Improve error reporting when serializing non-Unicode strings to JSON

* src/coding.c (utf8_string_p): New helper function.
(syms_of_coding) <utf-8-unix>: Move from json.c.

* src/json.c (json_check_utf8): New helper function.
(lisp_to_json_toplevel_1, lisp_to_json): Use it.  To save a bit of
time, check for invalid UTF-8 strings only after encountering an
error, since Jansson already rejects them.

* test/src/json-tests.el (json-serialize/invalid-unicode): Adapt
expected error symbol.

1 files changed, 22 insertions, 0 deletions

diff --git a/src/coding.c b/src/coding.c
index 1705838ffad..5ea1e395f20 100644
--- a/src/coding.c
+++ b/src/coding.c
@@ -6360,6 +6360,27 @@ check_utf_8 (struct coding_system *coding)
 }
 
 
+/* Return whether STRING is a valid UTF-8 string.  STRING must be a
+   unibyte string.  */
+
+bool
+utf8_string_p (Lisp_Object string)
+{
+  eassert (!STRING_MULTIBYTE (string));
+  struct coding_system coding;
+  setup_coding_system (Qutf_8_unix, &coding);
+  /* We initialize only the fields that check_utf_8 accesses.  */
+  coding.head_ascii = -1;
+  coding.src_pos = 0;
+  coding.src_pos_byte = 0;
+  coding.src_chars = SCHARS (string);
+  coding.src_bytes = SBYTES (string);
+  coding.src_object = string;
+  coding.eol_seen = EOL_SEEN_NONE;
+  return check_utf_8 (&coding) != -1;
+}
+
+
 /* Detect how end-of-line of a text of length SRC_BYTES pointed by
    SOURCE is encoded.  If CATEGORY is one of
    coding_category_utf_16_XXXX, assume that CR and LF are encoded by
@@ -10846,6 +10867,7 @@ syms_of_coding (void)
   DEFSYM (Qiso_2022, "iso-2022");
 
   DEFSYM (Qutf_8, "utf-8");
+  DEFSYM (Qutf_8_unix, "utf-8-unix");
   DEFSYM (Qutf_8_emacs, "utf-8-emacs");
 
 #if defined (WINDOWSNT) || defined (CYGWIN)