fix(windows): handle invalid wide chars in file names (fixes gh #241)

For some reason, Windows allows invalid UTF-16 characters in file names.
Try to handle these gracefully when converting to UTF-8.

include/dwarfs/util.h

@@ -55,6 +55,7 @@ void utf8_sanitize(std::string& str);
 void shorten_path_string(std::string& path, char separator, size_t max_len);
 
 std::filesystem::path canonical_path(std::filesystem::path p);
+std::string path_to_utf8_string_sanitized(std::filesystem::path const& p);
 
 bool getenv_is_enabled(char const* var);
 

include/dwarfs/writer/internal/entry.h

@@ -121,6 +121,9 @@ class entry : public entry_interface {
  private:
   std::u8string u8name() const;
 
+#ifdef _WIN32
+  std::filesystem::path path_;
+#endif
   std::string name_;
   std::weak_ptr<entry> parent_;
   file_stat stat_;

src/util.cpp

@@ -26,6 +26,7 @@
 #include <cstdio>
 #include <cstdlib>
 #include <iostream>
+#include <type_traits>
 
 #if __has_include(<utf8cpp/utf8.h>)
 #include <utf8cpp/utf8.h>
@@ -264,6 +265,25 @@ std::filesystem::path canonical_path(std::filesystem::path p) {
   return p;
 }
 
+std::string path_to_utf8_string_sanitized(std::filesystem::path const& p) {
+#ifdef _WIN32
+  if constexpr (std::is_same_v<std::filesystem::path::value_type, wchar_t>) {
+    auto const& in = p.native();
+    if (in.empty()) {
+      return {};
+    }
+    int size_needed = ::WideCharToMultiByte(
+        CP_UTF8, 0, in.data(), (int)in.size(), NULL, 0, NULL, NULL);
+    std::string out(size_needed, 0);
+    ::WideCharToMultiByte(CP_UTF8, 0, in.data(), (int)in.size(), &out[0],
+                          size_needed, NULL, NULL);
+    return out;
+  }
+#endif
+
+  return u8string_to_string(p.u8string());
+}
+
 bool getenv_is_enabled(char const* var) {
   if (auto val = std::getenv(var)) {
     if (auto maybeBool = try_to<bool>(val); maybeBool && *maybeBool) {

src/writer/internal/entry.cpp

@@ -58,20 +58,19 @@ bool is_root_path(std::string_view path) {
 #endif
 }
 
-std::string entry_name(fs::path const& path, bool has_parent) {
-  if (has_parent) {
-    return u8string_to_string(path.filename().u8string());
-  }
-  return u8string_to_string(path.u8string());
-}
-
 } // namespace
 
 entry::entry(fs::path const& path, std::shared_ptr<entry> parent,
              file_stat const& st)
-    : name_{entry_name(path, static_cast<bool>(parent))}
+#ifdef _WIN32
+    : path_{parent ? path.filename() : path}
+    , name_{path_to_utf8_string_sanitized(path_)}
+#else
+    : name_{path_to_utf8_string_sanitized(parent ? path.filename() : path)}
+#endif
     , parent_{std::move(parent)}
-    , stat_{st} {}
+    , stat_{st} {
+}
 
 bool entry::has_parent() const {
   if (parent_.lock()) {
@@ -88,11 +87,17 @@ void entry::set_name(const std::string& name) { name_ = name; }
 std::u8string entry::u8name() const { return string_to_u8string(name_); }
 
 fs::path entry::fs_path() const {
+#ifdef _WIN32
+  fs::path self = path_;
+#else
+  fs::path self = name_;
+#endif
+
   if (auto parent = parent_.lock()) {
-    return parent->fs_path() / u8name();
+    return parent->fs_path() / self;
   }
 
-  return fs::path(u8name());
+  return self;
 }
 
 std::string entry::path_as_string() const {

src/writer/scanner.cpp

@@ -31,6 +31,7 @@
 #include <stdexcept>
 #include <string>
 #include <system_error>
+#include <unordered_set>
 #include <utility>
 #include <vector>
 
@@ -330,6 +331,7 @@ class scanner_ final : public scanner::impl {
   os_access const& os_;
   std::vector<std::unique_ptr<entry_filter>> filters_;
   std::vector<std::unique_ptr<entry_transformer>> transformers_;
+  std::unordered_set<std::string> invalid_filenames_;
 };
 
 template <typename LoggerPolicy>
@@ -362,6 +364,24 @@ scanner_<LoggerPolicy>::add_entry(std::filesystem::path const& name,
                                   file_scanner& fs, bool debug_filter) {
   try {
     auto pe = entry_factory_.create(os_, name, parent);
+
+    if constexpr (!std::is_same_v<std::filesystem::path::value_type, char>) {
+      try {
+        auto tmp [[maybe_unused]] = name.filename().u8string();
+      } catch (std::system_error const& e) {
+        LOG_ERROR << fmt::format(
+            "invalid file name in \"{}\", storing as \"{}\": {}",
+            path_to_utf8_string_sanitized(name.parent_path()), pe->name(),
+            e.what());
+        if (!invalid_filenames_.emplace(path_to_utf8_string_sanitized(name))
+                 .second) {
+          LOG_ERROR << fmt::format(
+              "cannot store \"{}\" as the name already exists", pe->name());
+          return nullptr;
+        }
+      }
+    }
+
     bool const exclude =
         std::any_of(filters_.begin(), filters_.end(), [&pe](auto const& f) {
           return f->filter(*pe) == filter_action::remove;
@@ -452,7 +472,8 @@ scanner_<LoggerPolicy>::add_entry(std::filesystem::path const& name,
 
     return pe;
   } catch (const std::system_error& e) {
-    LOG_ERROR << fmt::format("error reading entry (path={}): {}", name.string(),
+    LOG_ERROR << fmt::format("error reading entry (path={}): {}",
+                             path_to_utf8_string_sanitized(name),
                              exception_str(e));
     prog.errors++;
   }