X-Git-Url: https://diplodocus.org/git/nmh/blobdiff_plain/fc6ab5f54b16af90d04ca2dc70af82b9e97ebb9c..37c403585f3a2831347ab6785ab5cb86d5064433:/sbr/encode_rfc2047.c

diff --git a/sbr/encode_rfc2047.c b/sbr/encode_rfc2047.c
index 0df5f706..ac2b6dc3 100644
--- a/sbr/encode_rfc2047.c
+++ b/sbr/encode_rfc2047.c
@@ -125,14 +125,14 @@ encode_rfc2047(const char *name, char **value, int encoding,
      * On the encoding we choose, and the specifics of encoding:
      *
      * - If a specified encoding is passed in, we use that.
-     * - If more than 50% of the characters are high-bit, we use base64
-     *   and encode the whole field as one atom (possibly split).
-     * - Otherwise, we use quoted-printable.
+     * - Otherwise, pick which encoding is shorter.
+     *
+     * We don't quite handle continuation right here, but it should be
+     * pretty close.
      */
 
     if (encoding == CE_UNKNOWN)
-    	encoding = (eightbitcount * 10 / (asciicount + eightbitcount) > 5) ?
-						CE_BASE64 : CE_QUOTED;
+        encoding = pref_encoding(asciicount, qpspecialcount, eightbitcount);
 
     unfold_header(value, asciicount + eightbitcount);
 
@@ -286,10 +286,13 @@ field_encode_quoted(const char *name, char **value, const char *charset,
 	}
     }
 
-    strcat(q, "?=");
+    *q++ = '?';
+    *q++ = '=';
 
     if (prefixlen)
-    	strcat(q, "\n");
+    	*q++ = '\n';
+
+    *q = '\0';
 
     free(*value);
 
@@ -514,8 +517,8 @@ unfold_header(char **value, int len)
 	     * This has the side effect of stripping off the final newline
 	     * for the header; we put it back in the encoding routine.
 	     */
-	    while (is_fws(*q++))
-	    	;
+	    while (is_fws(*q))
+	    	q++;
 	    if (*q == '\0')
 	    	break;
 
@@ -711,13 +714,27 @@ do_reformat:
 	 * we can use m_text directly.
 	 */
 
+	/*
+	 * If we were in a group but are no longer, make sure we add a
+	 * semicolon (which needs to be FIRST, as it needs to be at the end
+	 * of the last address).
+	 */
+
+	if (groupflag && ! mn->m_ingrp) {
+	    output = add(";", output);
+	    column += 1;
+	}
+
+	groupflag = mn->m_ingrp;
+
+	if (mn->m_gname) {
+	    cp = add(mn->m_gname, NULL);
+	}
+
 	if (reformat) {
-	    if (mn->m_gname) {
-	    	cp = add(mn->m_gname, NULL);
-	    }
 	    cp = add(adrformat(mn), cp);
 	} else {
-	    cp = add(mn->m_text, NULL);
+	    cp = add(mn->m_text, cp);
 	}
 
 	len = strlen(cp);
@@ -749,21 +766,21 @@ do_reformat:
 	column += len;
 	free(cp);
 	cp = NULL;
+    }
 
-	/*
-	 * If we were in a group but are no longer, make sure we add a
-	 * trailing semicolon.
-	 */
-
-	if (groupflag && ! mn->m_ingrp) {
-	    output = add(";", output);
-	}
+    /*
+     * Just in case we're at the end of a list
+     */
 
-	groupflag = mn->m_ingrp;
+    if (groupflag) {
+	output = add(";", output);
     }
 
-   *value = output;
-   output = NULL;
+    output = add("\n", output);
+
+    free(*value);
+    *value = output;
+    output = NULL;
 
 out:
 
@@ -790,7 +807,12 @@ scanstring(const char *string, int *asciilen, int *eightbitchars,
     for (; *string != '\0'; string++) {
     	if ((isascii((unsigned char) *string))) {
 	    (*asciilen)++;
-	    if (!qphrasevalid((unsigned char) *string))
+	    /*
+	     * So, a space is not a valid phrase character, but we're counting
+	     * an exception here, because in q-p a space can be directly
+	     * encoded as an underscore.
+	     */
+	    if (!qphrasevalid((unsigned char) *string) && *string != ' ')
 	    	(*specialchars)++;
 	} else {
 	    (*eightbitchars)++;