fix: sanitize UTF-8 strings, plus tests (fixes gh #191)

include/dwarfs/util.h

@@ -63,6 +63,7 @@ int call_sys_main_iolayer(std::span<std::string> args, iolayer const& iol,
 size_t utf8_display_width(char const* p, size_t len);
 size_t utf8_display_width(std::string const& str);
 void utf8_truncate(std::string& str, size_t len);
+void utf8_sanitize(std::string& str);
 
 void shorten_path_string(std::string& path, char separator, size_t max_len);
 

src/dwarfs/console_writer.cpp

@@ -91,9 +91,15 @@ void output_context_line(terminal const& term, std::ostream& os,
 
   assert(width >= progress_w + speed_w + 1);
 
+  std::string path;
+  if (st.path) {
+    path = *st.path;
+    utf8_sanitize(path);
+  }
+
   size_t status_w = width - (progress_w + speed_w + 1);
-  auto path_len = st.path ? utf8_display_width(*st.path) : 0;
-  size_t extra_len = st.path && !st.status_string.empty() ? 2 : 0;
+  auto path_len = !path.empty() ? utf8_display_width(path) : 0;
+  size_t extra_len = !path.empty() && !st.status_string.empty() ? 2 : 0;
 
   if (status_w <
       st.context.size() + st.status_string.size() + path_len + extra_len) {
@@ -106,11 +112,10 @@ void output_context_line(terminal const& term, std::ostream& os,
 
       if (max_path_len > 0) {
         shorten_path_string(
-            *st.path,
-            static_cast<char>(std::filesystem::path::preferred_separator),
+            path, static_cast<char>(std::filesystem::path::preferred_separator),
             max_path_len);
 
-        path_len = utf8_display_width(*st.path);
+        path_len = utf8_display_width(path);
       }
     }
 
@@ -128,7 +133,7 @@ void output_context_line(terminal const& term, std::ostream& os,
     if (!st.status_string.empty()) {
       st.status_string += ": ";
     }
-    st.status_string += *st.path;
+    st.status_string += path;
   }
 
   std::string progress;

src/dwarfs/scanner.cpp

@@ -265,6 +265,7 @@ std::string status_string(progress const& p, size_t width) {
       label = "writing: ";
       path = i->any()->path_as_string();
     }
+    utf8_sanitize(path);
     shorten_path_string(
         path, static_cast<char>(std::filesystem::path::preferred_separator),
         width - label.size());

src/dwarfs/util.cpp

@@ -265,6 +265,12 @@ void utf8_truncate(std::string& str, size_t len) {
   str.resize(p - str.data());
 }
 
+void utf8_sanitize(std::string& str) {
+  if (!utf8::is_valid(str)) [[unlikely]] {
+    str = utf8::replace_invalid(str);
+  }
+}
+
 void shorten_path_string(std::string& path, char separator, size_t max_len) {
   auto len = utf8_display_width(path);
 

test/tool_main_test.cpp

@@ -154,6 +154,7 @@ struct random_file_tree_options {
   int dimension{20};
   int max_name_len{50};
   bool with_errors{false};
+  bool with_invalid_utf8{false};
 };
 
 class mkdwarfs_tester : public tester_common {
@@ -193,10 +194,14 @@ class mkdwarfs_tester : public tester_common {
     std::mt19937_64 rng{42};
     std::exponential_distribution<> size_dist{1 / opt.avg_size};
     std::uniform_int_distribution<> path_comp_size_dist{0, opt.max_name_len};
+    std::uniform_int_distribution<> invalid_dist{0, 1};
     std::vector<std::pair<fs::path, std::string>> paths;
 
     auto random_path_component = [&] {
       auto size = path_comp_size_dist(rng);
+      if (opt.with_invalid_utf8 && invalid_dist(rng) == 0) {
+        return test::create_random_string(size, 96, 255, rng);
+      }
       return test::create_random_string(size, 'A', 'Z', rng);
     };
 
@@ -213,7 +218,7 @@ class mkdwarfs_tester : public tester_common {
           auto size = std::min(max_size, static_cast<size_t>(size_dist(rng)));
           std::string data;
 
-          if (rng() % 2 == 0) {
+          if (size < 1024 * 1024 && rng() % 2 == 0) {
             data = test::create_random_string(size, rng);
           } else {
             data = test::loremipsum(size);
@@ -1734,7 +1739,14 @@ TEST_P(mkdwarfs_progress_test, basic) {
   t.iol->set_terminal_fancy(true);
 
   t.add_root_dir();
-  t.add_random_file_tree();
+  t.add_random_file_tree({
+      .avg_size = 20.0 * 1024 * 1024,
+      .dimension = 2,
+#ifndef _WIN32
+      // Windows can't deal with non-UTF-8 filenames
+      .with_invalid_utf8 = true,
+#endif
+  });
   t.os->add_local_files(audio_data_dir);
   t.os->add_local_files(fits_data_dir);