[Support/Path] Make handling of paths like "///" consistent

[labath]

Summary

Various path functions were not treating paths consisting of slashes
alone consistently. For example, the iterator-based accessors decomposed the
path "///" into two elements: "/" and ".". This is not too bad, but it
is different from the behavior specified by posix:

A pathname that contains ***at least one non-slash character*** and that
ends with one or more trailing slashes shall be resolved as if a single
dot character ( '.' ) were appended to the pathname.

More importantly, this was different from how we treated the same path
in the filename+parent_path functions, which decomposed this path into
"." and "". This was completely wrong as it lost the information that
this was an absolute path which referred to the root directory.

This patch fixes this behavior by making sure all functions treat paths
consisting of (back)slashes alone the same way as "/". I.e., the
iterator-based functions will just report one component ("/"), and the
filename+parent_path will decompose them into "/" and "".

A slightly controversial topic here may be the treatment of "". Posix
says that paths beginning with "" may have special meaning and indeed
we have code which parses paths like "net/foo/bar" specially. However,
as we were already not being consistent in parsing the "" string
alone, and any special parsing for it would complicate the code further,
I chose to treat it the same way as longer sequences of slashes (which
are guaranteed to be the same as "/").

Another slight change of behavior is in the parsing of paths like
"net". Previously the last component of this path was ".". However,
as in our parsing the "net" part in this path was the same as the
"drive" part in "c:\" and the next slash was the "root directory", it
made sense to treat "net" the same way as "net/" (i.e., not to add
the extra "." component at the end).

[zturner]

unittests/Support/Path.cpp

186–190

A comment here maybe in order. Should this behavior be true for both path styles?

[labath]

unittests/Support/Path.cpp

186–190

I am not sure which aspect of this do you want to me comment on, as there several potentially controversial things going on here:

• decomposition of //: posix states that paths beginning with // are magic (I'm not sure if windows has a spec, but they are magic there as well), so treating it the same way as "/" is not entirely correct. It might be better to decompose this to ""+"//", but that is not completely consistent either, as elsewhere we treat "//net" as a single component, not as // followed by "net". (this holds for both path styles)
• treatment of \\\ on windows. It seems windows just refuses to resolve paths like this (though it is fine with multiple backslashes elsewhere in the path), so treating this the same way as \ is not entirely correct. However, I think it is somewhat better than treating it as ""+".", which we were doing before. (this is a windows speciality as posix will happily accept paths like this).

In general, I'm not trying to be overly-prescriptive about the *right* behavior here. I am just trying to make sure we do something vaguely reasonable, and I am totally open to changing there is a need for it, now or in the future. (Maybe I should put that in the comment?) It's just that right now this was the easiest to implement.

[labath]

Added an explanatory comment to the tests for "//", etc.

[labath]

Thank you for the review.

lib/Support/Path.cpp

StringRef find_first_component(StringRef path, Style style) {
     if (is_separator(path[0], style))
       return path.substr(0, 1);
 
     // * {file,directory}name
     size_t end = path.find_first_of(separators(style));
     return path.substr(0, end);
   }
 
+  // Returns the first character of the filename in str. For paths ending in
+  // '/', it returns the position of the '/'.
   size_t filename_pos(StringRef str, Style style) {
-    if (str.size() == 2 && is_separator(str[0], style) && str[0] == str[1])
-      return 0;
-
     if (str.size() > 0 && is_separator(str[str.size() - 1], style))
       return str.size() - 1;
 
     size_t pos = str.find_last_of(separators(style), str.size() - 1);
 
     if (real_style(style) == Style::windows) {
       if (pos == StringRef::npos)
         pos = str.find_last_of(':', str.size() - 2);
     }
 
     if (pos == StringRef::npos || (pos == 1 && is_separator(str[0], style)))
       return 0;
 
     return pos + 1;
   }
 
+  // Returns the position of the root directory in str. If there is no root
+  // directory in str, it returns StringRef::npos.
   size_t root_dir_start(StringRef str, Style style) {
     // case "c:/"
     if (real_style(style) == Style::windows) {
       if (str.size() > 2 && str[1] == ':' && is_separator(str[2], style))
         return 2;
     }
 
-    // case "//"
-    if (str.size() == 2 && is_separator(str[0], style) && str[0] == str[1])
-      return StringRef::npos;
-
     // case "//net"
     if (str.size() > 3 && is_separator(str[0], style) && str[0] == str[1] &&
         !is_separator(str[2], style)) {
       return str.find_first_of(separators(style), 2);
     }
 
     // case "/"
     if (str.size() > 0 && is_separator(str[0], style))
       return 0;
 
     return StringRef::npos;
   }
 
+  // Returns the position past the end of the "parent path" of path. The parent
+  // path will not end in '/', unless the parent is the root directory. If the
+  // path has no parent, 0 is returned.
   size_t parent_path_end(StringRef path, Style style) {
     size_t end_pos = filename_pos(path, style);
 
     bool filename_was_sep =
         path.size() > 0 && is_separator(path[end_pos], style);
 
-    // Skip separators except for root dir.
-    size_t root_dir_pos = root_dir_start(path.substr(0, end_pos), style);
-
-    while (end_pos > 0 && (end_pos - 1) != root_dir_pos &&
+    // Skip separators until we reach root dir (or the start of the string).
+    size_t root_dir_pos = root_dir_start(path, style);
+    while (end_pos > 0 &&
+           (root_dir_pos == StringRef::npos || end_pos > root_dir_pos) &&
            is_separator(path[end_pos - 1], style))
       --end_pos;
 
-    if (end_pos == 1 && root_dir_pos == 0 && filename_was_sep)
-      return StringRef::npos;
+    if (end_pos == root_dir_pos && !filename_was_sep) {
+      // We've reached the root dir and the input path was *not* ending in a
+      // sequence of slashes. Include the root dir in the parent path.
+      return root_dir_pos + 1;
+    }
 
+    // Otherwise, just include before the last slash.
     return end_pos;
   }
 } // end unnamed namespace
 
 enum FSEntity {
   FS_Dir,
   FS_File,
   FS_Name
if (was_net ||
       return *this;
     }
 
     // Skip extra separators.
     while (Position != Path.size() && is_separator(Path[Position], S)) {
       ++Position;
     }
 
-    // Treat trailing '/' as a '.'.
-    if (Position == Path.size()) {
+    // Treat trailing '/' as a '.', unless it is the root dir.
+    if (Position == Path.size() && Component != "/") {
       --Position;
       Component = ".";
       return *this;
     }
   }
 
   // Find next component.
   size_t end_pos = Path.find_first_of(separators(S), Position);
reverse_iterator rend(StringRef Path) {
   reverse_iterator I;
   I.Path = Path;
   I.Component = Path.substr(0, 0);
   I.Position = 0;
   return I;
 }
 
 reverse_iterator &reverse_iterator::operator++() {
-  // If we're at the end and the previous char was a '/', return '.' unless
-  // we are the root path.
   size_t root_dir_pos = root_dir_start(Path, S);
-  if (Position == Path.size() && Path.size() > root_dir_pos + 1 &&
-      is_separator(Path[Position - 1], S)) {
-    --Position;
-    Component = ".";
-    return *this;
-  }
 
   // Skip separators unless it's the root directory.
   size_t end_pos = Position;
-
   while (end_pos > 0 && (end_pos - 1) != root_dir_pos &&
          is_separator(Path[end_pos - 1], S))
     --end_pos;
 
+  // Treat trailing '/' as a '.', unless it is the root dir.
+  if (Position == Path.size() && !Path.empty() &&
+      is_separator(Path.back(), S) &&
+      (root_dir_pos == StringRef::npos || end_pos - 1 > root_dir_pos)) {
+    --Position;
+    Component = ".";
+    return *this;
+  }
+
   // Find next separator.
   size_t start_pos = filename_pos(Path.substr(0, end_pos), S);
   Component = Path.slice(start_pos, end_pos);
   Position = start_pos;
   return *this;
 }
 
 bool reverse_iterator::operator==(const reverse_iterator &RHS) const {

unittests/Support/Path.cpp

 
 TEST(Support, FilenameParent) {
   EXPECT_EQ("/", path::filename("/"));
   EXPECT_EQ("", path::parent_path("/"));
 
   EXPECT_EQ("\\", path::filename("c:\\", path::Style::windows));
   EXPECT_EQ("c:", path::parent_path("c:\\", path::Style::windows));
 
+  EXPECT_EQ("/", path::filename("//"));
+  EXPECT_EQ("", path::parent_path("//"));
+
+  EXPECT_EQ("/", path::filename("///"));
+  EXPECT_EQ("", path::parent_path("///"));
+
+  EXPECT_EQ("\\", path::filename("c:\\\\", path::Style::windows));
+  EXPECT_EQ("c:", path::parent_path("c:\\\\", path::Style::windows));
+
   EXPECT_EQ("bar", path::filename("/foo/bar"));
   EXPECT_EQ("/foo", path::parent_path("/foo/bar"));
 
   EXPECT_EQ("foo", path::filename("/foo"));
   EXPECT_EQ("/", path::parent_path("/foo"));
 
   EXPECT_EQ("foo", path::filename("foo"));
   EXPECT_EQ("", path::parent_path("foo"));

 static std::vector<StringRef>
 GetComponents(StringRef Path, path::Style S = path::Style::native) {
   return {path::begin(Path, S), path::end(Path)};
 }
 
 TEST(Support, PathIterator) {
   EXPECT_THAT(GetComponents("/foo"), testing::ElementsAre("/", "foo"));
   EXPECT_THAT(GetComponents("/"), testing::ElementsAre("/"));
+  EXPECT_THAT(GetComponents("//"), testing::ElementsAre("/"));
+  EXPECT_THAT(GetComponents("///"), testing::ElementsAre("/"));
   EXPECT_THAT(GetComponents("c/d/e/foo.txt"),
               testing::ElementsAre("c", "d", "e", "foo.txt"));
   EXPECT_THAT(GetComponents(".c/.d/../."),
               testing::ElementsAre(".c", ".d", "..", "."));
   EXPECT_THAT(GetComponents("/c/d/e/foo.txt"),
               testing::ElementsAre("/", "c", "d", "e", "foo.txt"));
   EXPECT_THAT(GetComponents("/.c/.d/../."),
               testing::ElementsAre("/", ".c", ".d", "..", "."));
   EXPECT_THAT(GetComponents("c:\\c\\e\\foo.txt", path::Style::windows),
               testing::ElementsAre("c:", "\\", "c", "e", "foo.txt"));
   EXPECT_THAT(GetComponents("//net/"), testing::ElementsAre("//net", "/"));
   EXPECT_THAT(GetComponents("//net/c/foo.txt"),
               testing::ElementsAre("//net", "/", "c", "foo.txt"));
 }
 
 TEST(Support, AbsolutePathIteratorEnd) {
   // Trailing slashes are converted to '.' unless they are part of the root path.
   SmallVector<std::pair<StringRef, path::Style>, 4> Paths;
   Paths.emplace_back("/foo/", path::Style::native);
   Paths.emplace_back("/foo//", path::Style::native);
-  Paths.emplace_back("//net//", path::Style::native);
   Paths.emplace_back("//net/foo/", path::Style::native);
   Paths.emplace_back("c:\\foo\\", path::Style::windows);
-  Paths.emplace_back("c:\\\\", path::Style::windows);
 
   for (auto &Path : Paths) {
     SCOPED_TRACE(Path.first);
     StringRef LastComponent = *path::rbegin(Path.first, Path.second);
     EXPECT_EQ(".", LastComponent);
   }
 
   SmallVector<std::pair<StringRef, path::Style>, 3> RootPaths;
   RootPaths.emplace_back("/", path::Style::native);
   RootPaths.emplace_back("//net/", path::Style::native);
   RootPaths.emplace_back("c:\\", path::Style::windows);
+  RootPaths.emplace_back("//net//", path::Style::native);
+  RootPaths.emplace_back("c:\\\\", path::Style::windows);
 
   for (auto &Path : RootPaths) {
     SCOPED_TRACE(Path.first);
     StringRef LastComponent = *path::rbegin(Path.first, Path.second);
     EXPECT_EQ(1u, LastComponent.size());
     EXPECT_TRUE(path::is_separator(LastComponent[0], Path.second));
   }
 }