fix looks_like_string for empty strings and UTF-8, add cross-module includes in codegen

2026-05-03 00:27:20 -05:00
parent 3d71db4958
commit e180baf776
4 changed files with 207 additions and 100 deletions
@@ -2939,8 +2939,13 @@ static int looks_like_string(el_val_t v) {
    const unsigned char* s = (const unsigned char*)p;
    for (int i = 0; i < 16; i++) {
        unsigned char c = s[i];
-        if (c == '\0') return i > 0; /* terminated string */
-        if (c < 0x09 || (c > 0x0d && c < 0x20) || c >= 0x7f) return 0;
+        if (c == '\0') return 1; /* terminated string (empty string is still a valid string) */
+        /* Reject C0 control chars (non-whitespace), allow UTF-8 high bytes.
+         * 0x09-0x0d = tab/newline/cr/vt/ff (whitespace, OK)
+         * 0x20-0x7e = printable ASCII (OK)
+         * 0x7f      = DEL (reject)
+         * 0x80-0xff = UTF-8 continuation/lead bytes (OK for multi-byte chars) */
+        if (c < 0x09 || (c > 0x0d && c < 0x20) || c == 0x7f) return 0;
    }
    return 1; /* 16+ printable bytes — call it a string */
 }