rustfmt libcoretest

CONTRIBUTING.md

@@ -122,6 +122,8 @@ To see a full list of options, run `./configure --help`.
 
 Some common make targets are:
 
+- `make tips` - show useful targets, variables and other tips for working with
+   the build system.
 - `make rustc-stage1` - build up to (and including) the first stage. For most
   cases we don't need to build the stage2 compiler, so we can save time by not
   building it. The stage1 compiler is a fully functioning compiler and

configure

@@ -599,6 +599,7 @@ opt debug-assertions 0 "build with debugging assertions"
 opt fast-make 0 "use .gitmodules as timestamp for submodule deps"
 opt ccache 0 "invoke gcc/clang via ccache to reuse object files between builds"
 opt local-rust 0 "use an installed rustc rather than downloading a snapshot"
+opt local-rebuild 0 "use an installed rustc matching the current version, for rebuilds"
 opt llvm-static-stdcpp 0 "statically link to libstdc++ for LLVM"
 opt rpath 1 "build rpaths into rustc itself"
 opt stage0-landing-pads 1 "enable landing pads during bootstrap with stage0"
@@ -847,6 +848,16 @@ then
     BIN_SUF=.exe
 fi
 
+# --enable-local-rebuild implies --enable-local-rust too
+if [ -n "$CFG_ENABLE_LOCAL_REBUILD" ]
+then
+    if [ -z "$CFG_ENABLE_LOCAL_RUST" ]
+    then
+        CFG_ENABLE_LOCAL_RUST=1
+        putvar CFG_ENABLE_LOCAL_RUST
+    fi
+fi
+
 if [ -n "$CFG_ENABLE_LOCAL_RUST" ]
 then
     system_rustc=$(which rustc)

mk/main.mk

@@ -13,7 +13,7 @@
 ######################################################################
 
 # The version number
-CFG_RELEASE_NUM=1.10.0
+CFG_RELEASE_NUM=1.11.0
 
 # An optional number to put after the label, e.g. '.2' -> '-beta.2'
 # NB Make sure it starts with a dot to conform to semver pre-release
@@ -34,7 +34,14 @@ CFG_FILENAME_EXTRA=$(shell printf '%s' $(CFG_RELEASE)$(CFG_EXTRA_FILENAME) | $(C
 # intentionally not "secure" by any definition, this is largely just a deterrent
 # from users enabling unstable features on the stable compiler.
 CFG_BOOTSTRAP_KEY=$(CFG_FILENAME_EXTRA)
+
+# The stage0 compiler needs to use the previous key recorded in src/stage0.txt,
+# except for local-rebuild when it just uses the same current key.
+ifdef CFG_ENABLE_LOCAL_REBUILD
+CFG_BOOTSTRAP_KEY_STAGE0=$(CFG_BOOTSTRAP_KEY)
+else
 CFG_BOOTSTRAP_KEY_STAGE0=$(shell grep 'rustc_key' $(S)src/stage0.txt | sed 's/rustc_key: '//)
+endif
 
 ifeq ($(CFG_RELEASE_CHANNEL),stable)
 # This is the normal semver version string, e.g. "0.12.0", "0.12.0-nightly"
@@ -526,6 +533,11 @@ ifneq ($(strip $(CFG_BUILD)),$(strip $(3)))
 CFGFLAG$(1)_T_$(2)_H_$(3) = stage1
 
 RPATH_VAR$(1)_T_$(2)_H_$(3) := $$(TARGET_RPATH_VAR1_T_$(2)_H_$$(CFG_BUILD))
+else
+ifdef CFG_ENABLE_LOCAL_REBUILD
+# Assume the local-rebuild rustc already has stage1 features too.
+CFGFLAG$(1)_T_$(2)_H_$(3) = stage1
+endif
 endif
 endif
 

src/bootstrap/build/config.rs

@@ -65,8 +65,9 @@ pub struct Config {
     pub build: String,
     pub host: Vec<String>,
     pub target: Vec<String>,
-    pub rustc: Option<String>,
-    pub cargo: Option<String>,
+    pub rustc: Option<PathBuf>,
+    pub cargo: Option<PathBuf>,
+    pub local_rebuild: bool,
 
     // libstd features
     pub debug_jemalloc: bool,
@@ -207,8 +208,8 @@ impl Config {
                 config.target.push(target.clone());
             }
         }
-        config.rustc = build.rustc;
-        config.cargo = build.cargo;
+        config.rustc = build.rustc.map(PathBuf::from);
+        config.cargo = build.cargo.map(PathBuf::from);
         set(&mut config.compiler_docs, build.compiler_docs);
         set(&mut config.docs, build.docs);
 
@@ -315,6 +316,7 @@ impl Config {
                 ("RPATH", self.rust_rpath),
                 ("OPTIMIZE_TESTS", self.rust_optimize_tests),
                 ("DEBUGINFO_TESTS", self.rust_debuginfo_tests),
+                ("LOCAL_REBUILD", self.local_rebuild),
             }
 
             match key {
@@ -377,6 +379,10 @@ impl Config {
                                      .or_insert(Target::default());
                     target.ndk = Some(PathBuf::from(value));
                 }
+                "CFG_LOCAL_RUST_ROOT" if value.len() > 0 => {
+                    self.rustc = Some(PathBuf::from(value).join("bin/rustc"));
+                    self.cargo = Some(PathBuf::from(value).join("bin/cargo"));
+                }
                 _ => {}
             }
         }

src/bootstrap/build/mod.rs

@@ -510,6 +510,14 @@ impl Build {
              .arg("-j").arg(self.jobs().to_string())
              .arg("--target").arg(target);
 
+        let stage;
+        if compiler.stage == 0 && self.config.local_rebuild {
+            // Assume the local-rebuild rustc already has stage1 features.
+            stage = 1;
+        } else {
+            stage = compiler.stage;
+        }
+
         // Customize the compiler we're running. Specify the compiler to cargo
         // as our shim and then pass it some various options used to configure
         // how the actual compiler itself is called.
@@ -518,7 +526,7 @@ impl Build {
         // src/bootstrap/{rustc,rustdoc.rs}
         cargo.env("RUSTC", self.out.join("bootstrap/debug/rustc"))
              .env("RUSTC_REAL", self.compiler_path(compiler))
-             .env("RUSTC_STAGE", compiler.stage.to_string())
+             .env("RUSTC_STAGE", stage.to_string())
              .env("RUSTC_DEBUGINFO", self.config.rust_debuginfo.to_string())
              .env("RUSTC_CODEGEN_UNITS",
                   self.config.rust_codegen_units.to_string())
@@ -744,7 +752,7 @@ impl Build {
         // In stage0 we're using a previously released stable compiler, so we
         // use the stage0 bootstrap key. Otherwise we use our own build's
         // bootstrap key.
-        let bootstrap_key = if compiler.is_snapshot(self) {
+        let bootstrap_key = if compiler.is_snapshot(self) && !self.config.local_rebuild {
             &self.bootstrap_key_stage0
         } else {
             &self.bootstrap_key

src/doc/book/no-stdlib.md

@@ -84,4 +84,4 @@ which do not trigger a panic can be assured that this function is never
 called. The second function, `panic_fmt`, is also used by the failure
 mechanisms of the compiler.
 
-[unwind]: https://github.com/rust-lang/rust/blob/master/src/libstd/sys/common/unwind/gcc.rs
+[unwind]: https://github.com/rust-lang/rust/blob/master/src/libpanic_unwind/gcc.rs

src/doc/book/primitive-types.md

@@ -175,8 +175,6 @@ You can use a combo of `&` and `[]` to create a slice from various things. The
 detail later in this section. The `[]`s, with a range, let you define the
 length of the slice:
 
-[references]: references-and-borrowing.html
-
 ```rust
 let a = [0, 1, 2, 3, 4];
 let complete = &a[..]; // A slice containing all of the elements in a

src/doc/reference.md

@@ -2433,6 +2433,8 @@ The currently implemented features of the reference compiler are:
 * - `abi_vectorcall` - Allows the usage of the vectorcall calling convention
                              (e.g. `extern "vectorcall" func fn_();`)
 
+* - `dotdot_in_tuple_patterns` - Allows `..` in tuple (struct) patterns.
+
 If a feature is promoted to a language feature, then all existing programs will
 start to receive compilation warnings about `#![feature]` directives which enabled
 the new feature (because the directive is no longer necessary). However, if a

src/etc/unicode.py

@@ -25,6 +25,9 @@
 
 import fileinput, re, os, sys, operator
 
+bytes_old = 0
+bytes_new = 0
+
 preamble = '''// 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.
@@ -307,12 +310,137 @@ def emit_table(f, name, t_data, t_type = "&'static [(char, char)]", is_pub=True,
     format_table_content(f, data, 8)
     f.write("\n    ];\n\n")
 
+def emit_trie_lookup_range_table(f):
+    f.write("""
+
+// BoolTrie is a trie for representing a set of Unicode codepoints. It is
+// implemented with postfix compression (sharing of identical child nodes),
+// which gives both compact size and fast lookup.
+//
+// The space of Unicode codepoints is divided into 3 subareas, each
+// represented by a trie with different depth. In the first (0..0x800), there
+// is no trie structure at all; each u64 entry corresponds to a bitvector
+// effectively holding 64 bool values.
+//
+// In the second (0x800..0x10000), each child of the root node represents a
+// 64-wide subrange, but instead of storing the full 64-bit value of the leaf,
+// the trie stores an 8-bit index into a shared table of leaf values. This
+// exploits the fact that in reasonable sets, many such leaves can be shared.
+//
+// In the third (0x10000..0x110000), each child of the root node represents a
+// 4096-wide subrange, and the trie stores an 8-bit index into a 64-byte slice
+// of a child tree. Each of these 64 bytes represents an index into the table
+// of shared 64-bit leaf values. This exploits the sparse structure in the
+// non-BMP range of most Unicode sets.
+pub struct BoolTrie {
+    // 0..0x800 (corresponding to 1 and 2 byte utf-8 sequences)
+    r1: [u64; 32],   // leaves
+
+    // 0x800..0x10000 (corresponding to 3 byte utf-8 sequences)
+    r2: [u8; 992],      // first level
+    r3: &'static [u64],  // leaves
+
+    // 0x10000..0x110000 (corresponding to 4 byte utf-8 sequences)
+    r4: [u8; 256],       // first level
+    r5: &'static [u8],   // second level
+    r6: &'static [u64],  // leaves
+}
+
+fn trie_range_leaf(c: usize, bitmap_chunk: u64) -> bool {
+    ((bitmap_chunk >> (c & 63)) & 1) != 0
+}
+
+fn trie_lookup_range_table(c: char, r: &'static BoolTrie) -> bool {
+    let c = c as usize;
+    if c < 0x800 {
+        trie_range_leaf(c, r.r1[c >> 6])
+    } else if c < 0x10000 {
+        let child = r.r2[(c >> 6) - 0x20];
+        trie_range_leaf(c, r.r3[child as usize])
+    } else {
+        let child = r.r4[(c >> 12) - 0x10];
+        let leaf = r.r5[((child as usize) << 6) + ((c >> 6) & 0x3f)];
+        trie_range_leaf(c, r.r6[leaf as usize])
+    }
+}\n
+""")
+
+def compute_trie(rawdata, chunksize):
+    root = []
+    childmap = {}
+    child_data = []
+    for i in range(len(rawdata) / chunksize):
+        data = rawdata[i * chunksize: (i + 1) * chunksize]
+        child = '|'.join(map(str, data))
+        if child not in childmap:
+            childmap[child] = len(childmap)
+            child_data.extend(data)
+        root.append(childmap[child])
+    return (root, child_data)
+
+def emit_bool_trie(f, name, t_data, is_pub=True):
+    global bytes_old, bytes_new
+    bytes_old += 8 * len(t_data)
+    CHUNK = 64
+    rawdata = [False] * 0x110000;
+    for (lo, hi) in t_data:
+        for cp in range(lo, hi + 1):
+            rawdata[cp] = True
+
+    # convert to bitmap chunks of 64 bits each
+    chunks = []
+    for i in range(0x110000 / CHUNK):
+        chunk = 0
+        for j in range(64):
+            if rawdata[i * 64 + j]:
+                chunk |= 1 << j
+        chunks.append(chunk)
+
+    pub_string = ""
+    if is_pub:
+        pub_string = "pub "
+    f.write("    %sconst %s: &'static super::BoolTrie = &super::BoolTrie {\n" % (pub_string, name))
+    f.write("        r1: [\n")
+    data = ','.join('0x%016x' % chunk for chunk in chunks[0:0x800 / CHUNK])
+    format_table_content(f, data, 12)
+    f.write("\n        ],\n")
+
+    # 0x800..0x10000 trie
+    (r2, r3) = compute_trie(chunks[0x800 / CHUNK : 0x10000 / CHUNK], 64 / CHUNK)
+    f.write("        r2: [\n")
+    data = ','.join(str(node) for node in r2)
+    format_table_content(f, data, 12)
+    f.write("\n        ],\n")
+    f.write("        r3: &[\n")
+    data = ','.join('0x%016x' % chunk for chunk in r3)
+    format_table_content(f, data, 12)
+    f.write("\n        ],\n")
+
+    # 0x10000..0x110000 trie
+    (mid, r6) = compute_trie(chunks[0x10000 / CHUNK : 0x110000 / CHUNK], 64 / CHUNK)
+    (r4, r5) = compute_trie(mid, 64)
+    f.write("        r4: [\n")
+    data = ','.join(str(node) for node in r4)
+    format_table_content(f, data, 12)
+    f.write("\n        ],\n")
+    f.write("        r5: &[\n")
+    data = ','.join(str(node) for node in r5)
+    format_table_content(f, data, 12)
+    f.write("\n        ],\n")
+    f.write("        r6: &[\n")
+    data = ','.join('0x%016x' % chunk for chunk in r6)
+    format_table_content(f, data, 12)
+    f.write("\n        ],\n")
+
+    f.write("    };\n\n")
+    bytes_new += 256 + 992 + 256 + 8 * len(r3) + len(r5) + 8 * len(r6)
+
 def emit_property_module(f, mod, tbl, emit):
     f.write("pub mod %s {\n" % mod)
     for cat in sorted(emit):
-        emit_table(f, "%s_table" % cat, tbl[cat])
+        emit_bool_trie(f, "%s_table" % cat, tbl[cat])
         f.write("    pub fn %s(c: char) -> bool {\n" % cat)
-        f.write("        super::bsearch_range_table(c, %s_table)\n" % cat)
+        f.write("        super::trie_lookup_range_table(c, %s_table)\n" % cat)
         f.write("    }\n\n")
     f.write("}\n\n")
 
@@ -402,8 +530,9 @@ def emit_norm_module(f, canon, compat, combine, norm_props):
         norm_props = load_properties("DerivedNormalizationProps.txt",
                      ["Full_Composition_Exclusion"])
 
-        # bsearch_range_table is used in all the property modules below
-        emit_bsearch_range_table(rf)
+        # trie_lookup_table is used in all the property modules below
+        emit_trie_lookup_range_table(rf)
+        # emit_bsearch_range_table(rf)
 
         # category tables
         for (name, cat, pfuns) in ("general_category", gencats, ["N", "Cc"]), \
@@ -414,3 +543,4 @@ def emit_norm_module(f, canon, compat, combine, norm_props):
         # normalizations and conversions module
         emit_norm_module(rf, canon_decomp, compat_decomp, combines, norm_props)
         emit_conversions_module(rf, to_upper, to_lower, to_title)
+    #print 'bytes before = %d, bytes after = %d' % (bytes_old, bytes_new)