function was also moved to `mem`, but had the same treatment as `copy_lifetime`. * copy_lifetime_vec - This function was removed because it is not used today, and its existence is not necessary with DST (copy_lifetime will suffice). In summary, the cast module was stripped down to these functions, and then the functions were moved to the `mem` module. transmute - #[unstable] transmute_copy - #[stable] forget - #[stable] copy_lifetime - #[unstable] copy_mut_lifetime - #[unstable] [breaking-change]
2014-05-09 10:34:51 -07:00
mem::transmute(ecx)
},
new_rbml_w: &mut new_rbml_w,
});
rbml_w.end_tag();
}
fn encode_side_tables_for_id(ecx: &e::EncodeContext,
rbml_w: &mut Encoder,
id: ast::NodeId) {
let tcx = ecx.tcx;
debug!("Encoding side tables for id {}", id);
2014-03-20 19:49:20 -07:00
for def in tcx.def_map.borrow().find(&id).iter() {
rbml_w.tag(c::tag_table_def, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| (*def).encode(rbml_w).unwrap());
2014-03-20 19:49:20 -07:00
})
}
2014-03-20 19:49:20 -07:00
for &ty in tcx.node_types.borrow().find(&(id as uint)).iter() {
rbml_w.tag(c::tag_table_node_type, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
rbml_w.emit_ty(ecx, *ty);
})
2014-03-20 19:49:20 -07:00
})
}
for &item_substs in tcx.item_substs.borrow().find(&id).iter() {
rbml_w.tag(c::tag_table_item_subst, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
rbml_w.emit_substs(ecx, &item_substs.substs);
})
2014-03-20 19:49:20 -07:00
})
}
2014-03-20 19:49:20 -07:00
for &fv in tcx.freevars.borrow().find(&id).iter() {
rbml_w.tag(c::tag_table_freevars, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
rbml_w.emit_from_vec(fv.as_slice(), |rbml_w, fv_entry| {
Ok(encode_freevar_entry(rbml_w, fv_entry))
});
})
});
for freevar in fv.iter() {
match freevars::get_capture_mode(tcx, id) {
freevars::CaptureByRef => {
rbml_w.tag(c::tag_table_upvar_borrow_map, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
let var_id = freevar.def.def_id().node;
let upvar_id = ty::UpvarId {
var_id: var_id,
closure_expr_id: id
};
let upvar_borrow = tcx.upvar_borrow_map.borrow()
.get_copy(&upvar_id);
var_id.encode(rbml_w);
upvar_borrow.encode(rbml_w);
})
})
}
_ => {}
}
}
}
for &cm in tcx.capture_modes.borrow().find(&id).iter() {
rbml_w.tag(c::tag_table_capture_modes, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
encode_capture_mode(rbml_w, *cm);
})
})
}
let lid = ast::DefId { krate: ast::LOCAL_CRATE, node: id };
for &pty in tcx.tcache.borrow().find(&lid).iter() {
rbml_w.tag(c::tag_table_tcache, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
rbml_w.emit_polytype(ecx, pty.clone());
})
2014-03-20 19:49:20 -07:00
})
}
2014-03-20 19:49:20 -07:00
for &type_param_def in tcx.ty_param_defs.borrow().find(&id).iter() {
rbml_w.tag(c::tag_table_param_defs, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
rbml_w.emit_type_param_def(ecx, type_param_def)
})
2014-03-20 19:49:20 -07:00
})
}
let method_call = MethodCall::expr(id);
for &method in tcx.method_map.borrow().find(&method_call).iter() {
rbml_w.tag(c::tag_table_method_map, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
encode_method_callee(ecx, rbml_w, method_call.adjustment, method)
})
})
}
for &trait_ref in tcx.object_cast_map.borrow().find(&id).iter() {
rbml_w.tag(c::tag_table_object_cast_map, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
rbml_w.emit_trait_ref(ecx, &**trait_ref);
})
2014-03-20 19:49:20 -07:00
})
}
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
for &adjustment in tcx.adjustments.borrow().find(&id).iter() {
match *adjustment {
_ if ty::adjust_is_object(adjustment) => {
let method_call = MethodCall::autoobject(id);
for &method in tcx.method_map.borrow().find(&method_call).iter() {
2014-08-06 11:59:40 +02:00
rbml_w.tag(c::tag_table_method_map, |rbml_w| {
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
encode_method_callee(ecx, rbml_w, method_call.adjustment, method)
})
})
}
}
ty::AdjustDerefRef(ref adj) => {
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
assert!(!ty::adjust_is_object(adjustment));
for autoderef in range(0, adj.autoderefs) {
let method_call = MethodCall::autoderef(id, autoderef);
for &method in tcx.method_map.borrow().find(&method_call).iter() {
rbml_w.tag(c::tag_table_method_map, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
encode_method_callee(ecx, rbml_w,
method_call.adjustment, method)
})
})
}
}
}
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
_ => {
assert!(!ty::adjust_is_object(adjustment));
}
}
rbml_w.tag(c::tag_table_adjustments, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
2014-08-06 11:59:40 +02:00
rbml_w.emit_auto_adjustment(ecx, adjustment);
})
2014-03-20 19:49:20 -07:00
})
}
for unboxed_closure in tcx.unboxed_closures
.borrow()
.find(&ast_util::local_def(id))
.iter() {
rbml_w.tag(c::tag_table_unboxed_closures, |rbml_w| {
rbml_w.id(id);
rbml_w.tag(c::tag_table_val, |rbml_w| {
rbml_w.emit_closure_type(ecx, &unboxed_closure.closure_type);
encode_unboxed_closure_kind(rbml_w, unboxed_closure.kind)
})
})
}
}
trait doc_decoder_helpers {
2013-02-22 01:41:37 -05:00
fn as_int(&self) -> int;
fn opt_child(&self, tag: c::astencode_tag) -> Option<Self>;
}
impl<'a> doc_decoder_helpers for rbml::Doc<'a> {
2013-02-22 01:41:37 -05:00
fn as_int(&self) -> int { reader::doc_as_u64(*self) as int }
fn opt_child(&self, tag: c::astencode_tag) -> Option<rbml::Doc<'a>> {
2013-02-22 01:41:37 -05:00
reader::maybe_get_doc(*self, tag as uint)
}
}
trait rbml_decoder_decoder_helpers {
fn read_method_origin(&mut self, dcx: &DecodeContext) -> typeck::MethodOrigin;
2014-09-07 20:09:06 +03:00
fn read_ty(&mut self, dcx: &DecodeContext) -> ty::t;
fn read_tys(&mut self, dcx: &DecodeContext) -> Vec<ty::t>;
fn read_trait_ref(&mut self, dcx: &DecodeContext) -> Rc<ty::TraitRef>;
2014-09-07 20:09:06 +03:00
fn read_type_param_def(&mut self, dcx: &DecodeContext)
-> ty::TypeParameterDef;
2014-09-07 20:09:06 +03:00
fn read_polytype(&mut self, dcx: &DecodeContext)
-> ty::Polytype;
2014-09-07 20:09:06 +03:00
fn read_existential_bounds(&mut self, dcx: &DecodeContext) -> ty::ExistentialBounds;
fn read_substs(&mut self, dcx: &DecodeContext) -> subst::Substs;
fn read_auto_adjustment(&mut self, dcx: &DecodeContext) -> ty::AutoAdjustment;
fn read_unboxed_closure(&mut self, dcx: &DecodeContext)
-> ty::UnboxedClosure;
2014-09-07 20:09:06 +03:00
fn read_auto_deref_ref(&mut self, dcx: &DecodeContext) -> ty::AutoDerefRef;
fn read_autoref(&mut self, dcx: &DecodeContext) -> ty::AutoRef;
fn read_unsize_kind(&mut self, dcx: &DecodeContext) -> ty::UnsizeKind;
fn convert_def_id(&mut self,
2014-09-07 20:09:06 +03:00
dcx: &DecodeContext,
source: DefIdSource,
did: ast::DefId)
-> ast::DefId;
// Versions of the type reading functions that don't need the full
2014-09-07 20:09:06 +03:00
// DecodeContext.
fn read_ty_nodcx(&mut self,
2014-04-17 15:06:25 +03:00
tcx: &ty::ctxt, cdata: &cstore::crate_metadata) -> ty::t;
2014-09-07 20:09:06 +03:00
fn read_tys_nodcx(&mut self,
2014-03-06 05:07:47 +02:00
tcx: &ty::ctxt,
2014-04-17 15:06:25 +03:00
cdata: &cstore::crate_metadata) -> Vec<ty::t>;
2014-09-07 20:09:06 +03:00
fn read_substs_nodcx(&mut self, tcx: &ty::ctxt,
cdata: &cstore::crate_metadata)
-> subst::Substs;
}
impl<'a> rbml_decoder_decoder_helpers for reader::Decoder<'a> {
2014-09-07 20:09:06 +03:00
fn read_ty_nodcx(&mut self,
2014-04-17 15:06:25 +03:00
tcx: &ty::ctxt, cdata: &cstore::crate_metadata) -> ty::t {
2014-03-29 01:05:46 +01:00
self.read_opaque(|_, doc| {
Ok(tydecode::parse_ty_data(
doc.data,
cdata.cnum,
doc.start,
tcx,
|_, id| decoder::translate_def_id(cdata, id)))
2014-03-29 01:05:46 +01:00
}).unwrap()
}
2014-09-07 20:09:06 +03:00
fn read_tys_nodcx(&mut self,
2014-03-06 05:07:47 +02:00
tcx: &ty::ctxt,
2014-04-17 15:06:25 +03:00
cdata: &cstore::crate_metadata) -> Vec<ty::t> {
2014-09-07 20:09:06 +03:00
self.read_to_vec(|this| Ok(this.read_ty_nodcx(tcx, cdata)) )
2014-03-29 01:05:46 +01:00
.unwrap()
2014-09-14 20:27:36 -07:00
.into_iter()
.collect()
}
2014-09-07 20:09:06 +03:00
fn read_substs_nodcx(&mut self,
tcx: &ty::ctxt,
cdata: &cstore::crate_metadata)
-> subst::Substs
{
self.read_opaque(|_, doc| {
Ok(tydecode::parse_substs_data(
doc.data,
cdata.cnum,
doc.start,
tcx,
|_, id| decoder::translate_def_id(cdata, id)))
}).unwrap()
}
fn read_method_origin(&mut self, dcx: &DecodeContext)
-> typeck::MethodOrigin
{
self.read_enum("MethodOrigin", |this| {
let variants = ["MethodStatic", "MethodStaticUnboxedClosure",
"MethodTypeParam", "MethodTraitObject"];
this.read_enum_variant(variants, |this, i| {
Ok(match i {
0 => {
let def_id = this.read_def_id(dcx);
typeck::MethodStatic(def_id)
}
1 => {
let def_id = this.read_def_id(dcx);
typeck::MethodStaticUnboxedClosure(def_id)
}
2 => {
this.read_struct("MethodTypeParam", 2, |this| {
Ok(typeck::MethodTypeParam(
typeck::MethodParam {
trait_ref: {
this.read_struct_field("trait_ref", 0, |this| {
Ok(this.read_trait_ref(dcx))
}).unwrap()
},
method_num: {
this.read_struct_field("method_num", 1, |this| {
this.read_uint()
}).unwrap()
}
}))
}).unwrap()
}
3 => {
this.read_struct("MethodTraitObject", 2, |this| {
Ok(typeck::MethodTraitObject(
typeck::MethodObject {
trait_ref: {
this.read_struct_field("trait_ref", 0, |this| {
Ok(this.read_trait_ref(dcx))
}).unwrap()
},
object_trait_id: {
this.read_struct_field("object_trait_id", 1, |this| {
Ok(this.read_def_id(dcx))
}).unwrap()
},
method_num: {
this.read_struct_field("method_num", 2, |this| {
this.read_uint()
}).unwrap()
},
real_index: {
this.read_struct_field("real_index", 3, |this| {
this.read_uint()
}).unwrap()
},
}))
}).unwrap()
}
_ => fail!("..")
})
})
}).unwrap()
}
2014-09-07 20:09:06 +03:00
fn read_ty(&mut self, dcx: &DecodeContext) -> ty::t {
// Note: regions types embed local node ids. In principle, we
// should translate these node ids into the new decode
// context. However, we do not bother, because region types
// are not used during trans.
2014-03-29 01:05:46 +01:00
return self.read_opaque(|this, doc| {
debug!("read_ty({})", type_string(doc));
let ty = tydecode::parse_ty_data(
doc.data,
2014-09-07 20:09:06 +03:00
dcx.cdata.cnum,
doc.start,
2014-09-07 20:09:06 +03:00
dcx.tcx,
|s, a| this.convert_def_id(dcx, s, a));
2014-03-29 01:05:46 +01:00
Ok(ty)
}).unwrap();
fn type_string(doc: rbml::Doc) -> String {
let mut str = String::new();
for i in range(doc.start, doc.end) {
str.push_char(doc.data[i] as char);
}
str
}
}
2014-09-07 20:09:06 +03:00
fn read_tys(&mut self, dcx: &DecodeContext) -> Vec<ty::t> {
2014-09-14 20:27:36 -07:00
self.read_to_vec(|this| Ok(this.read_ty(dcx))).unwrap().into_iter().collect()
}
fn read_trait_ref(&mut self, dcx: &DecodeContext) -> Rc<ty::TraitRef> {
Rc::new(self.read_opaque(|this, doc| {
let ty = tydecode::parse_trait_ref_data(
doc.data,
dcx.cdata.cnum,
doc.start,
dcx.tcx,
|s, a| this.convert_def_id(dcx, s, a));
Ok(ty)
}).unwrap())
}
2014-09-07 20:09:06 +03:00
fn read_type_param_def(&mut self, dcx: &DecodeContext)
-> ty::TypeParameterDef {
2014-03-29 01:05:46 +01:00
self.read_opaque(|this, doc| {
Ok(tydecode::parse_type_param_def_data(
doc.data,
doc.start,
2014-09-07 20:09:06 +03:00
dcx.cdata.cnum,
dcx.tcx,
|s, a| this.convert_def_id(dcx, s, a)))
2014-03-29 01:05:46 +01:00
}).unwrap()
}
2014-09-07 20:09:06 +03:00
fn read_polytype(&mut self, dcx: &DecodeContext)
-> ty::Polytype {
self.read_struct("Polytype", 2, |this| {
Ok(ty::Polytype {
2014-03-29 01:05:46 +01:00
generics: this.read_struct_field("generics", 0, |this| {
this.read_struct("Generics", 2, |this| {
2014-03-29 01:05:46 +01:00
Ok(ty::Generics {
types:
this.read_struct_field("types", 0, |this| {
Ok(this.read_vec_per_param_space(
2014-09-07 20:09:06 +03:00
|this| this.read_type_param_def(dcx)))
2014-03-29 01:05:46 +01:00
}).unwrap(),
regions:
this.read_struct_field("regions", 1, |this| {
Ok(this.read_vec_per_param_space(
|this| Decodable::decode(this).unwrap()))
}).unwrap()
})
})
2014-03-29 01:05:46 +01:00
}).unwrap(),
ty: this.read_struct_field("ty", 1, |this| {
2014-09-07 20:09:06 +03:00
Ok(this.read_ty(dcx))
2014-03-29 01:05:46 +01:00
}).unwrap()
})
2014-03-29 01:05:46 +01:00
}).unwrap()
}
2014-09-07 20:09:06 +03:00
fn read_existential_bounds(&mut self, dcx: &DecodeContext) -> ty::ExistentialBounds
{
self.read_opaque(|this, doc| {
Ok(tydecode::parse_existential_bounds_data(doc.data,
2014-09-07 20:09:06 +03:00
dcx.cdata.cnum,
doc.start,
2014-09-07 20:09:06 +03:00
dcx.tcx,
|s, a| this.convert_def_id(dcx, s, a)))
}).unwrap()
}
2014-09-07 20:09:06 +03:00
fn read_substs(&mut self, dcx: &DecodeContext) -> subst::Substs {
2014-03-29 01:05:46 +01:00
self.read_opaque(|this, doc| {
Ok(tydecode::parse_substs_data(doc.data,
2014-09-07 20:09:06 +03:00
dcx.cdata.cnum,
doc.start,
2014-09-07 20:09:06 +03:00
dcx.tcx,
|s, a| this.convert_def_id(dcx, s, a)))
2014-03-29 01:05:46 +01:00
}).unwrap()
}
2014-09-07 20:09:06 +03:00
fn read_auto_adjustment(&mut self, dcx: &DecodeContext) -> ty::AutoAdjustment {
2014-03-29 01:05:46 +01:00
self.read_enum("AutoAdjustment", |this| {
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
let variants = ["AutoAddEnv", "AutoDerefRef"];
this.read_enum_variant(variants, |this, i| {
2014-03-29 01:05:46 +01:00
Ok(match i {
0 => {
let store: ty::TraitStore =
2014-03-29 01:05:46 +01:00
this.read_enum_variant_arg(0, |this| Decodable::decode(this)).unwrap();
ty::AdjustAddEnv(store.tr(dcx))
}
1 => {
let auto_deref_ref: ty::AutoDerefRef =
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
this.read_enum_variant_arg(0,
2014-09-07 20:09:06 +03:00
|this| Ok(this.read_auto_deref_ref(dcx))).unwrap();
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
ty::AdjustDerefRef(auto_deref_ref)
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
}
_ => fail!("bad enum variant for ty::AutoAdjustment")
})
})
}).unwrap()
}
2014-09-07 20:09:06 +03:00
fn read_auto_deref_ref(&mut self, dcx: &DecodeContext) -> ty::AutoDerefRef {
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
self.read_struct("AutoDerefRef", 2, |this| {
Ok(ty::AutoDerefRef {
autoderefs: this.read_struct_field("autoderefs", 0, |this| {
Decodable::decode(this)
}).unwrap(),
autoref: this.read_struct_field("autoref", 1, |this| {
this.read_option(|this, b| {
if b {
2014-09-07 20:09:06 +03:00
Ok(Some(this.read_autoref(dcx)))
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
} else {
Ok(None)
}
})
}).unwrap(),
})
}).unwrap()
}
2014-09-07 20:09:06 +03:00
fn read_autoref(&mut self, dcx: &DecodeContext) -> ty::AutoRef {
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
self.read_enum("AutoRef", |this| {
let variants = ["AutoPtr",
"AutoUnsize",
"AutoUnsizeUniq",
"AutoUnsafe"];
this.read_enum_variant(variants, |this, i| {
Ok(match i {
0 => {
let r: ty::Region =
2014-03-29 01:05:46 +01:00
this.read_enum_variant_arg(0, |this| Decodable::decode(this)).unwrap();
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
let m: ast::Mutability =
this.read_enum_variant_arg(1, |this| Decodable::decode(this)).unwrap();
let a: Option<Box<ty::AutoRef>> =
this.read_enum_variant_arg(2, |this| this.read_option(|this, b| {
if b {
2014-09-07 20:09:06 +03:00
Ok(Some(box this.read_autoref(dcx)))
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
} else {
Ok(None)
}
})).unwrap();
2014-09-07 20:09:06 +03:00
ty::AutoPtr(r.tr(dcx), m, a)
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
}
1 => {
let uk: ty::UnsizeKind =
this.read_enum_variant_arg(0,
2014-09-07 20:09:06 +03:00
|this| Ok(this.read_unsize_kind(dcx))).unwrap();
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
ty::AutoUnsize(uk)
}
2 => {
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
let uk: ty::UnsizeKind =
this.read_enum_variant_arg(0,
2014-09-07 20:09:06 +03:00
|this| Ok(this.read_unsize_kind(dcx))).unwrap();
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
ty::AutoUnsizeUniq(uk)
}
3 => {
let m: ast::Mutability =
2014-03-29 01:05:46 +01:00
this.read_enum_variant_arg(0, |this| Decodable::decode(this)).unwrap();
let a: Option<Box<ty::AutoRef>> =
this.read_enum_variant_arg(1, |this| this.read_option(|this, b| {
if b {
2014-09-07 20:09:06 +03:00
Ok(Some(box this.read_autoref(dcx)))
} else {
Ok(None)
}
})).unwrap();
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
ty::AutoUnsafe(m, a)
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
}
_ => fail!("bad enum variant for ty::AutoRef")
})
})
}).unwrap()
}
2014-09-07 20:09:06 +03:00
fn read_unsize_kind(&mut self, dcx: &DecodeContext) -> ty::UnsizeKind {
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
self.read_enum("UnsizeKind", |this| {
let variants = ["UnsizeLength", "UnsizeStruct", "UnsizeVtable"];
this.read_enum_variant(variants, |this, i| {
Ok(match i {
0 => {
let len: uint =
this.read_enum_variant_arg(0, |this| Decodable::decode(this)).unwrap();
ty::UnsizeLength(len)
}
1 => {
let uk: ty::UnsizeKind =
this.read_enum_variant_arg(0,
2014-09-07 20:09:06 +03:00
|this| Ok(this.read_unsize_kind(dcx))).unwrap();
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
let idx: uint =
this.read_enum_variant_arg(1, |this| Decodable::decode(this)).unwrap();
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
ty::UnsizeStruct(box uk, idx)
}
2 => {
let b =
this.read_enum_variant_arg(
2014-09-07 20:09:06 +03:00
0, |this| Ok(this.read_existential_bounds(dcx))).unwrap();
let def_id: ast::DefId =
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
this.read_enum_variant_arg(1, |this| Decodable::decode(this)).unwrap();
let self_ty =
this.read_enum_variant_arg(2, |this| Ok(this.read_ty(dcx))).unwrap();
let substs = this.read_enum_variant_arg(3,
2014-09-07 20:09:06 +03:00
|this| Ok(this.read_substs(dcx))).unwrap();
let ty_trait = ty::TyTrait { def_id: def_id.tr(dcx),
bounds: b,
substs: substs };
ty::UnsizeVtable(ty_trait, self_ty)
}
DST coercions and DST structs [breaking-change] 1. The internal layout for traits has changed from (vtable, data) to (data, vtable). If you were relying on this in unsafe transmutes, you might get some very weird and apparently unrelated errors. You should not be doing this! Prefer not to do this at all, but if you must, you should use raw::TraitObject rather than hardcoding rustc's internal representation into your code. 2. The minimal type of reference-to-vec-literals (e.g., `&[1, 2, 3]`) is now a fixed size vec (e.g., `&[int, ..3]`) where it used to be an unsized vec (e.g., `&[int]`). If you want the unszied type, you must explicitly give the type (e.g., `let x: &[_] = &[1, 2, 3]`). Note in particular where multiple blocks must have the same type (e.g., if and else clauses, vec elements), the compiler will not coerce to the unsized type without a hint. E.g., `[&[1], &[1, 2]]` used to be a valid expression of type '[&[int]]'. It no longer type checks since the first element now has type `&[int, ..1]` and the second has type &[int, ..2]` which are incompatible. 3. The type of blocks (including functions) must be coercible to the expected type (used to be a subtype). Mostly this makes things more flexible and not less (in particular, in the case of coercing function bodies to the return type). However, in some rare cases, this is less flexible. TBH, I'm not exactly sure of the exact effects. I think the change causes us to resolve inferred type variables slightly earlier which might make us slightly more restrictive. Possibly it only affects blocks with unreachable code. E.g., `if ... { fail!(); "Hello" }` used to type check, it no longer does. The fix is to add a semicolon after the string.
2014-08-04 14:20:11 +02:00
_ => fail!("bad enum variant for ty::UnsizeKind")
})
})
2014-03-29 01:05:46 +01:00
}).unwrap()
}
2014-09-07 20:09:06 +03:00
fn read_unboxed_closure(&mut self, dcx: &DecodeContext)
-> ty::UnboxedClosure {
let closure_type = self.read_opaque(|this, doc| {
Ok(tydecode::parse_ty_closure_data(
doc.data,
2014-09-07 20:09:06 +03:00
dcx.cdata.cnum,
doc.start,
2014-09-07 20:09:06 +03:00
dcx.tcx,
|s, a| this.convert_def_id(dcx, s, a)))
}).unwrap();
let variants = [
"FnUnboxedClosureKind",
"FnMutUnboxedClosureKind",
"FnOnceUnboxedClosureKind"
];
let kind = self.read_enum_variant(variants, |_, i| {
Ok(match i {
0 => ty::FnUnboxedClosureKind,
1 => ty::FnMutUnboxedClosureKind,
2 => ty::FnOnceUnboxedClosureKind,
_ => fail!("bad enum variant for ty::UnboxedClosureKind"),
})
}).unwrap();
ty::UnboxedClosure {
closure_type: closure_type,
kind: kind,
}
}
fn convert_def_id(&mut self,
2014-09-07 20:09:06 +03:00
dcx: &DecodeContext,
source: tydecode::DefIdSource,
did: ast::DefId)
-> ast::DefId {
/*!
* Converts a def-id that appears in a type. The correct
* translation will depend on what kind of def-id this is.
* This is a subtle point: type definitions are not
* inlined into the current crate, so if the def-id names
* a nominal type or type alias, then it should be
* translated to refer to the source crate.
*
* However, *type parameters* are cloned along with the function
* they are attached to. So we should translate those def-ids
* to refer to the new, cloned copy of the type parameter.
* We only see references to free type parameters in the body of
* an inlined function. In such cases, we need the def-id to
* be a local id so that the TypeContents code is able to lookup
* the relevant info in the ty_param_defs table.
*
* *Region parameters*, unfortunately, are another kettle of fish.
* In such cases, def_id's can appear in types to distinguish
* shadowed bound regions and so forth. It doesn't actually
* matter so much what we do to these, since regions are erased
* at trans time, but it's good to keep them consistent just in
* case. We translate them with `tr_def_id()` which will map
* the crate numbers back to the original source crate.
*
* It'd be really nice to refactor the type repr to not include
* def-ids so that all these distinctions were unnecessary.
*/
let r = match source {
2014-09-07 20:09:06 +03:00
NominalType | TypeWithId | RegionParameter => dcx.tr_def_id(did),
TypeParameter => dcx.tr_intern_def_id(did)
};
debug!("convert_def_id(source={:?}, did={:?})={:?}", source, did, r);
return r;
}
}
2014-09-07 20:09:06 +03:00
fn decode_side_tables(dcx: &DecodeContext,
ast_doc: rbml::Doc) {
let tbl_doc = ast_doc.get(c::tag_table as uint);
reader::docs(tbl_doc, |tag, entry_doc| {
let id0 = entry_doc.get(c::tag_table_id as uint).as_int();
2014-09-07 20:09:06 +03:00
let id = dcx.tr_id(id0 as ast::NodeId);
debug!(">> Side table document with tag 0x{:x} \
2013-09-27 22:38:08 -07:00
found for id {} (orig {})",
2012-08-22 17:24:52 -07:00
tag, id, id0);
match c::astencode_tag::from_uint(tag) {
None => {
2014-09-07 20:09:06 +03:00
dcx.tcx.sess.bug(
format!("unknown tag found in side tables: {:x}",
tag).as_slice());
}
Some(value) => {
let val_doc = entry_doc.get(c::tag_table_val as uint);
let mut val_dsr = reader::Decoder::new(val_doc);
let val_dsr = &mut val_dsr;
match value {
c::tag_table_def => {
2014-09-07 20:09:06 +03:00
let def = decode_def(dcx, val_doc);
2014-03-20 19:49:20 -07:00
dcx.tcx.def_map.borrow_mut().insert(id, def);
}
c::tag_table_node_type => {
2014-09-07 20:09:06 +03:00
let ty = val_dsr.read_ty(dcx);
debug!("inserting ty for node {:?}: {}",
id, ty_to_string(dcx.tcx, ty));
2014-03-20 19:49:20 -07:00
dcx.tcx.node_types.borrow_mut().insert(id as uint, ty);
}
c::tag_table_item_subst => {
let item_substs = ty::ItemSubsts {
2014-09-07 20:09:06 +03:00
substs: val_dsr.read_substs(dcx)
};
dcx.tcx.item_substs.borrow_mut().insert(
id, item_substs);
}
c::tag_table_freevars => {
2014-04-14 17:43:57 +03:00
let fv_info = val_dsr.read_to_vec(|val_dsr| {
2014-09-07 20:09:06 +03:00
Ok(val_dsr.read_freevar_entry(dcx))
2014-09-14 20:27:36 -07:00
}).unwrap().into_iter().collect();
2014-03-20 19:49:20 -07:00
dcx.tcx.freevars.borrow_mut().insert(id, fv_info);
}
c::tag_table_upvar_borrow_map => {
let var_id: ast::NodeId = Decodable::decode(val_dsr).unwrap();
let upvar_id = ty::UpvarId {
2014-09-07 20:09:06 +03:00
var_id: dcx.tr_id(var_id),
closure_expr_id: id
};
let ub: ty::UpvarBorrow = Decodable::decode(val_dsr).unwrap();
2014-09-07 20:09:06 +03:00
dcx.tcx.upvar_borrow_map.borrow_mut().insert(upvar_id, ub.tr(dcx));
}
c::tag_table_capture_modes => {
let capture_mode = val_dsr.read_capture_mode();
dcx.tcx
.capture_modes
.borrow_mut()
.insert(id, capture_mode);
}
c::tag_table_tcache => {
2014-09-07 20:09:06 +03:00
let pty = val_dsr.read_polytype(dcx);
let lid = ast::DefId { krate: ast::LOCAL_CRATE, node: id };
dcx.tcx.tcache.borrow_mut().insert(lid, pty);
}
c::tag_table_param_defs => {
2014-09-07 20:09:06 +03:00
let bounds = val_dsr.read_type_param_def(dcx);
2014-03-20 19:49:20 -07:00
dcx.tcx.ty_param_defs.borrow_mut().insert(id, bounds);
}
c::tag_table_method_map => {
2014-09-07 20:09:06 +03:00
let (adjustment, method) = val_dsr.read_method_callee(dcx);
let method_call = MethodCall {
expr_id: id,
adjustment: adjustment
};
dcx.tcx.method_map.borrow_mut().insert(method_call, method);
}
c::tag_table_object_cast_map => {
let trait_ref = val_dsr.read_trait_ref(dcx);
dcx.tcx.object_cast_map.borrow_mut()
.insert(id, trait_ref);
}
c::tag_table_adjustments => {
2014-09-07 20:09:06 +03:00
let adj: ty::AutoAdjustment = val_dsr.read_auto_adjustment(dcx);
2014-03-20 19:49:20 -07:00
dcx.tcx.adjustments.borrow_mut().insert(id, adj);
}
c::tag_table_unboxed_closures => {
let unboxed_closure =
2014-09-07 20:09:06 +03:00
val_dsr.read_unboxed_closure(dcx);
dcx.tcx
.unboxed_closures
.borrow_mut()
.insert(ast_util::local_def(id),
unboxed_closure);
}
_ => {
2014-09-07 20:09:06 +03:00
dcx.tcx.sess.bug(
format!("unknown tag found in side tables: {:x}",
tag).as_slice());
}
}
}
}
debug!(">< Side table doc loaded");
true
});
}
// ______________________________________________________________________
// Testing of astencode_gen
#[cfg(test)]
2014-09-07 20:09:06 +03:00
fn encode_item_ast(rbml_w: &mut Encoder, item: &ast::Item) {
rbml_w.start_tag(c::tag_tree as uint);
(*item).encode(rbml_w);
rbml_w.end_tag();
}
#[cfg(test)]
2014-09-07 20:09:06 +03:00
fn decode_item_ast(par_doc: rbml::Doc) -> ast::Item {
let chi_doc = par_doc.get(c::tag_tree as uint);
let mut d = reader::Decoder::new(chi_doc);
2014-09-07 20:09:06 +03:00
Decodable::decode(&mut d).unwrap()
}
#[cfg(test)]
trait fake_ext_ctxt {
fn cfg(&self) -> ast::CrateConfig;
2014-03-09 16:54:34 +02:00
fn parse_sess<'a>(&'a self) -> &'a parse::ParseSess;
fn call_site(&self) -> Span;
2013-09-02 02:50:59 +02:00
fn ident_of(&self, st: &str) -> ast::Ident;
}
#[cfg(test)]
2014-03-09 16:54:34 +02:00
impl fake_ext_ctxt for parse::ParseSess {
fn cfg(&self) -> ast::CrateConfig {
Vec::new()
}
2014-03-09 16:54:34 +02:00
fn parse_sess<'a>(&'a self) -> &'a parse::ParseSess { self }
fn call_site(&self) -> Span {
codemap::Span {
lo: codemap::BytePos(0),
hi: codemap::BytePos(0),
expn_info: None
}
}
2013-09-02 02:50:59 +02:00
fn ident_of(&self, st: &str) -> ast::Ident {
token::str_to_ident(st)
}
}
#[cfg(test)]
2014-03-09 16:54:34 +02:00
fn mk_ctxt() -> parse::ParseSess {
2014-02-07 00:38:33 +02:00
parse::new_parse_sess()
}
#[cfg(test)]
2014-09-07 20:09:06 +03:00
fn roundtrip(in_item: Option<P<ast::Item>>) {
let in_item = in_item.unwrap();
remove seek from std::io::MemWriter, add SeekableMemWriter to librustc Not all users of MemWriter need to seek, but having MemWriter seekable adds between 3-29% in overhead in certain circumstances. This fixes that performance gap by making a non-seekable MemWriter, and creating a new SeekableMemWriter for those circumstances when that functionality is actually needed. ``` test io::mem::test::bench_buf_reader ... bench: 682 ns/iter (+/- 85) test io::mem::test::bench_buf_writer ... bench: 580 ns/iter (+/- 57) test io::mem::test::bench_mem_reader ... bench: 793 ns/iter (+/- 99) test io::mem::test::bench_mem_writer_001_0000 ... bench: 48 ns/iter (+/- 27) test io::mem::test::bench_mem_writer_001_0010 ... bench: 65 ns/iter (+/- 27) = 153 MB/s test io::mem::test::bench_mem_writer_001_0100 ... bench: 132 ns/iter (+/- 12) = 757 MB/s test io::mem::test::bench_mem_writer_001_1000 ... bench: 802 ns/iter (+/- 151) = 1246 MB/s test io::mem::test::bench_mem_writer_100_0000 ... bench: 481 ns/iter (+/- 28) test io::mem::test::bench_mem_writer_100_0010 ... bench: 1957 ns/iter (+/- 126) = 510 MB/s test io::mem::test::bench_mem_writer_100_0100 ... bench: 8222 ns/iter (+/- 434) = 1216 MB/s test io::mem::test::bench_mem_writer_100_1000 ... bench: 82496 ns/iter (+/- 11191) = 1212 MB/s test io::mem::test::bench_seekable_mem_writer_001_0000 ... bench: 48 ns/iter (+/- 2) test io::mem::test::bench_seekable_mem_writer_001_0010 ... bench: 64 ns/iter (+/- 2) = 156 MB/s test io::mem::test::bench_seekable_mem_writer_001_0100 ... bench: 129 ns/iter (+/- 7) = 775 MB/s test io::mem::test::bench_seekable_mem_writer_001_1000 ... bench: 801 ns/iter (+/- 159) = 1248 MB/s test io::mem::test::bench_seekable_mem_writer_100_0000 ... bench: 711 ns/iter (+/- 51) test io::mem::test::bench_seekable_mem_writer_100_0010 ... bench: 2532 ns/iter (+/- 227) = 394 MB/s test io::mem::test::bench_seekable_mem_writer_100_0100 ... bench: 8962 ns/iter (+/- 947) = 1115 MB/s test io::mem::test::bench_seekable_mem_writer_100_1000 ... bench: 85086 ns/iter (+/- 11555) = 1175 MB/s ``` [breaking-change]
2014-07-29 16:31:39 -07:00
let mut wr = SeekableMemWriter::new();
2014-09-07 20:09:06 +03:00
encode_item_ast(&mut writer::Encoder::new(&mut wr), &*in_item);
let rbml_doc = rbml::Doc::new(wr.get_ref());
let out_item = decode_item_ast(rbml_doc);
2014-09-07 20:09:06 +03:00
assert!(*in_item == out_item);
}
#[test]
fn test_basic() {
let cx = mk_ctxt();
roundtrip(quote_item!(&cx,
fn foo() {}
));
}
2014-07-08 22:28:52 -07:00
/* NOTE: When there's a snapshot, update this (yay quasiquoter!)
#[test]
fn test_smalltalk() {
let cx = mk_ctxt();
roundtrip(quote_item!(&cx,
fn foo() -> int { 3 + 4 } // first smalltalk program ever executed.
));
}
2014-07-08 22:28:52 -07:00
*/
#[test]
fn test_more() {
let cx = mk_ctxt();
roundtrip(quote_item!(&cx,
fn foo(x: uint, y: uint) -> uint {
let z = x + y;
2012-08-01 17:30:05 -07:00
return z;
}
));
}
#[test]
fn test_simplification() {
let cx = mk_ctxt();
2014-03-17 09:55:41 +02:00
let item = quote_item!(&cx,
fn new_int_alist<B>() -> alist<int, B> {
2013-05-07 17:57:58 -07:00
fn eq_int(a: int, b: int) -> bool { a == b }
return alist {eq_fn: eq_int, data: Vec::new()};
}
2014-01-06 14:00:46 +02:00
).unwrap();
let item_in = e::IIItemRef(&*item);
2014-01-06 14:00:46 +02:00
let item_out = simplify_ast(item_in);
let item_exp = ast::IIItem(quote_item!(&cx,
fn new_int_alist<B>() -> alist<int, B> {
return alist {eq_fn: eq_int, data: Vec::new()};
}
).unwrap());
2012-08-06 12:34:08 -07:00
match (item_out, item_exp) {
(ast::IIItem(item_out), ast::IIItem(item_exp)) => {
assert!(pprust::item_to_string(&*item_out) ==
pprust::item_to_string(&*item_exp));
}
_ => fail!()
}
}