rust/src/librustc/lint/context.rs

1060 lines
37 KiB
Rust
Raw Normal View History

// Copyright 2012-2015 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.
//! Implementation of lint checking.
//!
//! The lint checking is mostly consolidated into one pass which runs just
//! before translation to LLVM bytecode. Throughout compilation, lint warnings
//! can be added via the `add_lint` method on the Session structure. This
//! requires a span and an id of the node that the lint is being added to. The
//! lint isn't actually emitted at that time because it is unknown what the
//! actual lint level at that location is.
//!
//! To actually emit lint warnings/errors, a separate pass is used just before
//! translation. A context keeps track of the current state of all lint levels.
//! Upon entering a node of the ast which can modify the lint settings, the
//! previous lint state is pushed onto a stack and the ast is then recursed
//! upon. As the ast is traversed, this keeps track of the current lint level
//! for all lint attributes.
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
use self::TargetLint::*;
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
use rustc_back::slice;
use lint::{EarlyLintPassObject, LateLintPassObject};
use lint::{Level, Lint, LintId, LintPass, LintBuffer};
use lint::levels::{LintLevelSets, LintLevelsBuilder};
use middle::privacy::AccessLevels;
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
use rustc_serialize::{Decoder, Decodable, Encoder, Encodable};
use session::{config, early_error, Session};
use traits::Reveal;
use ty::{self, TyCtxt};
use util::nodemap::FxHashMap;
use std::default::Default as StdDefault;
use std::cell::{Ref, RefCell};
2015-11-15 17:17:50 +01:00
use syntax::ast;
use syntax_pos::{MultiSpan, Span};
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
use errors::DiagnosticBuilder;
2016-03-29 08:50:44 +03:00
use hir;
use hir::def_id::LOCAL_CRATE;
2016-03-29 08:50:44 +03:00
use hir::intravisit as hir_visit;
use syntax::visit as ast_visit;
/// Information about the registered lints.
///
/// This is basically the subset of `Context` that we can
/// build early in the compile pipeline.
pub struct LintStore {
/// Registered lints. The bool is true if the lint was
/// added by a plugin.
lints: Vec<(&'static Lint, bool)>,
/// Trait objects for each lint pass.
/// This is only `None` while performing a lint pass. See the definition
/// of `LintSession::new`.
early_passes: Option<Vec<EarlyLintPassObject>>,
late_passes: Option<Vec<LateLintPassObject>>,
/// Lints indexed by name.
by_name: FxHashMap<String, TargetLint>,
/// Map of registered lint groups to what lints they expand to. The bool
/// is true if the lint group was added by a plugin.
lint_groups: FxHashMap<&'static str, (Vec<LintId>, bool)>,
2017-08-11 20:34:14 +02:00
/// Extra info for future incompatibility lints, describing the
/// issue or RFC that caused the incompatibility.
future_incompatible: FxHashMap<LintId, FutureIncompatibleInfo>,
}
pub struct LintSession<'a, PassObject> {
/// Reference to the store of registered lints.
lints: Ref<'a, LintStore>,
/// Trait objects for each lint pass.
passes: Option<Vec<PassObject>>,
}
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
/// Lints that are buffered up early on in the `Session` before the
/// `LintLevels` is calculated
#[derive(PartialEq, RustcEncodable, RustcDecodable, Debug)]
pub struct BufferedEarlyLint {
pub lint_id: LintId,
pub ast_id: ast::NodeId,
pub span: MultiSpan,
pub msg: String,
2016-10-12 16:38:58 -04:00
}
/// Extra information for a future incompatibility lint. See the call
/// to `register_future_incompatible` in `librustc_lint/lib.rs` for
/// guidelines.
pub struct FutureIncompatibleInfo {
pub id: LintId,
pub reference: &'static str // e.g., a URL for an issue/PR/RFC or error code
}
/// The targed of the `by_name` map, which accounts for renaming/deprecation.
enum TargetLint {
/// A direct lint target
Id(LintId),
/// Temporary renaming, used for easing migration pain; see #16545
Renamed(String, LintId),
/// Lint with this name existed previously, but has been removed/deprecated.
/// The string argument is the reason for removal.
Removed(String),
}
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
pub enum FindLintError {
NotFound,
2017-03-13 02:12:13 +02:00
Removed,
}
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
pub enum CheckLintNameResult<'a> {
Ok(&'a [LintId]),
2017-08-17 16:15:40 +02:00
/// Lint doesn't exist
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
NoLint,
2017-08-17 16:15:40 +02:00
/// The lint is either renamed or removed. This is the warning
/// message.
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
Warning(String),
}
impl LintStore {
pub fn new() -> LintStore {
LintStore {
lints: vec![],
early_passes: Some(vec![]),
late_passes: Some(vec![]),
by_name: FxHashMap(),
future_incompatible: FxHashMap(),
lint_groups: FxHashMap(),
}
}
pub fn get_lints<'t>(&'t self) -> &'t [(&'static Lint, bool)] {
&self.lints
}
pub fn get_lint_groups<'t>(&'t self) -> Vec<(&'static str, Vec<LintId>, bool)> {
self.lint_groups.iter().map(|(k, v)| (*k,
2014-12-09 12:21:18 -05:00
v.0.clone(),
v.1)).collect()
}
pub fn register_early_pass(&mut self,
sess: Option<&Session>,
from_plugin: bool,
pass: EarlyLintPassObject) {
self.push_pass(sess, from_plugin, &pass);
self.early_passes.as_mut().unwrap().push(pass);
}
pub fn register_late_pass(&mut self,
sess: Option<&Session>,
from_plugin: bool,
pass: LateLintPassObject) {
self.push_pass(sess, from_plugin, &pass);
self.late_passes.as_mut().unwrap().push(pass);
}
// Helper method for register_early/late_pass
fn push_pass<P: LintPass + ?Sized + 'static>(&mut self,
sess: Option<&Session>,
from_plugin: bool,
pass: &Box<P>) {
2015-01-31 12:20:46 -05:00
for &lint in pass.get_lints() {
2014-10-06 17:30:54 -07:00
self.lints.push((*lint, from_plugin));
2014-10-06 17:30:54 -07:00
let id = LintId::of(*lint);
2014-11-06 12:25:16 -05:00
if self.by_name.insert(lint.name_lower(), Id(id)).is_some() {
let msg = format!("duplicate specification of lint {}", lint.name_lower());
match (sess, from_plugin) {
// We load builtin lints first, so a duplicate is a compiler bug.
// Use early_error when handling -W help with no crate.
(None, _) => early_error(config::ErrorOutputType::default(), &msg[..]),
(Some(_), false) => bug!("{}", msg),
// A duplicate name from a plugin is a user error.
(Some(sess), true) => sess.err(&msg[..]),
}
}
}
}
pub fn register_future_incompatible(&mut self,
sess: Option<&Session>,
lints: Vec<FutureIncompatibleInfo>) {
let ids = lints.iter().map(|f| f.id).collect();
self.register_group(sess, false, "future_incompatible", ids);
for info in lints {
self.future_incompatible.insert(info.id, info);
}
}
pub fn future_incompatible(&self, id: LintId) -> Option<&FutureIncompatibleInfo> {
self.future_incompatible.get(&id)
}
pub fn register_group(&mut self, sess: Option<&Session>,
from_plugin: bool, name: &'static str,
to: Vec<LintId>) {
2014-11-06 12:25:16 -05:00
let new = self.lint_groups.insert(name, (to, from_plugin)).is_none();
if !new {
let msg = format!("duplicate specification of lint group {}", name);
match (sess, from_plugin) {
// We load builtin lints first, so a duplicate is a compiler bug.
// Use early_error when handling -W help with no crate.
(None, _) => early_error(config::ErrorOutputType::default(), &msg[..]),
(Some(_), false) => bug!("{}", msg),
// A duplicate name from a plugin is a user error.
(Some(sess), true) => sess.err(&msg[..]),
}
}
}
pub fn register_renamed(&mut self, old_name: &str, new_name: &str) {
let target = match self.by_name.get(new_name) {
Some(&Id(lint_id)) => lint_id.clone(),
2016-03-26 19:59:04 +01:00
_ => bug!("invalid lint renaming of {} to {}", old_name, new_name)
};
self.by_name.insert(old_name.to_string(), Renamed(new_name.to_string(), target));
}
pub fn register_removed(&mut self, name: &str, reason: &str) {
self.by_name.insert(name.into(), Removed(reason.into()));
}
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
pub fn find_lints(&self, lint_name: &str) -> Result<Vec<LintId>, FindLintError> {
match self.by_name.get(lint_name) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
Some(&Id(lint_id)) => Ok(vec![lint_id]),
Some(&Renamed(_, lint_id)) => {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
Ok(vec![lint_id])
},
Some(&Removed(_)) => {
Err(FindLintError::Removed)
},
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
None => {
match self.lint_groups.get(lint_name) {
Some(v) => Ok(v.0.clone()),
None => Err(FindLintError::Removed)
}
}
}
}
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
// Checks the validity of lint names derived from the command line
pub fn check_lint_name_cmdline(&self,
sess: &Session,
lint_name: &str,
level: Level) {
let db = match self.check_lint_name(lint_name) {
CheckLintNameResult::Ok(_) => None,
CheckLintNameResult::Warning(ref msg) => {
Some(sess.struct_warn(msg))
},
CheckLintNameResult::NoLint => {
Some(struct_err!(sess, E0602, "unknown lint: `{}`", lint_name))
}
};
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
if let Some(mut db) = db {
let msg = format!("requested on the command line with `{} {}`",
match level {
Level::Allow => "-A",
Level::Warn => "-W",
Level::Deny => "-D",
Level::Forbid => "-F",
},
lint_name);
db.note(&msg);
db.emit();
}
}
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
/// Checks the name of a lint for its existence, and whether it was
/// renamed or removed. Generates a DiagnosticBuilder containing a
/// warning for renamed and removed lints. This is over both lint
/// names from attributes and those passed on the command line. Since
/// it emits non-fatal warnings and there are *two* lint passes that
/// inspect attributes, this is only run from the late pass to avoid
/// printing duplicate warnings.
pub fn check_lint_name(&self, lint_name: &str) -> CheckLintNameResult {
match self.by_name.get(lint_name) {
Some(&Renamed(ref new_name, _)) => {
CheckLintNameResult::Warning(
format!("lint {} has been renamed to {}", lint_name, new_name)
)
},
Some(&Removed(ref reason)) => {
CheckLintNameResult::Warning(
format!("lint {} has been removed: {}", lint_name, reason)
)
},
None => {
match self.lint_groups.get(lint_name) {
None => CheckLintNameResult::NoLint,
Some(ids) => CheckLintNameResult::Ok(&ids.0),
}
}
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
Some(&Id(ref id)) => CheckLintNameResult::Ok(slice::ref_slice(id)),
}
}
}
impl<'a, PassObject: LintPassObject> LintSession<'a, PassObject> {
/// Creates a new `LintSession`, by moving out the `LintStore`'s initial
/// lint levels and pass objects. These can be restored using the `restore`
/// method.
fn new(store: &'a RefCell<LintStore>) -> LintSession<'a, PassObject> {
let mut s = store.borrow_mut();
let passes = PassObject::take_passes(&mut *s);
drop(s);
LintSession {
lints: store.borrow(),
passes,
}
}
/// Restores the levels back to the original lint store.
fn restore(self, store: &RefCell<LintStore>) {
drop(self.lints);
let mut s = store.borrow_mut();
PassObject::restore_passes(&mut *s, self.passes);
}
}
/// Context for lint checking after type checking.
pub struct LateContext<'a, 'tcx: 'a> {
/// Type context we're checking in.
pub tcx: TyCtxt<'a, 'tcx, 'tcx>,
/// Side-tables for the body we are in.
2017-01-25 16:24:00 -05:00
pub tables: &'a ty::TypeckTables<'tcx>,
/// Parameter environment for the item we are in.
pub param_env: ty::ParamEnv<'tcx>,
/// Items accessible from the crate being checked.
pub access_levels: &'a AccessLevels,
/// The store of registered lints and the lint levels.
lint_sess: LintSession<'tcx, LateLintPassObject>,
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
last_ast_node_with_lint_attrs: ast::NodeId,
}
/// Context for lint checking of the AST, after expansion, before lowering to
/// HIR.
pub struct EarlyContext<'a> {
/// Type context we're checking in.
pub sess: &'a Session,
/// The crate being checked.
pub krate: &'a ast::Crate,
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
builder: LintLevelsBuilder<'a>,
/// The store of registered lints and the lint levels.
lint_sess: LintSession<'a, EarlyLintPassObject>,
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
buffered: LintBuffer,
}
/// Convenience macro for calling a `LintPass` method on every pass in the context.
macro_rules! run_lints { ($cx:expr, $f:ident, $ps:ident, $($args:expr),*) => ({
// Move the vector of passes out of `$cx` so that we can
// iterate over it mutably while passing `$cx` to the methods.
let mut passes = $cx.lint_sess_mut().passes.take().unwrap();
for obj in &mut passes {
obj.$f($cx, $($args),*);
}
$cx.lint_sess_mut().passes = Some(passes);
}) }
pub trait LintPassObject: Sized {
fn take_passes(store: &mut LintStore) -> Option<Vec<Self>>;
fn restore_passes(store: &mut LintStore, passes: Option<Vec<Self>>);
}
impl LintPassObject for EarlyLintPassObject {
fn take_passes(store: &mut LintStore) -> Option<Vec<Self>> {
store.early_passes.take()
}
fn restore_passes(store: &mut LintStore, passes: Option<Vec<Self>>) {
store.early_passes = passes;
}
}
impl LintPassObject for LateLintPassObject {
fn take_passes(store: &mut LintStore) -> Option<Vec<Self>> {
store.late_passes.take()
}
fn restore_passes(store: &mut LintStore, passes: Option<Vec<Self>>) {
store.late_passes = passes;
}
}
pub trait LintContext<'tcx>: Sized {
type PassObject: LintPassObject;
fn sess(&self) -> &Session;
fn lints(&self) -> &LintStore;
fn lint_sess(&self) -> &LintSession<'tcx, Self::PassObject>;
fn lint_sess_mut(&mut self) -> &mut LintSession<'tcx, Self::PassObject>;
fn enter_attrs(&mut self, attrs: &'tcx [ast::Attribute]);
fn exit_attrs(&mut self, attrs: &'tcx [ast::Attribute]);
fn lookup_and_emit<S: Into<MultiSpan>>(&self,
lint: &'static Lint,
span: Option<S>,
msg: &str) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.lookup(lint, span, msg).emit();
}
2016-10-12 17:00:30 -04:00
fn lookup<S: Into<MultiSpan>>(&self,
lint: &'static Lint,
span: Option<S>,
msg: &str)
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
-> DiagnosticBuilder;
2015-12-21 10:00:43 +13:00
/// Emit a lint at the appropriate level, for a particular span.
fn span_lint<S: Into<MultiSpan>>(&self, lint: &'static Lint, span: S, msg: &str) {
self.lookup_and_emit(lint, Some(span), msg);
}
2016-10-12 17:00:30 -04:00
fn struct_span_lint<S: Into<MultiSpan>>(&self,
lint: &'static Lint,
span: S,
msg: &str)
-> DiagnosticBuilder {
2015-12-21 10:00:43 +13:00
self.lookup(lint, Some(span), msg)
}
/// Emit a lint and note at the appropriate level, for a particular span.
fn span_lint_note(&self, lint: &'static Lint, span: Span, msg: &str,
note_span: Span, note: &str) {
2015-12-23 19:27:20 +13:00
let mut err = self.lookup(lint, Some(span), msg);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
if note_span == span {
err.note(note);
} else {
err.span_note(note_span, note);
}
2015-12-21 10:00:43 +13:00
err.emit();
}
/// Emit a lint and help at the appropriate level, for a particular span.
fn span_lint_help(&self, lint: &'static Lint, span: Span,
msg: &str, help: &str) {
2015-12-23 19:27:20 +13:00
let mut err = self.lookup(lint, Some(span), msg);
self.span_lint(lint, span, msg);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
err.span_help(span, help);
2015-12-21 10:00:43 +13:00
err.emit();
}
/// Emit a lint at the appropriate level, with no associated span.
fn lint(&self, lint: &'static Lint, msg: &str) {
self.lookup_and_emit(lint, None as Option<Span>, msg);
}
/// Merge the lints specified by any lint attributes into the
/// current lint context, call the provided function, then reset the
/// lints in effect to their previous state.
2014-12-08 20:26:43 -05:00
fn with_lint_attrs<F>(&mut self,
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
id: ast::NodeId,
attrs: &'tcx [ast::Attribute],
f: F)
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
where F: FnOnce(&mut Self);
}
impl<'a> EarlyContext<'a> {
fn new(sess: &'a Session,
krate: &'a ast::Crate) -> EarlyContext<'a> {
EarlyContext {
sess,
krate,
lint_sess: LintSession::new(&sess.lint_store),
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
builder: LintLevelSets::builder(sess),
buffered: sess.buffered_lints.borrow_mut().take().unwrap(),
}
}
fn check_id(&mut self, id: ast::NodeId) {
for early_lint in self.buffered.take(id) {
self.lookup_and_emit(early_lint.lint_id.lint,
Some(early_lint.span.clone()),
&early_lint.msg);
}
}
}
impl<'a, 'tcx> LintContext<'tcx> for LateContext<'a, 'tcx> {
type PassObject = LateLintPassObject;
/// Get the overall compiler `Session` object.
fn sess(&self) -> &Session {
&self.tcx.sess
}
fn lints(&self) -> &LintStore {
&*self.lint_sess.lints
}
fn lint_sess(&self) -> &LintSession<'tcx, Self::PassObject> {
&self.lint_sess
}
fn lint_sess_mut(&mut self) -> &mut LintSession<'tcx, Self::PassObject> {
&mut self.lint_sess
}
fn enter_attrs(&mut self, attrs: &'tcx [ast::Attribute]) {
debug!("late context: enter_attrs({:?})", attrs);
run_lints!(self, enter_lint_attrs, late_passes, attrs);
}
fn exit_attrs(&mut self, attrs: &'tcx [ast::Attribute]) {
debug!("late context: exit_attrs({:?})", attrs);
run_lints!(self, exit_lint_attrs, late_passes, attrs);
}
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
fn lookup<S: Into<MultiSpan>>(&self,
lint: &'static Lint,
span: Option<S>,
msg: &str)
-> DiagnosticBuilder {
let id = self.last_ast_node_with_lint_attrs;
match span {
Some(s) => self.tcx.struct_span_lint_node(lint, id, s, msg),
None => self.tcx.struct_lint_node(lint, id, msg),
}
}
fn with_lint_attrs<F>(&mut self,
id: ast::NodeId,
attrs: &'tcx [ast::Attribute],
f: F)
where F: FnOnce(&mut Self)
{
let prev = self.last_ast_node_with_lint_attrs;
self.last_ast_node_with_lint_attrs = id;
self.enter_attrs(attrs);
f(self);
self.exit_attrs(attrs);
self.last_ast_node_with_lint_attrs = prev;
}
}
impl<'a> LintContext<'a> for EarlyContext<'a> {
type PassObject = EarlyLintPassObject;
/// Get the overall compiler `Session` object.
fn sess(&self) -> &Session {
&self.sess
}
fn lints(&self) -> &LintStore {
&*self.lint_sess.lints
}
fn lint_sess(&self) -> &LintSession<'a, Self::PassObject> {
&self.lint_sess
}
fn lint_sess_mut(&mut self) -> &mut LintSession<'a, Self::PassObject> {
&mut self.lint_sess
}
fn enter_attrs(&mut self, attrs: &'a [ast::Attribute]) {
debug!("early context: enter_attrs({:?})", attrs);
run_lints!(self, enter_lint_attrs, early_passes, attrs);
}
fn exit_attrs(&mut self, attrs: &'a [ast::Attribute]) {
debug!("early context: exit_attrs({:?})", attrs);
run_lints!(self, exit_lint_attrs, early_passes, attrs);
}
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
fn lookup<S: Into<MultiSpan>>(&self,
lint: &'static Lint,
span: Option<S>,
msg: &str)
-> DiagnosticBuilder {
self.builder.struct_lint(lint, span.map(|s| s.into()), msg)
}
fn with_lint_attrs<F>(&mut self,
id: ast::NodeId,
attrs: &'a [ast::Attribute],
f: F)
where F: FnOnce(&mut Self)
{
let push = self.builder.push(attrs);
self.check_id(id);
self.enter_attrs(attrs);
f(self);
self.exit_attrs(attrs);
self.builder.pop(push);
}
}
impl<'a, 'tcx> LateContext<'a, 'tcx> {
fn with_param_env<F>(&mut self, id: ast::NodeId, f: F)
where F: FnOnce(&mut Self),
{
let old_param_env = self.param_env;
self.param_env = self.tcx.param_env(self.tcx.hir.local_def_id(id));
f(self);
self.param_env = old_param_env;
}
}
impl<'a, 'tcx> hir_visit::Visitor<'tcx> for LateContext<'a, 'tcx> {
/// Because lints are scoped lexically, we want to walk nested
/// items in the context of the outer item, so enable
/// deep-walking.
fn nested_visit_map<'this>(&'this mut self) -> hir_visit::NestedVisitorMap<'this, 'tcx> {
hir_visit::NestedVisitorMap::All(&self.tcx.hir)
}
fn visit_nested_body(&mut self, body: hir::BodyId) {
let old_tables = self.tables;
self.tables = self.tcx.body_tables(body);
let body = self.tcx.hir.body(body);
self.visit_body(body);
self.tables = old_tables;
}
2017-03-07 12:51:09 +01:00
fn visit_body(&mut self, body: &'tcx hir::Body) {
run_lints!(self, check_body, late_passes, body);
hir_visit::walk_body(self, body);
run_lints!(self, check_body_post, late_passes, body);
}
fn visit_item(&mut self, it: &'tcx hir::Item) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(it.id, &it.attrs, |cx| {
cx.with_param_env(it.id, |cx| {
run_lints!(cx, check_item, late_passes, it);
hir_visit::walk_item(cx, it);
run_lints!(cx, check_item_post, late_passes, it);
});
})
}
fn visit_foreign_item(&mut self, it: &'tcx hir::ForeignItem) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(it.id, &it.attrs, |cx| {
cx.with_param_env(it.id, |cx| {
run_lints!(cx, check_foreign_item, late_passes, it);
hir_visit::walk_foreign_item(cx, it);
run_lints!(cx, check_foreign_item_post, late_passes, it);
});
})
}
fn visit_pat(&mut self, p: &'tcx hir::Pat) {
run_lints!(self, check_pat, late_passes, p);
hir_visit::walk_pat(self, p);
}
fn visit_expr(&mut self, e: &'tcx hir::Expr) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(e.id, &e.attrs, |cx| {
run_lints!(cx, check_expr, late_passes, e);
hir_visit::walk_expr(cx, e);
run_lints!(cx, check_expr_post, late_passes, e);
})
}
fn visit_stmt(&mut self, s: &'tcx hir::Stmt) {
// statement attributes are actually just attributes on one of
// - item
// - local
// - expression
// so we keep track of lint levels there
run_lints!(self, check_stmt, late_passes, s);
hir_visit::walk_stmt(self, s);
}
fn visit_fn(&mut self, fk: hir_visit::FnKind<'tcx>, decl: &'tcx hir::FnDecl,
body_id: hir::BodyId, span: Span, id: ast::NodeId) {
// Wrap in tables here, not just in visit_nested_body,
// in order for `check_fn` to be able to use them.
let old_tables = self.tables;
self.tables = self.tcx.body_tables(body_id);
let body = self.tcx.hir.body(body_id);
run_lints!(self, check_fn, late_passes, fk, decl, body, span, id);
2016-10-28 22:58:32 +02:00
hir_visit::walk_fn(self, fk, decl, body_id, span, id);
run_lints!(self, check_fn_post, late_passes, fk, decl, body, span, id);
self.tables = old_tables;
}
fn visit_variant_data(&mut self,
s: &'tcx hir::VariantData,
2015-09-20 03:34:12 +03:00
name: ast::Name,
g: &'tcx hir::Generics,
item_id: ast::NodeId,
_: Span) {
run_lints!(self, check_struct_def, late_passes, s, name, g, item_id);
hir_visit::walk_struct_def(self, s);
run_lints!(self, check_struct_def_post, late_passes, s, name, g, item_id);
}
fn visit_struct_field(&mut self, s: &'tcx hir::StructField) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(s.id, &s.attrs, |cx| {
run_lints!(cx, check_struct_field, late_passes, s);
hir_visit::walk_struct_field(cx, s);
})
}
fn visit_variant(&mut self,
v: &'tcx hir::Variant,
g: &'tcx hir::Generics,
item_id: ast::NodeId) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(v.node.data.id(), &v.node.attrs, |cx| {
run_lints!(cx, check_variant, late_passes, v, g);
hir_visit::walk_variant(cx, v, g, item_id);
run_lints!(cx, check_variant_post, late_passes, v, g);
})
}
fn visit_ty(&mut self, t: &'tcx hir::Ty) {
run_lints!(self, check_ty, late_passes, t);
hir_visit::walk_ty(self, t);
}
2015-09-20 03:34:12 +03:00
fn visit_name(&mut self, sp: Span, name: ast::Name) {
run_lints!(self, check_name, late_passes, sp, name);
}
fn visit_mod(&mut self, m: &'tcx hir::Mod, s: Span, n: ast::NodeId) {
run_lints!(self, check_mod, late_passes, m, s, n);
hir_visit::walk_mod(self, m, n);
run_lints!(self, check_mod_post, late_passes, m, s, n);
}
fn visit_local(&mut self, l: &'tcx hir::Local) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(l.id, &l.attrs, |cx| {
run_lints!(cx, check_local, late_passes, l);
hir_visit::walk_local(cx, l);
})
}
fn visit_block(&mut self, b: &'tcx hir::Block) {
run_lints!(self, check_block, late_passes, b);
hir_visit::walk_block(self, b);
run_lints!(self, check_block_post, late_passes, b);
}
fn visit_arm(&mut self, a: &'tcx hir::Arm) {
run_lints!(self, check_arm, late_passes, a);
hir_visit::walk_arm(self, a);
}
fn visit_decl(&mut self, d: &'tcx hir::Decl) {
run_lints!(self, check_decl, late_passes, d);
hir_visit::walk_decl(self, d);
}
fn visit_generics(&mut self, g: &'tcx hir::Generics) {
run_lints!(self, check_generics, late_passes, g);
hir_visit::walk_generics(self, g);
}
fn visit_trait_item(&mut self, trait_item: &'tcx hir::TraitItem) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(trait_item.id, &trait_item.attrs, |cx| {
cx.with_param_env(trait_item.id, |cx| {
run_lints!(cx, check_trait_item, late_passes, trait_item);
hir_visit::walk_trait_item(cx, trait_item);
run_lints!(cx, check_trait_item_post, late_passes, trait_item);
});
});
}
fn visit_impl_item(&mut self, impl_item: &'tcx hir::ImplItem) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(impl_item.id, &impl_item.attrs, |cx| {
cx.with_param_env(impl_item.id, |cx| {
run_lints!(cx, check_impl_item, late_passes, impl_item);
hir_visit::walk_impl_item(cx, impl_item);
run_lints!(cx, check_impl_item_post, late_passes, impl_item);
});
});
}
fn visit_lifetime(&mut self, lt: &'tcx hir::Lifetime) {
run_lints!(self, check_lifetime, late_passes, lt);
Fix lint attributes on non-item nodes. Currently, late lint checking uses two HIR visitors: LateContext and IdVisitor. IdVisitor only overrides visit_id, and for each node searches for builtin lints previously added to the session; LateContext overrides a number of methods, and runs late lints. When LateContext encounters an item, it first has IdVisitor walk everything in it except nested items (OnlyBodies), then recurses into it itself - i.e. there are two separate walks. Aside from apparently being unnecessary, this separation prevents lint attributes (allow/deny/warn) on non-item HIR nodes from working properly. Test case: // generates warning without this change fn main() { #[allow(unreachable_code)] loop { break; break; } } LateContext contains logic to merge attributes seen into the current lint settings while walking (with_lint_attrs), but IdVisitor does not. So such attributes will affect late lints (because they are called from LateContext), and if the node contains any items within it, they will affect builtin lints within those items (because that IdVisitor is run while LateContext is within the attributed node), but otherwise the attributes will be ignored for builtin lints. This change simply removes IdVisitor and moves its visit_id into LateContext itself. Hopefully this doesn't break anything... Also added walk calls to visit_lifetime and visit_lifetime_def respectively, so visit_lifetime_def will recurse into the lifetime and visit_lifetime will recurse into the name. In principle this could confuse lint plugins. This is "necessary" because walk_lifetime calls visit_id on the lifetime; of course, an alternative would be directly calling visit_id (which would require manually iterating over the lifetimes in visit_lifetime_def), but that seems less clean.
2017-01-03 20:40:29 +01:00
hir_visit::walk_lifetime(self, lt);
}
fn visit_lifetime_def(&mut self, lt: &'tcx hir::LifetimeDef) {
run_lints!(self, check_lifetime_def, late_passes, lt);
Fix lint attributes on non-item nodes. Currently, late lint checking uses two HIR visitors: LateContext and IdVisitor. IdVisitor only overrides visit_id, and for each node searches for builtin lints previously added to the session; LateContext overrides a number of methods, and runs late lints. When LateContext encounters an item, it first has IdVisitor walk everything in it except nested items (OnlyBodies), then recurses into it itself - i.e. there are two separate walks. Aside from apparently being unnecessary, this separation prevents lint attributes (allow/deny/warn) on non-item HIR nodes from working properly. Test case: // generates warning without this change fn main() { #[allow(unreachable_code)] loop { break; break; } } LateContext contains logic to merge attributes seen into the current lint settings while walking (with_lint_attrs), but IdVisitor does not. So such attributes will affect late lints (because they are called from LateContext), and if the node contains any items within it, they will affect builtin lints within those items (because that IdVisitor is run while LateContext is within the attributed node), but otherwise the attributes will be ignored for builtin lints. This change simply removes IdVisitor and moves its visit_id into LateContext itself. Hopefully this doesn't break anything... Also added walk calls to visit_lifetime and visit_lifetime_def respectively, so visit_lifetime_def will recurse into the lifetime and visit_lifetime will recurse into the name. In principle this could confuse lint plugins. This is "necessary" because walk_lifetime calls visit_id on the lifetime; of course, an alternative would be directly calling visit_id (which would require manually iterating over the lifetimes in visit_lifetime_def), but that seems less clean.
2017-01-03 20:40:29 +01:00
hir_visit::walk_lifetime_def(self, lt);
}
fn visit_path(&mut self, p: &'tcx hir::Path, id: ast::NodeId) {
run_lints!(self, check_path, late_passes, p, id);
hir_visit::walk_path(self, p);
}
fn visit_attribute(&mut self, attr: &'tcx ast::Attribute) {
run_lints!(self, check_attribute, late_passes, attr);
}
}
impl<'a> ast_visit::Visitor<'a> for EarlyContext<'a> {
fn visit_item(&mut self, it: &'a ast::Item) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(it.id, &it.attrs, |cx| {
run_lints!(cx, check_item, early_passes, it);
ast_visit::walk_item(cx, it);
2016-02-12 18:21:43 +01:00
run_lints!(cx, check_item_post, early_passes, it);
})
}
fn visit_foreign_item(&mut self, it: &'a ast::ForeignItem) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(it.id, &it.attrs, |cx| {
run_lints!(cx, check_foreign_item, early_passes, it);
ast_visit::walk_foreign_item(cx, it);
run_lints!(cx, check_foreign_item_post, early_passes, it);
})
}
fn visit_pat(&mut self, p: &'a ast::Pat) {
run_lints!(self, check_pat, early_passes, p);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(p.id);
ast_visit::walk_pat(self, p);
}
fn visit_expr(&mut self, e: &'a ast::Expr) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(e.id, &e.attrs, |cx| {
run_lints!(cx, check_expr, early_passes, e);
ast_visit::walk_expr(cx, e);
})
}
fn visit_stmt(&mut self, s: &'a ast::Stmt) {
run_lints!(self, check_stmt, early_passes, s);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(s.id);
ast_visit::walk_stmt(self, s);
}
fn visit_fn(&mut self, fk: ast_visit::FnKind<'a>, decl: &'a ast::FnDecl,
span: Span, id: ast::NodeId) {
run_lints!(self, check_fn, early_passes, fk, decl, span, id);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(id);
ast_visit::walk_fn(self, fk, decl, span);
run_lints!(self, check_fn_post, early_passes, fk, decl, span, id);
}
fn visit_variant_data(&mut self,
s: &'a ast::VariantData,
ident: ast::Ident,
g: &'a ast::Generics,
item_id: ast::NodeId,
_: Span) {
run_lints!(self, check_struct_def, early_passes, s, ident, g, item_id);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(s.id());
ast_visit::walk_struct_def(self, s);
run_lints!(self, check_struct_def_post, early_passes, s, ident, g, item_id);
}
fn visit_struct_field(&mut self, s: &'a ast::StructField) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(s.id, &s.attrs, |cx| {
run_lints!(cx, check_struct_field, early_passes, s);
ast_visit::walk_struct_field(cx, s);
})
}
fn visit_variant(&mut self, v: &'a ast::Variant, g: &'a ast::Generics, item_id: ast::NodeId) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(item_id, &v.node.attrs, |cx| {
run_lints!(cx, check_variant, early_passes, v, g);
ast_visit::walk_variant(cx, v, g, item_id);
run_lints!(cx, check_variant_post, early_passes, v, g);
})
}
fn visit_ty(&mut self, t: &'a ast::Ty) {
run_lints!(self, check_ty, early_passes, t);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(t.id);
ast_visit::walk_ty(self, t);
}
fn visit_ident(&mut self, sp: Span, id: ast::Ident) {
run_lints!(self, check_ident, early_passes, sp, id);
}
2017-05-12 07:15:29 +12:00
fn visit_mod(&mut self, m: &'a ast::Mod, s: Span, _a: &[ast::Attribute], n: ast::NodeId) {
run_lints!(self, check_mod, early_passes, m, s, n);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(n);
ast_visit::walk_mod(self, m);
run_lints!(self, check_mod_post, early_passes, m, s, n);
}
fn visit_local(&mut self, l: &'a ast::Local) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(l.id, &l.attrs, |cx| {
run_lints!(cx, check_local, early_passes, l);
ast_visit::walk_local(cx, l);
})
}
fn visit_block(&mut self, b: &'a ast::Block) {
run_lints!(self, check_block, early_passes, b);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(b.id);
ast_visit::walk_block(self, b);
run_lints!(self, check_block_post, early_passes, b);
}
fn visit_arm(&mut self, a: &'a ast::Arm) {
run_lints!(self, check_arm, early_passes, a);
ast_visit::walk_arm(self, a);
}
fn visit_expr_post(&mut self, e: &'a ast::Expr) {
run_lints!(self, check_expr_post, early_passes, e);
}
fn visit_generics(&mut self, g: &'a ast::Generics) {
run_lints!(self, check_generics, early_passes, g);
ast_visit::walk_generics(self, g);
}
fn visit_trait_item(&mut self, trait_item: &'a ast::TraitItem) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(trait_item.id, &trait_item.attrs, |cx| {
run_lints!(cx, check_trait_item, early_passes, trait_item);
ast_visit::walk_trait_item(cx, trait_item);
run_lints!(cx, check_trait_item_post, early_passes, trait_item);
});
}
fn visit_impl_item(&mut self, impl_item: &'a ast::ImplItem) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.with_lint_attrs(impl_item.id, &impl_item.attrs, |cx| {
run_lints!(cx, check_impl_item, early_passes, impl_item);
ast_visit::walk_impl_item(cx, impl_item);
run_lints!(cx, check_impl_item_post, early_passes, impl_item);
});
}
fn visit_lifetime(&mut self, lt: &'a ast::Lifetime) {
run_lints!(self, check_lifetime, early_passes, lt);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(lt.id);
}
fn visit_lifetime_def(&mut self, lt: &'a ast::LifetimeDef) {
run_lints!(self, check_lifetime_def, early_passes, lt);
}
fn visit_path(&mut self, p: &'a ast::Path, id: ast::NodeId) {
run_lints!(self, check_path, early_passes, p, id);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(id);
ast_visit::walk_path(self, p);
}
fn visit_path_list_item(&mut self, prefix: &'a ast::Path, item: &'a ast::PathListItem) {
2015-09-17 15:19:24 +03:00
run_lints!(self, check_path_list_item, early_passes, item);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(item.node.id);
2015-09-17 15:19:24 +03:00
ast_visit::walk_path_list_item(self, prefix, item);
}
fn visit_attribute(&mut self, attr: &'a ast::Attribute) {
run_lints!(self, check_attribute, early_passes, attr);
}
2017-03-17 21:58:48 +00:00
fn visit_mac_def(&mut self, _mac: &'a ast::MacroDef, id: ast::NodeId) {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
self.check_id(id);
}
}
/// Perform lint checking on a crate.
///
/// Consumes the `lint_store` field of the `Session`.
pub fn check_crate<'a, 'tcx>(tcx: TyCtxt<'a, 'tcx, 'tcx>) {
let access_levels = &tcx.privacy_access_levels(LOCAL_CRATE);
let krate = tcx.hir.krate();
let mut cx = LateContext {
tcx,
tables: &ty::TypeckTables::empty(None),
param_env: ty::ParamEnv::empty(Reveal::UserFacing),
access_levels,
lint_sess: LintSession::new(&tcx.sess.lint_store),
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
last_ast_node_with_lint_attrs: ast::CRATE_NODE_ID,
};
// Visit the whole crate.
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
cx.with_lint_attrs(ast::CRATE_NODE_ID, &krate.attrs, |cx| {
// since the root module isn't visited as an item (because it isn't an
// item), warn for it here.
run_lints!(cx, check_crate, late_passes, krate);
hir_visit::walk_crate(cx, krate);
run_lints!(cx, check_crate_post, late_passes, krate);
});
// Put the lint store levels and passes back in the session.
cx.lint_sess.restore(&tcx.sess.lint_store);
}
pub fn check_ast_crate(sess: &Session, krate: &ast::Crate) {
let mut cx = EarlyContext::new(sess, krate);
// Visit the whole crate.
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
cx.with_lint_attrs(ast::CRATE_NODE_ID, &krate.attrs, |cx| {
// since the root module isn't visited as an item (because it isn't an
// item), warn for it here.
run_lints!(cx, check_crate, early_passes, krate);
ast_visit::walk_crate(cx, krate);
run_lints!(cx, check_crate_post, early_passes, krate);
});
// Put the lint store levels and passes back in the session.
cx.lint_sess.restore(&sess.lint_store);
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
// Emit all buffered lints from early on in the session now that we've
// calculated the lint levels for all AST nodes.
for (_id, lints) in cx.buffered.map {
for early_lint in lints {
span_bug!(early_lint.span, "failed to process bufferd lint here");
}
}
}
impl Encodable for LintId {
fn encode<S: Encoder>(&self, s: &mut S) -> Result<(), S::Error> {
s.emit_str(&self.lint.name.to_lowercase())
}
}
impl Decodable for LintId {
#[inline]
fn decode<D: Decoder>(d: &mut D) -> Result<LintId, D::Error> {
let s = d.read_str()?;
ty::tls::with(|tcx| {
rustc: Rearchitect lints to be emitted more eagerly In preparation for incremental compilation this commit refactors the lint handling infrastructure in the compiler to be more "eager" and overall more incremental-friendly. Many passes of the compiler can emit lints at various points but before this commit all lints were buffered in a table to be emitted at the very end of compilation. This commit changes these lints to be emitted immediately during compilation using pre-calculated lint level-related data structures. Linting today is split into two phases, one set of "early" lints run on the `syntax::ast` and a "late" set of lints run on the HIR. This commit moves the "early" lints to running as late as possible in compilation, just before HIR lowering. This notably means that we're catching resolve-related lints just before HIR lowering. The early linting remains a pass very similar to how it was before, maintaining context of the current lint level as it walks the tree. Post-HIR, however, linting is structured as a method on the `TyCtxt` which transitively executes a query to calculate lint levels. Each request to lint on a `TyCtxt` will query the entire crate's 'lint level data structure' and then go from there about whether the lint should be emitted or not. The query depends on the entire HIR crate but should be very quick to calculate (just a quick walk of the HIR) and the red-green system should notice that the lint level data structure rarely changes, and should hopefully preserve incrementality. Overall this resulted in a pretty big change to the test suite now that lints are emitted much earlier in compilation (on-demand vs only at the end). This in turn necessitated the addition of many `#![allow(warnings)]` directives throughout the compile-fail test suite and a number of updates to the UI test suite.
2017-07-26 21:51:09 -07:00
match tcx.sess.lint_store.borrow().find_lints(&s) {
Ok(ids) => {
if ids.len() != 0 {
panic!("invalid lint-id `{}`", s);
}
Ok(ids[0])
}
Err(_) => panic!("invalid lint-id `{}`", s),
}
})
}
}