import driver::session; import driver::session::session; import middle::ty; import syntax::{ast, ast_util, visit}; import syntax::attr; import syntax::codemap::span; import std::map::{map,hashmap,int_hash,hash_from_strs}; import std::smallintmap::{map,smallintmap}; import io::WriterUtil; import util::ppaux::{ty_to_str}; import middle::pat_util::{pat_bindings}; import syntax::ast_util::{path_to_ident}; import syntax::print::pprust::{expr_to_str, mode_to_str, pat_to_str}; export lint, ctypes, unused_imports, while_true, path_statement, old_vecs; export unrecognized_lint, non_implicitly_copyable_typarams; export vecs_implicitly_copyable, implicit_copies; export level, allow, warn, deny, forbid; export lint_dict, get_lint_dict, level_to_str; export get_lint_level, get_lint_settings_level; export check_crate, build_settings_crate, mk_lint_settings; export lint_settings; /** * A 'lint' check is a kind of miscellaneous constraint that a user _might_ * want to enforce, but might reasonably want to permit as well, on a * module-by-module basis. They contrast with static constraints enforced by * other phases of the compiler, which are generally required to hold in order * to compile the program at all. * * We also build up a table containing information about lint settings, in * order to allow other passes to take advantage of the lint attribute * infrastructure. To save space, the table is keyed by the id of /items/, not * of every expression. When an item has the default settings, the entry will * be omitted. If we start allowing lint attributes on expressions, we will * start having entries for expressions that do not share their enclosing * items settings. * * This module then, exports two passes: one that populates the lint * settings table in the session and is run early in the compile process, and * one that does a variety of lint checks, and is run late in the compile * process. */ enum lint { ctypes, unused_imports, while_true, path_statement, implicit_copies, unrecognized_lint, non_implicitly_copyable_typarams, vecs_implicitly_copyable, deprecated_mode, deprecated_pattern, non_camel_case_types, managed_heap_memory, owned_heap_memory, heap_memory, // FIXME(#3266)--make liveness warnings lintable // unused_variable, // dead_assignment } impl lint : cmp::Eq { pure fn eq(&&other: lint) -> bool { (self as uint) == (other as uint) } } fn level_to_str(lv: level) -> ~str { match lv { allow => ~"allow", warn => ~"warn", deny => ~"deny", forbid => ~"forbid" } } enum level { allow, warn, deny, forbid } impl level : cmp::Eq { pure fn eq(&&other: level) -> bool { (self as uint) == (other as uint) } } type lint_spec = @{lint: lint, desc: ~str, default: level}; type lint_dict = hashmap<~str,lint_spec>; /* Pass names should not contain a '-', as the compiler normalizes '-' to '_' in command-line flags */ fn get_lint_dict() -> lint_dict { let v = ~[ (~"ctypes", @{lint: ctypes, desc: ~"proper use of core::libc types in foreign modules", default: warn}), (~"unused_imports", @{lint: unused_imports, desc: ~"imports that are never used", default: allow}), (~"while_true", @{lint: while_true, desc: ~"suggest using loop { } instead of while(true) { }", default: warn}), (~"path_statement", @{lint: path_statement, desc: ~"path statements with no effect", default: warn}), (~"unrecognized_lint", @{lint: unrecognized_lint, desc: ~"unrecognized lint attribute", default: warn}), (~"non_implicitly_copyable_typarams", @{lint: non_implicitly_copyable_typarams, desc: ~"passing non implicitly copyable types as copy type params", default: warn}), (~"vecs_implicitly_copyable", @{lint: vecs_implicitly_copyable, desc: ~"make vecs and strs not implicitly copyable \ (only checked at top level)", default: warn}), (~"implicit_copies", @{lint: implicit_copies, desc: ~"implicit copies of non implicitly copyable data", default: warn}), (~"deprecated_mode", @{lint: deprecated_mode, desc: ~"warn about deprecated uses of modes", default: allow}), (~"deprecated_pattern", @{lint: deprecated_pattern, desc: ~"warn about deprecated uses of pattern bindings", default: allow}), (~"non_camel_case_types", @{lint: non_camel_case_types, desc: ~"types, variants and traits must have camel case names", default: allow}), (~"managed_heap_memory", @{lint: managed_heap_memory, desc: ~"use of managed (@ type) heap memory", default: allow}), (~"owned_heap_memory", @{lint: owned_heap_memory, desc: ~"use of owned (~ type) heap memory", default: allow}), (~"heap_memory", @{lint: heap_memory, desc: ~"use of any (~ type or @ type) heap memory", default: allow}), /* FIXME(#3266)--make liveness warnings lintable (~"unused_variable", @{lint: unused_variable, desc: ~"detect variables which are not used in any way", default: warn}), (~"dead_assignment", @{lint: dead_assignment, desc: ~"detect assignments that will never be read", default: warn}), */ ]; hash_from_strs(v) } // This is a highly not-optimal set of data structure decisions. type lint_modes = smallintmap; type lint_mode_map = hashmap; // settings_map maps node ids of items with non-default lint settings // to their settings; default_settings contains the settings for everything // not in the map. type lint_settings = { default_settings: lint_modes, settings_map: lint_mode_map }; fn mk_lint_settings() -> lint_settings { {default_settings: std::smallintmap::mk(), settings_map: int_hash()} } fn get_lint_level(modes: lint_modes, lint: lint) -> level { match modes.find(lint as uint) { Some(c) => c, None => allow } } fn get_lint_settings_level(settings: lint_settings, lint_mode: lint, _expr_id: ast::node_id, item_id: ast::node_id) -> level { match settings.settings_map.find(item_id) { Some(modes) => get_lint_level(modes, lint_mode), None => get_lint_level(settings.default_settings, lint_mode) } } // This is kind of unfortunate. It should be somewhere else, or we should use // a persistent data structure... fn clone_lint_modes(modes: lint_modes) -> lint_modes { std::smallintmap::smallintmap_(@{v: copy modes.v}) } type ctxt_ = {dict: lint_dict, curr: lint_modes, is_default: bool, sess: session}; enum ctxt { ctxt_(ctxt_) } impl ctxt { fn get_level(lint: lint) -> level { get_lint_level(self.curr, lint) } fn set_level(lint: lint, level: level) { if level == allow { self.curr.remove(lint as uint); } else { self.curr.insert(lint as uint, level); } } fn span_lint(level: level, span: span, msg: ~str) { self.sess.span_lint_level(level, 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. */ fn with_lint_attrs(attrs: ~[ast::attribute], f: fn(ctxt)) { let mut new_ctxt = self; let mut triples = ~[]; for [allow, warn, deny, forbid].each |level| { let level_name = level_to_str(level); let metas = attr::attr_metas(attr::find_attrs_by_name(attrs, level_name)); for metas.each |meta| { match meta.node { ast::meta_list(_, metas) => { for metas.each |meta| { match meta.node { ast::meta_word(lintname) => { vec::push(triples, (meta, level, lintname)); } _ => { self.sess.span_err( meta.span, ~"malformed lint attribute"); } } } } _ => { self.sess.span_err(meta.span, ~"malformed lint attribute"); } } } } for triples.each |pair| { let (meta, level, lintname) = pair; match self.dict.find(lintname) { None => { self.span_lint( new_ctxt.get_level(unrecognized_lint), meta.span, fmt!("unknown `%s` attribute: `%s`", level_to_str(level), lintname)); } Some(lint) => { if new_ctxt.get_level(lint.lint) == forbid && level != forbid { self.span_lint( forbid, meta.span, fmt!("%s(%s) overruled by outer forbid(%s)", level_to_str(level), lintname, lintname)); } // we do multiple unneeded copies of the // map if many attributes are set, but // this shouldn't actually be a problem... let c = clone_lint_modes(new_ctxt.curr); new_ctxt = ctxt_({is_default: false, curr: c, with *new_ctxt}); new_ctxt.set_level(lint.lint, level); } } } f(new_ctxt); } } fn build_settings_item(i: @ast::item, &&cx: ctxt, v: visit::vt) { do cx.with_lint_attrs(i.attrs) |cx| { if !cx.is_default { cx.sess.lint_settings.settings_map.insert(i.id, cx.curr); } visit::visit_item(i, cx, v); } } fn build_settings_crate(sess: session::session, crate: @ast::crate) { let cx = ctxt_({dict: get_lint_dict(), curr: std::smallintmap::mk(), is_default: true, sess: sess}); // Install defaults. for cx.dict.each |_k, spec| { cx.set_level(spec.lint, spec.default); } // Install command-line options, overriding defaults. for sess.opts.lint_opts.each |pair| { let (lint,level) = pair; cx.set_level(lint, level); } do cx.with_lint_attrs(crate.node.attrs) |cx| { // Copy out the default settings for cx.curr.each |k, v| { sess.lint_settings.default_settings.insert(k, v); } let cx = ctxt_({is_default: true with *cx}); let visit = visit::mk_vt(@{ visit_item: build_settings_item with *visit::default_visitor() }); visit::visit_crate(*crate, cx, visit); } sess.abort_if_errors(); } fn check_item(i: @ast::item, cx: ty::ctxt) { check_item_ctypes(cx, i); check_item_while_true(cx, i); check_item_path_statement(cx, i); check_item_non_camel_case_types(cx, i); check_item_heap(cx, i); } // Take a visitor, and modify it so that it will not proceed past subitems. // This is used to make the simple visitors used for the lint passes // not traverse into subitems, since that is handled by the outer // lint visitor. fn item_stopping_visitor(v: visit::vt) -> visit::vt { visit::mk_vt(@{visit_item: |_i, _e, _v| { } with **v}) } fn check_item_while_true(cx: ty::ctxt, it: @ast::item) { let visit = item_stopping_visitor(visit::mk_simple_visitor(@{ visit_expr: fn@(e: @ast::expr) { match e.node { ast::expr_while(cond, _) => { match cond.node { ast::expr_lit(@{node: ast::lit_bool(true),_}) => { cx.sess.span_lint( while_true, e.id, it.id, e.span, ~"denote infinite loops with loop { ... }"); } _ => () } } _ => () } } with *visit::default_simple_visitor() })); visit::visit_item(it, (), visit); } fn check_item_ctypes(cx: ty::ctxt, it: @ast::item) { fn check_foreign_fn(cx: ty::ctxt, fn_id: ast::node_id, decl: ast::fn_decl) { let tys = vec::map(decl.inputs, |a| a.ty ); for vec::each(vec::append_one(tys, decl.output)) |ty| { match ty.node { ast::ty_path(_, id) => { match cx.def_map.get(id) { ast::def_prim_ty(ast::ty_int(ast::ty_i)) => { cx.sess.span_lint( ctypes, id, fn_id, ty.span, ~"found rust type `int` in foreign module, while \ libc::c_int or libc::c_long should be used"); } ast::def_prim_ty(ast::ty_uint(ast::ty_u)) => { cx.sess.span_lint( ctypes, id, fn_id, ty.span, ~"found rust type `uint` in foreign module, while \ libc::c_uint or libc::c_ulong should be used"); } _ => () } } _ => () } } } match it.node { ast::item_foreign_mod(nmod) if attr::foreign_abi(it.attrs) != either::Right(ast::foreign_abi_rust_intrinsic) => { for nmod.items.each |ni| { match ni.node { ast::foreign_item_fn(decl, _, _) => { check_foreign_fn(cx, it.id, decl); } ast::foreign_item_const(*) => {} // XXX: Not implemented. } } } _ => {/* nothing to do */ } } } fn check_item_heap(cx: ty::ctxt, it: @ast::item) { fn check_type_for_lint(cx: ty::ctxt, lint: lint, node: ast::node_id, item: ast::node_id, span: span, ty: ty::t) { if get_lint_settings_level(cx.sess.lint_settings, lint, node, item) != allow { let mut n_box = 0; let mut n_uniq = 0; ty::fold_ty(cx, ty, |t| { match ty::get(t).struct { ty::ty_box(_) => n_box += 1, ty::ty_uniq(_) => n_uniq += 1, _ => () }; t }); if (n_uniq > 0 && lint != managed_heap_memory) { let s = ty_to_str(cx, ty); let m = ~"type uses owned (~ type) pointers: " + s; cx.sess.span_lint(lint, node, item, span, m); } if (n_box > 0 && lint != owned_heap_memory) { let s = ty_to_str(cx, ty); let m = ~"type uses managed (@ type) pointers: " + s; cx.sess.span_lint(lint, node, item, span, m); } } } fn check_type(cx: ty::ctxt, node: ast::node_id, item: ast::node_id, span: span, ty: ty::t) { for [managed_heap_memory, owned_heap_memory, heap_memory].each |lint| { check_type_for_lint(cx, lint, node, item, span, ty); } } match it.node { ast::item_fn(*) | ast::item_ty(*) | ast::item_enum(*) | ast::item_class(*) | ast::item_trait(*) => check_type(cx, it.id, it.id, it.span, ty::node_id_to_type(cx, it.id)), _ => () } let visit = item_stopping_visitor(visit::mk_simple_visitor(@{ visit_expr: fn@(e: @ast::expr) { let ty = ty::expr_ty(cx, e); check_type(cx, e.id, it.id, e.span, ty); } with *visit::default_simple_visitor() })); visit::visit_item(it, (), visit); } fn check_item_path_statement(cx: ty::ctxt, it: @ast::item) { let visit = item_stopping_visitor(visit::mk_simple_visitor(@{ visit_stmt: fn@(s: @ast::stmt) { match s.node { ast::stmt_semi(@{id: id, callee_id: _, node: ast::expr_path(_), span: _}, _) => { cx.sess.span_lint( path_statement, id, it.id, s.span, ~"path statement with no effect"); } _ => () } } with *visit::default_simple_visitor() })); visit::visit_item(it, (), visit); } fn check_item_non_camel_case_types(cx: ty::ctxt, it: @ast::item) { fn is_camel_case(cx: ty::ctxt, ident: ast::ident) -> bool { let ident = cx.sess.str_of(ident); assert ident.is_not_empty(); let ident = ident_without_trailing_underscores(ident); let ident = ident_without_leading_underscores(ident); char::is_uppercase(str::char_at(ident, 0)) && !ident.contains_char('_') } fn ident_without_trailing_underscores(ident: ~str) -> ~str { match str::rfind(ident, |c| c != '_') { Some(idx) => (ident).slice(0, idx + 1), None => { ident } // all underscores } } fn ident_without_leading_underscores(ident: ~str) -> ~str { match str::find(ident, |c| c != '_') { Some(idx) => ident.slice(idx, ident.len()), None => { // all underscores ident } } } fn check_case(cx: ty::ctxt, ident: ast::ident, expr_id: ast::node_id, item_id: ast::node_id, span: span) { if !is_camel_case(cx, ident) { cx.sess.span_lint( non_camel_case_types, expr_id, item_id, span, ~"type, variant, or trait must be camel case"); } } match it.node { ast::item_ty(*) | ast::item_class(*) | ast::item_trait(*) => { check_case(cx, it.ident, it.id, it.id, it.span) } ast::item_enum(enum_definition, _) => { check_case(cx, it.ident, it.id, it.id, it.span); for enum_definition.variants.each |variant| { check_case(cx, variant.node.name, variant.node.id, it.id, variant.span); } } _ => () } } fn check_pat(tcx: ty::ctxt, pat: @ast::pat) { debug!("lint check_pat pat=%s", pat_to_str(pat, tcx.sess.intr())); do pat_bindings(tcx.def_map, pat) |binding_mode, id, span, path| { match binding_mode { ast::bind_by_ref(_) | ast::bind_by_value | ast::bind_by_move => {} ast::bind_by_implicit_ref => { let pat_ty = ty::node_id_to_type(tcx, id); let kind = ty::type_kind(tcx, pat_ty); if !ty::kind_is_safe_for_default_mode(kind) { tcx.sess.span_lint( deprecated_pattern, id, id, span, fmt!("binding `%s` should use ref or copy mode", tcx.sess.str_of(path_to_ident(path)))); } } } } } fn check_fn(tcx: ty::ctxt, fk: visit::fn_kind, decl: ast::fn_decl, _body: ast::blk, span: span, id: ast::node_id) { debug!("lint check_fn fk=%? id=%?", fk, id); // don't complain about blocks, since they tend to get their modes // specified from the outside match fk { visit::fk_fn_block(*) => { return; } _ => {} } let fn_ty = ty::node_id_to_type(tcx, id); match ty::get(fn_ty).struct { ty::ty_fn(fn_ty) => { let mut counter = 0; do vec::iter2(fn_ty.inputs, decl.inputs) |arg_ty, arg_ast| { counter += 1; debug!("arg %d, ty=%s, mode=%s", counter, ty_to_str(tcx, arg_ty.ty), mode_to_str(arg_ast.mode)); match arg_ast.mode { ast::expl(ast::by_copy) => { /* always allow by-copy */ } ast::expl(_) => { tcx.sess.span_lint( deprecated_mode, id, id, span, fmt!("argument %d uses an explicit mode", counter)); } ast::infer(_) => { let kind = ty::type_kind(tcx, arg_ty.ty); if !ty::kind_is_safe_for_default_mode(kind) { tcx.sess.span_lint( deprecated_mode, id, id, span, fmt!("argument %d uses the default mode \ but shouldn't", counter)); } } } } } _ => tcx.sess.impossible_case(span, ~"check_fn: function has \ non-fn type") } } fn check_crate(tcx: ty::ctxt, crate: @ast::crate) { let v = visit::mk_simple_visitor(@{ visit_item: |it| check_item(it, tcx), visit_fn: |fk, decl, body, span, id| check_fn(tcx, fk, decl, body, span, id), visit_pat: |pat| check_pat(tcx, pat), with *visit::default_simple_visitor() }); visit::visit_crate(*crate, (), v); tcx.sess.abort_if_errors(); } // // Local Variables: // mode: rust // fill-column: 78; // indent-tabs-mode: nil // c-basic-offset: 4 // buffer-file-coding-system: utf-8-unix // End: //