Introduce fsyspath subclass of std::filesystem::path.

Our std::strings are UTF-8 encoded, so conversion from std::string to
std::filesystem::path must use UTF-8 decoding. The native Windows
std::filesystem::path constructor and assignment operator accepting
std::string use "native narrow encoding," which mangles path strings
containing UTF-8 encoded non-ASCII characters.

fsyspath's std::string constructor and assignment operator explicitly engage
std::filesystem::u8path() to handle encoding. u8path() is deprecated in C++20,
but once we adapt fsyspath's conversion to C++20 conventions, consuming code
need not be modified.

(cherry picked from commit e399b02e3306a249cb161f07cac578d3f2617bab)

2 files changed, 75 insertions, 0 deletions

diff --git a/indra/llcommon/CMakeLists.txt b/indra/llcommon/CMakeLists.txt
index 60549d9d11..a504e71340 100644
--- a/indra/llcommon/CMakeLists.txt
+++ b/indra/llcommon/CMakeLists.txt
@@ -119,6 +119,7 @@ set(llcommon_HEADER_FILES
     commoncontrol.h
     ctype_workaround.h
     fix_macros.h
+    fsyspath.h
     function_types.h
     indra_constants.h
     lazyeventapi.h
diff --git a/indra/llcommon/fsyspath.h b/indra/llcommon/fsyspath.h
new file mode 100644
index 0000000000..aa4e0132bc
--- /dev/null
+++ b/indra/llcommon/fsyspath.h
@@ -0,0 +1,74 @@
+/**
+ * @file   fsyspath.h
+ * @author Nat Goodspeed
+ * @date   2024-04-03
+ * @brief  Adapt our UTF-8 std::strings for std::filesystem::path
+ * 
+ * $LicenseInfo:firstyear=2024&license=viewerlgpl$
+ * Copyright (c) 2024, Linden Research, Inc.
+ * $/LicenseInfo$
+ */
+
+#if ! defined(LL_FSYSPATH_H)
+#define LL_FSYSPATH_H
+
+#include <filesystem>
+
+// While std::filesystem::path can be directly constructed from std::string on
+// both Posix and Windows, that's not what we want on Windows. Per
+// https://en.cppreference.com/w/cpp/filesystem/path/path:
+
+// ... the method of conversion to the native character set depends on the
+// character type used by source.
+// 
+// * If the source character type is char, the encoding of the source is
+//   assumed to be the native narrow encoding (so no conversion takes place on
+//   POSIX systems).
+// * If the source character type is char8_t, conversion from UTF-8 to native
+//   filesystem encoding is used. (since C++20)
+// * If the source character type is wchar_t, the input is assumed to be the
+//   native wide encoding (so no conversion takes places on Windows).
+
+// The trouble is that on Windows, from std::string ("source character type is
+// char"), the "native narrow encoding" isn't UTF-8, so file paths containing
+// non-ASCII characters get mangled.
+//
+// Once we're building with C++20, we could pass a UTF-8 std::string through a
+// vector<char8_t> to engage std::filesystem::path's own UTF-8 conversion. But
+// sigh, as of 2024-04-03 we're not yet there.
+//
+// Anyway, encapsulating the important UTF-8 conversions in our own subclass
+// allows us to migrate forward to C++20 conventions without changing
+// referencing code.
+
+class fsyspath: public std::filesystem::path
+{
+    using super = std::filesystem::path;
+
+public:
+    // default
+    fsyspath() {}
+    // construct from UTF-8 encoded std::string
+    fsyspath(const std::string& path): super(std::filesystem::u8path(path)) {}
+    // construct from UTF-8 encoded const char*
+    fsyspath(const char* path): super(std::filesystem::u8path(path)) {}
+    // construct from existing path
+    fsyspath(const super& path): super(path) {}
+
+    fsyspath& operator=(const super& p) { super::operator=(p); return *this; }
+    fsyspath& operator=(const std::string& p)
+    {
+        super::operator=(std::filesystem::u8path(p));
+        return *this;
+    }
+    fsyspath& operator=(const char* p)
+    {
+        super::operator=(std::filesystem::u8path(p));
+        return *this;
+    }
+
+    // shadow base-class string() method with UTF-8 aware method
+    std::string string() const { return super::u8string(); }
+};
+
+#endif /* ! defined(LL_FSYSPATH_H) */