Clearer separation of WASI POSIX support with (or without) wasi-libc

lib/std/Build/Step/Run.zig

@@ -1089,7 +1089,7 @@ fn runCommand(
                         // the `--` before the module name. This appears to work for both old and
                         // new Wasmtime versions.
                         try interp_argv.append(bin_name);
-                        try interp_argv.append("--dir=.");
+                        try interp_argv.append("--dir=."); // Preopen '.' at file descriptor 3 for cwd
                         try interp_argv.append("--");
                         try interp_argv.append(argv[0]);
                         try interp_argv.appendSlice(argv[1..]);

lib/std/c.zig

@@ -133,9 +133,10 @@ pub const dev_t = switch (native_os) {
 pub const mode_t = switch (native_os) {
     .linux => linux.mode_t,
     .emscripten => emscripten.mode_t,
-    .openbsd, .haiku, .netbsd, .solaris, .illumos, .wasi => u32,
+    .openbsd, .haiku, .netbsd, .solaris, .illumos => u32,
     .freebsd, .macos, .ios, .tvos, .watchos, .visionos => u16,
-    else => u0,
+    .wasi, .windows => void,
+    else => void,
 };
 
 pub const nlink_t = switch (native_os) {
@@ -671,6 +672,7 @@ pub const F = switch (native_os) {
     .linux => linux.F,
     .emscripten => emscripten.F,
     .wasi => struct {
+        // Match F_* constants from lib/libc/include/wasm-wasi-musl/__header_fcntl.h
         pub const GETFD = 1;
         pub const SETFD = 2;
         pub const GETFL = 3;
@@ -1702,17 +1704,49 @@ pub const S = switch (native_os) {
     .linux => linux.S,
     .emscripten => emscripten.S,
     .wasi => struct {
-        pub const IEXEC = @compileError("TODO audit this");
+        // Match S_* constants from lib/libc/include/wasm-wasi-musl/__mode_t.h
+        //
+        // Note, a bug in wasi-libc means both IFIFO and IFSOCK have the same value (0xc000).
+        // IFIFO should be 0x1000 (see https://github.com/WebAssembly/wasi-libc/pull/463), and
+        // 0x1000 is used by the wasi-libc bottom-half.  So we use 0x1000 here.  But the actual bit
+        // values we get back from a wasi-libc may get masked with the wrong values, or may get
+        // mistranslated.  So the FIFO and Socket file-type bits are not trustworthy.
         pub const IFBLK = 0x6000;
         pub const IFCHR = 0x2000;
         pub const IFDIR = 0x4000;
-        pub const IFIFO = 0xc000;
+        pub const IFIFO = 0x1000; // buggy
         pub const IFLNK = 0xa000;
-        pub const IFMT = IFBLK | IFCHR | IFDIR | IFIFO | IFLNK | IFREG | IFSOCK;
         pub const IFREG = 0x8000;
-        /// There's no concept of UNIX domain socket but we define this value here
-        /// in order to line with other OSes.
-        pub const IFSOCK = 0x1;
+        pub const IFSOCK = 0xc000; // buggy
+        pub const IFMT = IFBLK | IFCHR | IFDIR | IFIFO | IFLNK | IFREG | IFSOCK;
+
+        pub fn ISBLK(m: u32) bool {
+            return m & IFMT == IFBLK;
+        }
+
+        pub fn ISCHR(m: u32) bool {
+            return m & IFMT == IFCHR;
+        }
+
+        pub fn ISDIR(m: u32) bool {
+            return m & IFMT == IFDIR;
+        }
+
+        pub fn ISFIFO(m: u32) bool {
+            return m & IFMT == IFIFO;
+        }
+
+        pub fn ISLNK(m: u32) bool {
+            return m & IFMT == IFLNK;
+        }
+
+        pub fn ISREG(m: u32) bool {
+            return m & IFMT == IFREG;
+        }
+
+        pub fn ISSOCK(m: u32) bool {
+            return m & IFMT == IFSOCK;
+        }
     },
     .macos, .ios, .tvos, .watchos, .visionos => struct {
         pub const IFMT = 0o170000;
@@ -6486,10 +6520,11 @@ pub const Stat = switch (native_os) {
     },
     .emscripten => emscripten.Stat,
     .wasi => extern struct {
+        // Matches wasi-libc's struct stat in lib/libc/include/wasm-wasi-musl/__struct_stat.h
         dev: dev_t,
         ino: ino_t,
         nlink: nlink_t,
-        mode: mode_t,
+        mode: c_uint, // only the file-type bits (S.IFMT), no permission bits
         uid: uid_t,
         gid: gid_t,
         __pad0: c_uint = 0,
@@ -7185,17 +7220,13 @@ pub const AT = switch (native_os) {
         pub const RECURSIVE = 0x8000;
     },
     .wasi => struct {
-        pub const SYMLINK_NOFOLLOW = 0x100;
-        pub const SYMLINK_FOLLOW = 0x400;
-        pub const REMOVEDIR: u32 = 0x4;
-        /// When linking libc, we follow their convention and use -2 for current working directory.
-        /// However, without libc, Zig does a different convention: it assumes the
-        /// current working directory is the first preopen. This behavior can be
-        /// overridden with a public function called `wasi_cwd` in the root source
-        /// file.
-        pub const FDCWD: fd_t = if (builtin.link_libc) -2 else 3;
+        // Match AT_* constants in lib/libc/include/wasm-wasi-musl/__header_fcntl.h
+        pub const FDCWD = -2;
+        pub const EACCESS = 0x0;
+        pub const SYMLINK_NOFOLLOW = 0x1;
+        pub const SYMLINK_FOLLOW = 0x2;
+        pub const REMOVEDIR = 0x4;
     },
-
     else => void,
 };
 
@@ -7224,6 +7255,7 @@ pub const O = switch (native_os) {
         _: u9 = 0,
     },
     .wasi => packed struct(u32) {
+        // Match O_* bits from lib/libc/include/wasm-wasi-musl/__header_fcntl.h
         APPEND: bool = false,
         DSYNC: bool = false,
         NONBLOCK: bool = false,
@@ -7236,10 +7268,17 @@ pub const O = switch (native_os) {
         TRUNC: bool = false,
         _16: u8 = 0,
         NOFOLLOW: bool = false,
+
+        // Logical O_ACCMODE is the following 4 bits, note O_SEARCH is between RD and WR,
+        // so can't easily reuse std.posix.ACCMODE in this struct.
         EXEC: bool = false,
-        read: bool = false,
+        RDONLY: bool = true, // equivalent to "ACCMODE = .RDONLY" default
         SEARCH: bool = false,
-        write: bool = false,
+        WRONLY: bool = false,
+
+        // CLOEXEC, TTY_ININT, NOCTTY are mapped 0, so they're silently
+        // ignored in C code.
+
         _: u3 = 0,
     },
     .solaris, .illumos => packed struct(u32) {
@@ -9069,6 +9108,7 @@ pub const gettimeofday = switch (native_os) {
 
 pub const msync = switch (native_os) {
     .netbsd => private.__msync13,
+    .wasi => {},
     else => private.msync,
 };
 
@@ -9211,8 +9251,8 @@ pub const fork = switch (native_os) {
 pub extern "c" fn access(path: [*:0]const u8, mode: c_uint) c_int;
 pub extern "c" fn faccessat(dirfd: fd_t, path: [*:0]const u8, mode: c_uint, flags: c_uint) c_int;
 pub extern "c" fn pipe(fds: *[2]fd_t) c_int;
-pub extern "c" fn mkdir(path: [*:0]const u8, mode: c_uint) c_int;
-pub extern "c" fn mkdirat(dirfd: fd_t, path: [*:0]const u8, mode: u32) c_int;
+pub extern "c" fn mkdir(path: [*:0]const u8, mode: usize) c_int;
+pub extern "c" fn mkdirat(dirfd: fd_t, path: [*:0]const u8, mode: usize) c_int;
 pub extern "c" fn symlink(existing: [*:0]const u8, new: [*:0]const u8) c_int;
 pub extern "c" fn symlinkat(oldpath: [*:0]const u8, newdirfd: fd_t, newpath: [*:0]const u8) c_int;
 pub extern "c" fn rename(old: [*:0]const u8, new: [*:0]const u8) c_int;

lib/std/debug.zig

@@ -1017,12 +1017,12 @@ test printLineFromFileAnyOs {
 
     var test_dir = std.testing.tmpDir(.{});
     defer test_dir.cleanup();
-    // Relies on testing.tmpDir internals which is not ideal, but SourceLocation requires paths.
+    // Relies on testing.tmpDir internals which is not ideal, but SourceLocation requires paths relative to cwd.
     const test_dir_path = try join(allocator, &.{ ".zig-cache", "tmp", test_dir.sub_path[0..] });
     defer allocator.free(test_dir_path);
 
-    // Cases
     {
+        // no newlines in file
         const path = try join(allocator, &.{ test_dir_path, "one_line.zig" });
         defer allocator.free(path);
         try test_dir.dir.writeFile(.{ .sub_path = "one_line.zig", .data = "no new lines in this file, but one is printed anyway" });
@@ -1034,7 +1034,8 @@ test printLineFromFileAnyOs {
         output.clearRetainingCapacity();
     }
     {
-        const path = try fs.path.join(allocator, &.{ test_dir_path, "three_lines.zig" });
+        // print 1 & 3 of 3-line file
+        const path = try join(allocator, &.{ test_dir_path, "three_lines.zig" });
         defer allocator.free(path);
         try test_dir.dir.writeFile(.{
             .sub_path = "three_lines.zig",
@@ -1054,9 +1055,10 @@ test printLineFromFileAnyOs {
         output.clearRetainingCapacity();
     }
     {
+        // mem.page_size boundary crossing line
         const file = try test_dir.dir.createFile("line_overlaps_page_boundary.zig", .{});
         defer file.close();
-        const path = try fs.path.join(allocator, &.{ test_dir_path, "line_overlaps_page_boundary.zig" });
+        const path = try join(allocator, &.{ test_dir_path, "line_overlaps_page_boundary.zig" });
         defer allocator.free(path);
 
         const overlap = 10;
@@ -1070,9 +1072,10 @@ test printLineFromFileAnyOs {
         output.clearRetainingCapacity();
     }
     {
+        // ends on mem.page_size boundary
         const file = try test_dir.dir.createFile("file_ends_on_page_boundary.zig", .{});
         defer file.close();
-        const path = try fs.path.join(allocator, &.{ test_dir_path, "file_ends_on_page_boundary.zig" });
+        const path = try join(allocator, &.{ test_dir_path, "file_ends_on_page_boundary.zig" });
         defer allocator.free(path);
 
         var writer = file.writer();
@@ -1083,9 +1086,10 @@ test printLineFromFileAnyOs {
         output.clearRetainingCapacity();
     }
     {
+        // multi-mem.page_size "line"
         const file = try test_dir.dir.createFile("very_long_first_line_spanning_multiple_pages.zig", .{});
         defer file.close();
-        const path = try fs.path.join(allocator, &.{ test_dir_path, "very_long_first_line_spanning_multiple_pages.zig" });
+        const path = try join(allocator, &.{ test_dir_path, "very_long_first_line_spanning_multiple_pages.zig" });
         defer allocator.free(path);
 
         var writer = file.writer();
@@ -1108,9 +1112,10 @@ test printLineFromFileAnyOs {
         output.clearRetainingCapacity();
     }
     {
+        // mem.page_size pages of newlines
         const file = try test_dir.dir.createFile("file_of_newlines.zig", .{});
         defer file.close();
-        const path = try fs.path.join(allocator, &.{ test_dir_path, "file_of_newlines.zig" });
+        const path = try join(allocator, &.{ test_dir_path, "file_of_newlines.zig" });
         defer allocator.free(path);
 
         var writer = file.writer();