Blame SOURCES/mailx-12.5-encsplit.patch

c88ac0
diff --git a/mime.c b/mime.c
c88ac0
index 45de80a..f9fbadf 100644
c88ac0
--- a/mime.c
c88ac0
+++ b/mime.c
c88ac0
@@ -1109,16 +1109,34 @@ fromhdr_end:
c88ac0
 }
c88ac0
 
c88ac0
 /*
c88ac0
+ * return length of this UTF-8 codepoint in bytes
c88ac0
+ */
c88ac0
+static size_t
c88ac0
+codepointsize(char tc)
c88ac0
+{
c88ac0
+	int rv = 0;
c88ac0
+	if ( ! ( tc & 0x80 ) )
c88ac0
+		return 1;
c88ac0
+	while ( tc & 0x80 )
c88ac0
+	{
c88ac0
+		rv++;
c88ac0
+		tc = tc<<1;
c88ac0
+	}
c88ac0
+	return rv;
c88ac0
+}
c88ac0
+
c88ac0
+/*
c88ac0
  * Convert header fields to RFC 1522 format and write to the file fo.
c88ac0
  */
c88ac0
 static size_t
c88ac0
 mime_write_tohdr(struct str *in, FILE *fo)
c88ac0
 {
c88ac0
 	char *upper, *wbeg, *wend, *charset, *lastwordend = NULL, *lastspc, b,
c88ac0
-		*charset7;
c88ac0
+		*charset7, *cp;
c88ac0
 	struct str cin, cout;
c88ac0
-	size_t sz = 0, col = 0, wr, charsetlen, charset7len;
c88ac0
+	size_t sz = 0, col = 0, wr, charsetlen, charset7len, cpsz;
c88ac0
 	int quoteany, mustquote, broken,
c88ac0
+		maxin, maxout, curin, cps,
c88ac0
 		maxcol = 65 /* there is the header field's name, too */;
c88ac0
 
c88ac0
 	upper = in->s + in->l;
c88ac0
@@ -1134,41 +1152,75 @@ mime_write_tohdr(struct str *in, FILE *fo)
c88ac0
 		if (mustquote_hdr(wbeg, wbeg == in->s, wbeg == &upper[-1]))
c88ac0
 			quoteany++;
c88ac0
 	}
c88ac0
+
c88ac0
+	/*
c88ac0
+	 * rfc2047 says we cannot split multi-byte characters over
c88ac0
+	 * encoded words, so we need to know if we're a multi-byte
c88ac0
+	 * source stream (UTF-8 specifically) or just an 8 bit
c88ac0
+	 * stream like ISO-8859-15
c88ac0
+	 * so test beginning of charset since it is valid to include
c88ac0
+	 * language in charset "UTF-8*DE" etc as per rfc 2184/2231
c88ac0
+	 */
c88ac0
+	char *thisset = b&0200 ? charset : charset7;
c88ac0
+	int is_utf8 = ( strncasecmp( thisset, "utf-8", 5 ) == 0 );
c88ac0
+
c88ac0
 	if (2 * quoteany > in->l) {
c88ac0
 		/*
c88ac0
 		 * Print the entire field in base64.
c88ac0
 		 */
c88ac0
-		for (wbeg = in->s; wbeg < upper; wbeg = wend) {
c88ac0
+		for (wbeg = in->s; wbeg < upper; ) {
c88ac0
 			wend = upper;
c88ac0
 			cin.s = wbeg;
c88ac0
-			for (;;) {
c88ac0
-				cin.l = wend - wbeg;
c88ac0
-				if (cin.l * 4/3 + 7 + charsetlen
c88ac0
-						< maxcol - col) {
c88ac0
-					fprintf(fo, "=?%s?B?",
c88ac0
-						b&0200 ? charset : charset7);
c88ac0
-					wr = mime_write_tob64(&cin, fo, 1);
c88ac0
-					fwrite("?=", sizeof (char), 2, fo);
c88ac0
-					wr += 7 + charsetlen;
c88ac0
-					sz += wr, col += wr;
c88ac0
-					if (wend < upper) {
c88ac0
-						fwrite("\n ", sizeof (char),
c88ac0
-								2, fo);
c88ac0
-						sz += 2;
c88ac0
-						col = 0;
c88ac0
-						maxcol = 76;
c88ac0
+			/*
c88ac0
+			 * we calculate the maximum number of bytes
c88ac0
+			 * we can use on this output line, and then what
c88ac0
+			 * this equates to as base64 encoded source bytes
c88ac0
+			 */
c88ac0
+			maxout = maxcol - col - 7 - charsetlen;
c88ac0
+			maxin = (maxout - (maxout & 0x03)) * 3/4;
c88ac0
+
c88ac0
+			/* short enough to finish ? */
c88ac0
+			if (maxin > upper - wbeg )
c88ac0
+			{
c88ac0
+				curin = upper - wbeg;
c88ac0
+				wbeg += curin;
c88ac0
+			}else
c88ac0
+			{
c88ac0
+				if (is_utf8)
c88ac0
+				{
c88ac0
+					/*
c88ac0
+					 * now scan the input from the beginning
c88ac0
+					 * to see how many codepoints will fit
c88ac0
+					 */
c88ac0
+					curin = 0;
c88ac0
+					while (curin < maxin
c88ac0
+						&& (cpsz = codepointsize(*wbeg)) <= (maxin - curin))
c88ac0
+					{
c88ac0
+						curin += cpsz;
c88ac0
+						wbeg += cpsz;
c88ac0
 					}
c88ac0
-					break;
c88ac0
-				} else {
c88ac0
-					if (col) {
c88ac0
-						fprintf(fo, "\n ");
c88ac0
-						sz += 2;
c88ac0
-						col = 0;
c88ac0
-						maxcol = 76;
c88ac0
-					} else
c88ac0
-						wend -= 4;
c88ac0
+				}else
c88ac0
+				{
c88ac0
+					curin = maxin;
c88ac0
+					wbeg += maxin;
c88ac0
 				}
c88ac0
 			}
c88ac0
+			cin.l = curin;
c88ac0
+			fprintf(fo, "%s=?%s?B?", (cin.s != in->s) ? " " : "", thisset );
c88ac0
+			wr = mime_write_tob64(&cin, fo, 1);
c88ac0
+
c88ac0
+			if (wbeg < upper)
c88ac0
+			{
c88ac0
+				wr += fwrite("?=\n ", sizeof (char), 4, fo) * sizeof (char);
c88ac0
+			}else
c88ac0
+			{
c88ac0
+				wr += fwrite("?=", sizeof (char), 2, fo) * sizeof (char);
c88ac0
+			}
c88ac0
+
c88ac0
+			/* and shuffle pointers and counts */
c88ac0
+			col = 1;
c88ac0
+			maxcol = 76;
c88ac0
+			sz += wr + 7 + charsetlen + ((cin.s != in->s) ? 1 : 0 );
c88ac0
 		}
c88ac0
 	} else {
c88ac0
 		/*
c88ac0
@@ -1243,7 +1295,29 @@ mime_write_tohdr(struct str *in, FILE *fo)
c88ac0
 								maxcol -= wbeg -
c88ac0
 									lastspc;
c88ac0
 						} else {
c88ac0
-							wend -= 4;
c88ac0
+							if (is_utf8)
c88ac0
+							{
c88ac0
+								/*
c88ac0
+								 * make sure wend is not pointing to
c88ac0
+								 * the middle of a codepoint
c88ac0
+								 */
c88ac0
+								cp = wend;
c88ac0
+								while (--cp > wbeg)
c88ac0
+								{
c88ac0
+									cps = codepointsize(*cp);
c88ac0
+									if (cps > 1)
c88ac0
+									{
c88ac0
+										if (wend - cp - cps > 4)
c88ac0
+											wend -= 4;
c88ac0
+										else
c88ac0
+											wend = cp;
c88ac0
+										break;
c88ac0
+									}
c88ac0
+								}
c88ac0
+								if (cp == wbeg)
c88ac0
+									wend -= 4;
c88ac0
+							} else
c88ac0
+								wend -= 4;
c88ac0
 						}
c88ac0
 						free(cout.s);
c88ac0
 					}