[libc] Add swab implementation

[Caslyn]

Summary

Swab implementation is added to libc/src/unistd.

[Caslyn]

Fix typo

[michaelrj]

libc/spec/posix.td

550

this should be SSizeTType

libc/src/unistd/linux/swab.cpp

18

this should be ssize_t

[Caslyn]

Fix: size_t -> ssize_t

[Caslyn]

libc/src/unistd/linux/swab.cpp

18

Thanks for catching - included size check as well.

[mcgrathr]

libc/test/src/unistd/swab_test.cpp

15

Since the argument type is ssize_t there should be a test that a negative size is also a no-op.

20

This only tests the compiler, not the function, since it won't do anything to change what pointer to is.
A more meaningful test might be to have nonempty buffers with known contents and verify that the contents weren't changed (and that the from and to bytes don't match after swapping so you could tell).
It's separately worth testing that zero or negative size works with nullptr buffers since otherwise that would crash when it shouldn't, which might not be tested by the case of real buffers that weren't touched.

27

Same logic applies here. This isn't really testing anything except that it didn't crash.

33

Just = {}; is equivalent here and seems more clear that it's simply zero-initializing the whole array.

[Caslyn]

Improve tests per review feedback

[mcgrathr]

lgtm!

libc/src/unistd/linux/swab.cpp

18

I don't see any reason for this short-circuit. The loop will naturally have zero iterations for these cases.

[Caslyn]

Remove redundant check

[michaelrj]

LGTM

[mcgrathr]

Sorry I failed to notice the placement and build plumbing are not right here. It should not be in the linux/ subdir, it has nothing to do with linux or any other OS.
Follow the model of e.g. getopt that is in unistd but is not OS-specific.

libc/config/linux/aarch64/entrypoints.txt

set(TARGET_LIBC_ENTRYPOINTS
     libc.src.unistd.linkat
     libc.src.unistd.lseek
     libc.src.unistd.pread
     libc.src.unistd.pwrite
     libc.src.unistd.read
     libc.src.unistd.readlink
     libc.src.unistd.readlinkat
     libc.src.unistd.rmdir
+    libc.src.unistd.swab
     libc.src.unistd.symlink
     libc.src.unistd.symlinkat
     libc.src.unistd.sysconf
     libc.src.unistd.truncate
     libc.src.unistd.unlink
     libc.src.unistd.unlinkat
     libc.src.unistd.write
 )

libc/config/linux/riscv64/entrypoints.txt

set(TARGET_LIBC_ENTRYPOINTS
     libc.src.unistd.linkat
     libc.src.unistd.lseek
     libc.src.unistd.pread
     libc.src.unistd.pwrite
     libc.src.unistd.read
     libc.src.unistd.readlink
     libc.src.unistd.readlinkat
     libc.src.unistd.rmdir
+    libc.src.unistd.swab
     libc.src.unistd.symlink
     libc.src.unistd.symlinkat
     libc.src.unistd.sysconf
     libc.src.unistd.truncate
     libc.src.unistd.unlink
     libc.src.unistd.unlinkat
     libc.src.unistd.write
 )

libc/config/linux/x86_64/entrypoints.txt

set(TARGET_LIBC_ENTRYPOINTS
     libc.src.unistd.linkat
     libc.src.unistd.lseek
     libc.src.unistd.pread
     libc.src.unistd.pwrite
     libc.src.unistd.read
     libc.src.unistd.readlink
     libc.src.unistd.readlinkat
     libc.src.unistd.rmdir
+    libc.src.unistd.swab
     libc.src.unistd.symlink
     libc.src.unistd.symlinkat
     libc.src.unistd.sysconf
     libc.src.unistd.truncate
     libc.src.unistd.unlink
     libc.src.unistd.unlinkat
     libc.src.unistd.write
 

libc/spec/posix.td

[
           [ArgSpec<ConstCharRestrictedPtr>, ArgSpec<CharRestrictedPtr>, ArgSpec<SizeTType>]
         >,
         FunctionSpec<
           "rmdir",
           RetValSpec<IntType>,
           [ArgSpec<ConstCharPtr>]
         >,
         FunctionSpec<
+          "swab",
+          RetValSpec<VoidType>,
+          [ArgSpec<ConstVoidRestrictedPtr>, ArgSpec<VoidPtr>, ArgSpec<SSizeTType>]
+        >,
+        FunctionSpec<
           "symlink",
           RetValSpec<IntType>,
           [ArgSpec<ConstCharPtr>, ArgSpec<ConstCharPtr>]
         >,
         FunctionSpec<
           "symlinkat",
           RetValSpec<IntType>,
           [ArgSpec<IntType>, ArgSpec<ConstCharPtr>, ArgSpec<IntType>, ArgSpec<ConstCharPtr>]

libc/src/unistd/CMakeLists.txt

 add_entrypoint_object(
   rmdir
   ALIAS
   DEPENDS
     .${LIBC_TARGET_OS}.rmdir
 )
 
 add_entrypoint_object(
+  swab
+  ALIAS
+  DEPENDS
+    .${LIBC_TARGET_OS}.swab
+)
+
+add_entrypoint_object(
   symlink
   ALIAS
   DEPENDS
     .${LIBC_TARGET_OS}.symlink
 )
 
 add_entrypoint_object(
   symlinkat

libc/src/unistd/linux/CMakeLists.txt

DEPENDS
     libc.include.fcntl
     libc.include.unistd
     libc.include.sys_syscall
     libc.src.__support.OSUtil.osutil
     libc.src.errno.errno
 )
 
 add_entrypoint_object(
+  swab
+  SRCS
+    swab.cpp
+  HDRS
+    ../swab.h
+)
+
+add_entrypoint_object(
   symlink
   SRCS
     symlink.cpp
   HDRS
     ../symlink.h
   DEPENDS
     libc.include.fcntl
     libc.include.unistd

libc/src/unistd/linux/swab.cpp

+//===-- Implementation of swab --------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/unistd/swab.h"
+
+#include "src/__support/common.h"
+
+namespace __llvm_libc {
+
+LLVM_LIBC_FUNCTION(void, swab,
+                   (const void *__restrict from, void *__restrict to,
+                    ssize_t n)) {
+  if (n < 2)
+    return;
+
+  const unsigned char *f = static_cast<const unsigned char *>(from);
+  unsigned char *t = static_cast<unsigned char *>(to);
+  for (ssize_t i = 1; i < n; i += 2) {
+    t[i - 1] = f[i];
+    t[i] = f[i - 1];
+  }
+}
+
+} // namespace __llvm_libc

libc/src/unistd/swab.h

+//===-- Implementation header for swab --------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_UNISTD_SWAB_H
+#define LLVM_LIBC_SRC_UNISTD_SWAB_H
+
+#include <unistd.h> // For ssize_t
+
+namespace __llvm_libc {
+
+void swab(const void *__restrict from, void *__restrict to, ssize_t n);
+
+} // namespace __llvm_libc
+
+#endif // LLVM_LIBC_SRC_UNISTD_SWAB_H

libc/test/src/unistd/CMakeLists.txt

DEPENDS
     libc.include.errno
     libc.include.fcntl
     libc.src.errno.errno
     libc.src.sys.stat.mkdir
     libc.src.unistd.rmdir
 )
 
 add_libc_unittest(
+  swab_test
+  SUITE
+    libc_unistd_unittests
+  SRCS
+    swab_test.cpp
+  DEPENDS
+    libc.src.unistd.swab
+)
+
+add_libc_unittest(
   readlink_test
   SUITE
     libc_unistd_unittests
   SRCS
     readlink_test.cpp
   DEPENDS
     libc.include.errno
     libc.include.unistd

libc/test/src/unistd/swab_test.cpp

+
+//===-- Unittests for swab ------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/unistd/swab.h"
+
+#include "src/string/string_utils.h"
+#include "test/UnitTest/Test.h"
+
+TEST(LlvmLibcSwabTest, NegativeSizeIsNoOp) {
+  const char *from = "abc";
+  char to[4] = {'x', 'y', 'z', '\0'};
+  __llvm_libc::swab(from, to, -1);
+  ASSERT_STREQ(to, "xyz");
+}
+
+TEST(LlvmLibcSwabTest, ZeroSizeIsNoOp) {
+  const char *from = "abc";
+  char to[4] = {'x', 'y', 'z', '\0'};
+  __llvm_libc::swab(from, to, 0);
+  ASSERT_STREQ(to, "xyz");
+}
+
+TEST(LlvmLibcSwabTest, SingleByteIsNoOp) {
+  char from[] = {'a'};
+  char to[4] = {'x', 'y', 'z', '\0'};
+  __llvm_libc::swab(from, to, sizeof(from));
+  ASSERT_STREQ(to, "xyz");
+}
+
+TEST(LlvmLibcSwabTest, NullPtrsAreNotDeRefedIfNIsLessThanTwo) {
+  // This test passes if a crash does not happen
+  __llvm_libc::swab(nullptr, nullptr, -1);
+  __llvm_libc::swab(nullptr, nullptr, 0);
+  __llvm_libc::swab(nullptr, nullptr, 1);
+}
+
+TEST(LlvmLibcSwabTest, BytesAreSwappedWithEvenN) {
+  {
+    const char *from = "ab";
+    char to[3] = {};
+    __llvm_libc::swab(from, to, __llvm_libc::internal::string_length(from));
+    ASSERT_STREQ(to, "ba");
+  }
+  {
+    const char *from = "abcd";
+    char to[5] = {};
+    __llvm_libc::swab(from, to, __llvm_libc::internal::string_length(from));
+    ASSERT_STREQ(to, "badc");
+  }
+  {
+    const char *from = "aAaAaA";
+    char to[7] = {};
+    __llvm_libc::swab(from, to, __llvm_libc::internal::string_length(from));
+    ASSERT_STREQ(to, "AaAaAa");
+  }
+}
+
+TEST(LlvmLibcSwabTest, LastByteIgnoredWithOddN) {
+  {
+    const char *from = "aba";
+    char to[3] = {};
+    __llvm_libc::swab(from, to, __llvm_libc::internal::string_length(from));
+    ASSERT_STREQ(to, "ba");
+  }
+  {
+    const char *from = "abcde";
+    char to[5] = {};
+    __llvm_libc::swab(from, to, __llvm_libc::internal::string_length(from));
+    ASSERT_STREQ(to, "badc");
+  }
+  {
+    const char *from = "aAaAaAx";
+    char to[7] = {};
+    __llvm_libc::swab(from, to, __llvm_libc::internal::string_length(from));
+    ASSERT_STREQ(to, "AaAaAa");
+  }
+}