fixed auto-completion of Unicode file names

* teco_string_diff() could return a number of bytes in the middle of
  an Unicode sequence. It now also requires Unicode strings.
* Added a missing Unicode-validity check when replacing command lines (`{` and `}`).
  teco_cmdline_insert() should really be refactored, though (FIXME).
* Added test case

2 files changed, 28 insertions, 6 deletions

diff --git a/src/cmdline.c b/src/cmdline.c
index fa69d91..9a7b356 100644
--- a/src/cmdline.c
+++ b/src/cmdline.c
@@ -162,12 +162,32 @@ teco_cmdline_insert(const gchar *data, gsize len, GError **error)
 				 * Result of command line replacement (}):
 				 * Exchange command lines
 				 */
+				g_clear_error(&tmp_error);
+
 				teco_qreg_t *cmdline_reg = teco_qreg_table_find(&teco_qreg_table_globals, "\e", 1);
 				g_auto(teco_string_t) new_cmdline = {NULL, 0};
 
 				if (!cmdline_reg->vtable->get_string(cmdline_reg, &new_cmdline.data, &new_cmdline.len,
-				                                     NULL, error))
+				                                     NULL, &tmp_error)) {
+					teco_error_add_frame_toplevel();
+					teco_error_display_short(tmp_error);
+					g_propagate_error(error, g_steal_pointer(&tmp_error));
 					return FALSE;
+				}
+
+				/*
+				 * SciTECO code must always be UTF-8, but you can smuggle arbitrary bytes
+				 * into the "\e" register.
+				 * This would be cumbersome to test for in teco_state_start_cmdline_pop().
+				 */
+				if (!teco_string_validate_utf8(new_cmdline)) {
+					g_set_error_literal(&tmp_error, TECO_ERROR, TECO_ERROR_CODEPOINT,
+					                    "Invalid UTF-8 byte sequence in command-line replacement");
+					teco_error_add_frame_toplevel();
+					teco_error_display_short(tmp_error);
+					g_propagate_error(error, g_steal_pointer(&tmp_error));
+					return FALSE;
+				}
 
 				/*
 				 * Search for first differing character in old and
@@ -229,7 +249,7 @@ teco_cmdline_insert(const gchar *data, gsize len, GError **error)
 				}
 			}
 
-			/* error is handled in teco_cmdline_keypress_c() */
+			/* error is handled in teco_cmdline_keypress() */
 			g_propagate_error(error, g_steal_pointer(&tmp_error));
 			return FALSE;
 		}
diff --git a/src/string-utils.c b/src/string-utils.c
index e9dd148..d98b6b0 100644
--- a/src/string-utils.c
+++ b/src/string-utils.c
@@ -87,8 +87,10 @@ teco_string_get_coord(const gchar *str, gsize off, guint *pos, guint *line, guin
 }
 
 /**
- * Get the length of the prefix common to two strings.
- * Works with UTF-8 and single-byte encodings.
+ * Get the length of the prefix common to two UTF-8 strings.
+ *
+ * The UTF-8 strings must be validated, which should be the case
+ * for help labels and short Q-Register names.
  *
  * @param a Left string.
  * @param b Right string.
@@ -103,8 +105,8 @@ teco_string_diff(teco_string_t a, const gchar *b, gsize b_len)
 	gsize len = 0;
 
 	while (len < a.len && len < b_len &&
-	       a.data[len] == b[len])
-		len++;
+	       g_utf8_get_char(a.data+len) == g_utf8_get_char(b+len))
+		len = g_utf8_next_char(b+len) - b;
 
 	return len;
 }