Auto merge of #27893 - nikomatsakis:mir, r=<try>

This PR contains a new crate, `rustc_mir`, which implements the MIR as specified in the RFC (more or less). There are no targeted unit tests at the moment, as I didn't decide what kind of infrastructure would be best and didn't take the time to implement it. 

~~NB: In packaging up this PR, I realized that MIR construction code is not triggering for methods right now, I think it's only for fixed fns. I'll push a fix for this soon. Hopefully it doesn't stop any crates from building. :)~~ Fixed. Everything still seems to work.

However, the MIR construction code (`librustc_mir/build`) is intentionally quite distinct from the code which munges the compiler's data structures (`librustc_mir/tcx`). The interface between the two is the `HIR` trait (`librustc_mir/hir`). To avoid confusion with @nrc's work, perhaps a better name for this trait is warranted, although ultimately this trait *will* be connected to the HIR, I imagine, so in a way the name is perfect. Anyway, I'm open to suggestions. The initial motivation for this split was to allow for the MIR construction code to be unit-tested. But while I didn't end up writing unit tests (yet), I did find the split made the code immensely easier to think about, since the messiness of our existing system, with its myriad hashtables, punning, and so forth, is confined to one part, which simply transforms to a more fully explicit AST-like form. I tried to separate out the commits somewhat, but since this mostly new code, it mostly winds up coming in one fell swoop in the MIR commit.

Quick guide to the MIR crate:

- `repr.rs` defines the MIR itself; each MIR instance is parameterized by some HIR `H`
- `build/` is the MIR construction code, parameterized by a particular HIR
- `hir/` is the definition of the HIR interface
- `tcx/` is the impl of the HIR interface for the tcx
- `dump.rs` is the minimal compiler pass that invokes the HIR

One open question:

- In the HIR trait, I used exclusively struct-like variants. I found I like this more, since things have names. Should I convert the repr code?

mk/crates.mk

@@ -54,7 +54,7 @@ TARGET_CRATES := libc std flate arena term \
                  log graphviz core rbml alloc \
                  rustc_unicode rustc_bitflags \
 		 alloc_system
-RUSTC_CRATES := rustc rustc_typeck rustc_borrowck rustc_resolve rustc_driver \
+RUSTC_CRATES := rustc rustc_typeck rustc_mir rustc_borrowck rustc_resolve rustc_driver \
                 rustc_trans rustc_back rustc_llvm rustc_privacy rustc_lint \
                 rustc_data_structures rustc_front rustc_platform_intrinsics
 HOST_CRATES := syntax $(RUSTC_CRATES) rustdoc fmt_macros
@@ -70,11 +70,12 @@ DEPS_std := core libc rand alloc collections rustc_unicode \
 DEPS_graphviz := std
 DEPS_syntax := std term serialize log fmt_macros arena libc rustc_bitflags
 DEPS_rustc_driver := arena flate getopts graphviz libc rustc rustc_back rustc_borrowck \
-                     rustc_typeck rustc_resolve log syntax serialize rustc_llvm \
+                     rustc_typeck rustc_mir rustc_resolve log syntax serialize rustc_llvm \
 		             rustc_trans rustc_privacy rustc_lint rustc_front
 
 DEPS_rustc_trans := arena flate getopts graphviz libc rustc rustc_back \
 	                log syntax serialize rustc_llvm rustc_front rustc_platform_intrinsics
+DEPS_rustc_mir := rustc rustc_front syntax
 DEPS_rustc_typeck := rustc syntax rustc_front rustc_platform_intrinsics
 DEPS_rustc_borrowck := rustc rustc_front log graphviz syntax
 DEPS_rustc_resolve := rustc rustc_front log syntax

mk/main.mk

@@ -172,6 +172,18 @@ RUST_LIB_FLAGS_ST3 += -C prefer-dynamic
 # by not emitting them.
 RUSTFLAGS_STAGE0 += -Z no-landing-pads
 
+# Enable MIR to "always build" for crates where this works. This is
+# just temporary while MIR is being actively built up -- it's just a
+# poor man's unit testing infrastructure. Anyway we only want this for
+# stage1/stage2.
+define ADD_MIR_FLAG
+RUSTFLAGS1_$(1) += -Z always-build-mir
+RUSTFLAGS2_$(1) += -Z always-build-mir
+endef
+$(foreach crate,$(TARGET_CRATES),$(eval $(call ADD_MIR_FLAG,$(crate))))
+$(foreach crate,$(RUSTC_CRATES),$(eval $(call ADD_MIR_FLAG,$(crate))))
+$(foreach crate,$(HOST_CRATES),$(eval $(call ADD_MIR_FLAG,$(crate))))
+
 # platform-specific auto-configuration
 include $(CFG_SRC_DIR)mk/platform.mk
 

mk/target.mk

@@ -93,6 +93,7 @@ $$(TLIB$(1)_T_$(2)_H_$(3))/stamp.$(4): \
 		$$(LLVM_LIBDIR_RUSTFLAGS_$(2)) \
 		$$(LLVM_STDCPP_RUSTFLAGS_$(2)) \
 		$$(RUSTFLAGS_$(4)) \
+		$$(RUSTFLAGS$(1)_$(4)) \
 		$$(RUSTFLAGS$(1)_$(4)_T_$(2)) \
 		--out-dir $$(@D) \
 		-C extra-filename=-$$(CFG_FILENAME_EXTRA) \

src/libgraphviz/lib.rs

@@ -313,6 +313,13 @@ pub enum LabelText<'a> {
     /// are also the escape sequences `\l` which left-justifies the
     /// preceding line and `\r` which right-justifies it.
     EscStr(Cow<'a, str>),
+
+    /// This uses a graphviz [HTML string label][html]. The string is
+    /// printed exactly as given, but between `<` and `>`. **No
+    /// escaping is performed.**
+    ///
+    /// [html]: http://www.graphviz.org/content/node-shapes#html
+    HtmlStr(Cow<'a, str>),
 }
 
 /// The style for a node or edge.
@@ -453,6 +460,14 @@ pub trait Labeller<'a,N,E> {
     /// is a valid DOT identifier.
     fn node_id(&'a self, n: &N) -> Id<'a>;
 
+    /// Maps `n` to one of the [graphviz `shape` names][1]. If `None`
+    /// is returned, no `shape` attribute is specified.
+    ///
+    /// [1]: http://www.graphviz.org/content/node-shapes
+    fn node_shape(&'a self, _node: &N) -> Option<LabelText<'a>> {
+        None
+    }
+
     /// Maps `n` to a label that will be used in the rendered output.
     /// The label need not be unique, and may be the empty string; the
     /// default is just the output from `node_id`.
@@ -479,6 +494,16 @@ pub trait Labeller<'a,N,E> {
     }
 }
 
+/// Escape tags in such a way that it is suitable for inclusion in a
+/// Graphviz HTML label.
+pub fn escape_html(s: &str) -> String {
+    s
+        .replace("&", "&amp;")
+        .replace("\"", "&quot;")
+        .replace("<", "&lt;")
+        .replace(">", "&gt;")
+}
+
 impl<'a> LabelText<'a> {
     pub fn label<S:IntoCow<'a, str>>(s: S) -> LabelText<'a> {
         LabelStr(s.into_cow())
@@ -488,6 +513,10 @@ impl<'a> LabelText<'a> {
         EscStr(s.into_cow())
     }
 
+    pub fn html<S:IntoCow<'a, str>>(s: S) -> LabelText<'a> {
+        HtmlStr(s.into_cow())
+    }
+
     fn escape_char<F>(c: char, mut f: F) where F: FnMut(char) {
         match c {
             // not escaping \\, since Graphviz escString needs to
@@ -505,10 +534,12 @@ impl<'a> LabelText<'a> {
     }
 
     /// Renders text as string suitable for a label in a .dot file.
-    pub fn escape(&self) -> String {
+    /// This includes quotes or suitable delimeters.
+    pub fn to_dot_string(&self) -> String {
         match self {
-            &LabelStr(ref s) => s.escape_default(),
-            &EscStr(ref s) => LabelText::escape_str(&s[..]),
+            &LabelStr(ref s) => format!("\"{}\"", s.escape_default()),
+            &EscStr(ref s) => format!("\"{}\"", LabelText::escape_str(&s[..])),
+            &HtmlStr(ref s) => format!("<{}>", s),
         }
     }
 
@@ -524,6 +555,7 @@ impl<'a> LabelText<'a> {
             } else {
                 s
             },
+            HtmlStr(s) => s,
         }
     }
 
@@ -612,14 +644,15 @@ pub fn render_opts<'a, N:Clone+'a, E:Clone+'a, G:Labeller<'a,N,E>+GraphWalk<'a,N
         try!(indent(w));
         let id = g.node_id(n);
 
-        let escaped = &g.node_label(n).escape();
+        let escaped = &g.node_label(n).to_dot_string();
+        let shape;
 
         let mut text = vec![id.as_slice()];
 
         if !options.contains(&RenderOption::NoNodeLabels) {
-            text.push("[label=\"");
+            text.push("[label=");
             text.push(escaped);
-            text.push("\"]");
+            text.push("]");
         }
 
         let style = g.node_style(n);
@@ -629,12 +662,19 @@ pub fn render_opts<'a, N:Clone+'a, E:Clone+'a, G:Labeller<'a,N,E>+GraphWalk<'a,N
             text.push("\"]");
         }
 
+        if let Some(s) = g.node_shape(n) {
+            shape = s.to_dot_string();
+            text.push("[shape=");
+            text.push(&shape);
+            text.push("]");
+        }
+
         text.push(";");
         try!(writeln(w, &text));
     }
 
     for e in g.edges().iter() {
-        let escaped_label = &g.edge_label(e).escape();
+        let escaped_label = &g.edge_label(e).to_dot_string();
         try!(indent(w));
         let source = g.source(e);
         let target = g.target(e);
@@ -644,9 +684,9 @@ pub fn render_opts<'a, N:Clone+'a, E:Clone+'a, G:Labeller<'a,N,E>+GraphWalk<'a,N
         let mut text = vec![source_id.as_slice(), " -> ", target_id.as_slice()];
 
         if !options.contains(&RenderOption::NoEdgeLabels) {
-            text.push("[label=\"");
+            text.push("[label=");
             text.push(escaped_label);
-            text.push("\"]");
+            text.push("]");
         }
 
         let style = g.edge_style(e);
@@ -667,7 +707,7 @@ pub fn render_opts<'a, N:Clone+'a, E:Clone+'a, G:Labeller<'a,N,E>+GraphWalk<'a,N
 mod tests {
     use self::NodeLabels::*;
     use super::{Id, Labeller, Nodes, Edges, GraphWalk, render, Style};
-    use super::LabelText::{self, LabelStr, EscStr};
+    use super::LabelText::{self, LabelStr, EscStr, HtmlStr};
     use std::io;
     use std::io::prelude::*;
     use std::borrow::IntoCow;
@@ -805,12 +845,12 @@ mod tests {
         fn node_id(&'a self, n: &Node) -> Id<'a> { self.graph.node_id(n) }
         fn node_label(&'a self, n: &Node) -> LabelText<'a> {
             match self.graph.node_label(n) {
-                LabelStr(s) | EscStr(s) => EscStr(s),
+                LabelStr(s) | EscStr(s) | HtmlStr(s) => EscStr(s),
             }
         }
         fn edge_label(&'a self, e: & &'a Edge) -> LabelText<'a> {
             match self.graph.edge_label(e) {
-                LabelStr(s) | EscStr(s) => EscStr(s),
+                LabelStr(s) | EscStr(s) | HtmlStr(s) => EscStr(s),
             }
         }
     }

src/librustc/middle/astencode.rs

@@ -470,8 +470,8 @@ impl tr for def::Def {
           def::DefPrimTy(p) => def::DefPrimTy(p),
           def::DefTyParam(s, index, def_id, n) => def::DefTyParam(s, index, def_id.tr(dcx), n),
           def::DefUse(did) => def::DefUse(did.tr(dcx)),
-          def::DefUpvar(nid1, nid2) => {
-            def::DefUpvar(dcx.tr_id(nid1), dcx.tr_id(nid2))
+          def::DefUpvar(nid1, index, nid2) => {
+            def::DefUpvar(dcx.tr_id(nid1), index, dcx.tr_id(nid2))
           }
           def::DefStruct(did) => def::DefStruct(did.tr(dcx)),
           def::DefRegion(nid) => def::DefRegion(dcx.tr_id(nid)),

src/librustc/middle/def.rs

@@ -38,6 +38,7 @@ pub enum Def {
     DefTyParam(ParamSpace, u32, DefId, ast::Name),
     DefUse(DefId),
     DefUpvar(ast::NodeId,  // id of closed over local
+             usize,        // index in the freevars list of the closure
              ast::NodeId), // expr node that creates the closure
 
     /// Note that if it's a tuple struct's definition, the node id of the DefId
@@ -129,7 +130,7 @@ impl Def {
                 id
             }
             DefLocal(id) |
-            DefUpvar(id, _) |
+            DefUpvar(id, _, _) |
             DefRegion(id) |
             DefLabel(id)  |
             DefSelfTy(_, Some((_, id))) => {

src/librustc/middle/expr_use_visitor.rs

@@ -246,7 +246,7 @@ impl OverloadedCallType {
 pub struct ExprUseVisitor<'d, 't, 'a: 't, 'tcx:'a+'d+'t> {
     typer: &'t infer::InferCtxt<'a, 'tcx>,
     mc: mc::MemCategorizationContext<'t, 'a, 'tcx>,
-    delegate: &'d mut (Delegate<'tcx>+'d),
+    delegate: &'d mut Delegate<'tcx>,
 }
 
 // If the TYPER results in an error, it's because the type check

src/librustc/middle/mem_categorization.rs

@@ -578,7 +578,7 @@ impl<'t, 'a,'tcx> MemCategorizationContext<'t, 'a, 'tcx> {
               }))
           }
 
-          def::DefUpvar(var_id, fn_node_id) => {
+          def::DefUpvar(var_id, _, fn_node_id) => {
               let ty = try!(self.node_ty(fn_node_id));
               match ty.sty {
                   ty::TyClosure(closure_id, _) => {

src/librustc/middle/region.rs

@@ -329,6 +329,9 @@ impl RegionMaps {
     pub fn item_extent(&self, n: ast::NodeId) -> CodeExtent {
         self.lookup_code_extent(CodeExtentData::DestructionScope(n))
     }
+    pub fn opt_destruction_extent(&self, n: ast::NodeId) -> Option<CodeExtent> {
+        self.code_extent_interner.borrow().get(&CodeExtentData::DestructionScope(n)).cloned()
+    }
     pub fn intern_code_extent(&self,
                               e: CodeExtentData,
                               parent: CodeExtent) -> CodeExtent {

src/librustc/middle/ty.rs

@@ -3475,6 +3475,13 @@ impl<'tcx, 'container> AdtDefData<'tcx, 'container> {
             .expect("variant_with_id: unknown variant")
     }
 
+    pub fn variant_index_with_id(&self, vid: DefId) -> usize {
+        self.variants
+            .iter()
+            .position(|v| v.did == vid)
+            .expect("variant_index_with_id: unknown variant")
+    }
+
     pub fn variant_of_def(&self, def: def::Def) -> &VariantDefData<'tcx, 'container> {
         match def {
             def::DefVariant(_, vid, _) => self.variant_with_id(vid),
@@ -5191,28 +5198,12 @@ impl<'tcx> TyS<'tcx> {
 
                         if !adjusted_ty.references_error() {
                             for i in 0..adj.autoderefs {
-                                let method_call = MethodCall::autoderef(expr_id, i as u32);
-                                match method_type(method_call) {
-                                    Some(method_ty) => {
-                                        // Overloaded deref operators have all late-bound
-                                        // regions fully instantiated and coverge.
-                                        let fn_ret =
-                                            cx.no_late_bound_regions(&method_ty.fn_ret()).unwrap();
-                                        adjusted_ty = fn_ret.unwrap();
-                                    }
-                                    None => {}
-                                }
-                                match adjusted_ty.builtin_deref(true, NoPreference) {
-                                    Some(mt) => { adjusted_ty = mt.ty; }
-                                    None => {
-                                        cx.sess.span_bug(
-                                            span,
-                                            &format!("the {}th autoderef failed: {}",
-                                                    i,
-                                                     adjusted_ty)
-                                            );
-                                    }
-                                }
+                                adjusted_ty =
+                                    adjusted_ty.adjust_for_autoderef(cx,
+                                                                     expr_id,
+                                                                     span,
+                                                                     i as u32,
+                                                                     &mut method_type);
                             }
                         }
 
@@ -5228,6 +5219,36 @@ impl<'tcx> TyS<'tcx> {
         };
     }
 
+    pub fn adjust_for_autoderef<F>(&'tcx self,
+                                   cx: &ctxt<'tcx>,
+                                   expr_id: ast::NodeId,
+                                   expr_span: Span,
+                                   autoderef: u32, // how many autoderefs so far?
+                                   mut method_type: F)
+                                   -> Ty<'tcx> where
+        F: FnMut(MethodCall) -> Option<Ty<'tcx>>,
+    {
+        let method_call = MethodCall::autoderef(expr_id, autoderef);
+        let mut adjusted_ty = self;
+        if let Some(method_ty) = method_type(method_call) {
+            // Method calls always have all late-bound regions
+            // fully instantiated.
+            let fn_ret = cx.no_late_bound_regions(&method_ty.fn_ret()).unwrap();
+            adjusted_ty = fn_ret.unwrap();
+        }
+        match adjusted_ty.builtin_deref(true, NoPreference) {
+            Some(mt) => mt.ty,
+            None => {
+                cx.sess.span_bug(
+                    expr_span,
+                    &format!("the {}th autoderef failed: {}",
+                             autoderef,
+                             adjusted_ty)
+                        );
+            }
+        }
+    }
+
     pub fn adjust_for_autoref(&'tcx self, cx: &ctxt<'tcx>,
                               autoref: Option<AutoRef<'tcx>>)
                               -> Ty<'tcx> {

src/librustc/session/config.rs

@@ -104,6 +104,7 @@ pub struct Options {
     pub parse_only: bool,
     pub no_trans: bool,
     pub treat_err_as_bug: bool,
+    pub always_build_mir: bool,
     pub no_analysis: bool,
     pub debugging_opts: DebuggingOptions,
     /// Whether to write dependency files. It's (enabled, optional filename).
@@ -216,6 +217,7 @@ pub fn basic_options() -> Options {
         parse_only: false,
         no_trans: false,
         treat_err_as_bug: false,
+        always_build_mir: false,
         no_analysis: false,
         debugging_opts: basic_debugging_options(),
         write_dependency_info: (false, None),
@@ -583,6 +585,8 @@ options! {DebuggingOptions, DebuggingSetter, basic_debugging_options,
           "Run all passes except translation; no output"),
     treat_err_as_bug: bool = (false, parse_bool,
           "Treat all errors that occur as bugs"),
+    always_build_mir: bool = (false, parse_bool,
+          "Always build MIR for all fns, even without a #[rustc_mir] annotation"),
     no_analysis: bool = (false, parse_bool,
           "Parse and expand the source, but run no analysis"),
     extra_plugins: Vec<String> = (Vec::new(), parse_list,
@@ -894,6 +898,7 @@ pub fn build_session_options(matches: &getopts::Matches) -> Options {
     let parse_only = debugging_opts.parse_only;
     let no_trans = debugging_opts.no_trans;
     let treat_err_as_bug = debugging_opts.treat_err_as_bug;
+    let always_build_mir = debugging_opts.always_build_mir;
     let no_analysis = debugging_opts.no_analysis;
 
     if debugging_opts.debug_llvm {
@@ -1049,6 +1054,7 @@ pub fn build_session_options(matches: &getopts::Matches) -> Options {
         parse_only: parse_only,
         no_trans: no_trans,
         treat_err_as_bug: treat_err_as_bug,
+        always_build_mir: always_build_mir,
         no_analysis: no_analysis,
         debugging_opts: debugging_opts,
         write_dependency_info: write_dependency_info,