Bug #53891 [Com]: iconv_mime_encode() fails converting UTF-8 string to quoted-printable

Edit report at https://bugs.php.net/bug.php?id=53891&edit=1 ID: 53891 Comment by: af at ins dot de Reported by: can at strg dot at Summary: iconv_mime_encode() fails converting UTF-8 string to quoted-printable Status: Verified Type: Bug Package: ICONV related Operating System: GNU/linux (gentoo/kernel 2.6.35) PHP Version: 5.3.5 Block user comment: N Private report: N New Comment: Sorry. ZendFramework-version is obviouisly not 5.6.32, it is: 2.4.13 Previous Comments: ------------------------------------------------------------------------ [2017-12-06 08:27:28] af at ins dot de We have encountered the same bug when encoding UTF-8 strings with PHP 5.6.32 and PHP 7.1.7. A fix after this long time would be really nice. As a sidenote: This error impacts also the ZendFramework (5.6.32) because Zend/Mail;Zend/Header is checking with iconv_mime_encode() if a subject line is valid and throws a "Subject value must be composed of printable US-ASCII or UTF-8 characters" instead of sending the mail when the bug occurs. ------------------------------------------------------------------------ [2016-08-13 11:12:32] cmb@php.net Only changing the threshold does not work, unfortunately. ------------------------------------------------------------------------ [2015-09-09 22:18:18] cmb@php.net The culprit is the check whether there is sufficient place in the line to add another encoded word[1]. out_charset_len + 12 is not sufficient for quoted-printable encoding of UTF-8. In the worst case it is out_charset_len + BOILERPLATE + CHARS_PER_BYTE * MAX_BYTES (where BOILERPLATE == 7, CHARS_PER_BYTE == 3 and MAX_BYTES == 6), i.e. out_charset_len + 25. AFAIK, this would be sufficient for all other character encodings and for Base64 encoding as well. Of course, changing this threshold might change the result of iconv_mime_encode with regard to folding, but that supposedly would only break some tests, if they're relying on the exact output. It would more severly affect any code that uses small line-length values, but presumably such code doesn't exist – actually, I can't see a reason to set line-length < 76. [1] <https://github.com/php/php-src/blob/php-7.0.0RC2/ext/iconv/iconv.c#L1238> ------------------------------------------------------------------------ [2015-09-09 18:09:21] cmb@php.net Confirmed: <https://3v4l.org/JLman> Error 7 is likely E2BIG, what occurs if there's not enough room at *outbuf when calling iconv(). ------------------------------------------------------------------------ [2011-01-31 14:25:20] can at strg dot at Description: ------------ iconv_mime_encode() fails on certain UTF-8 strings containing at least one multibyte character. The error is reproducable for these strings. I have further noticed that it is more likely that this error occurs if a multibyte character is found at the following character positions within the string: 24±1, 52±3, 82±4, 112±5, 143±5 The code that led me to the previous conclusion can be found here: http://pastebin.com/GXQZ129M Test script: --------------- $preferences = array( 'scheme' => 'Q', 'input-charset' => 'utf-8', 'output-charset' => 'utf-8', 'line-length' => 74, 'line-break-chars' => "\r\n", ); // Note the ellipsis character at the end of the following string: iconv_mime_encode('subject', "d obeybiubrsfqllpdtpge…", $preferences); Actual result: -------------- PHP Notice: iconv_mime_encode(): Unknown error (7) in /home/soulmerge/test.php on line 11 Notice: iconv_mime_encode(): Unknown error (7) in /home/soulmerge/test.php on line 11 ------------------------------------------------------------------------ -- Edit this bug report at https://bugs.php.net/bug.php?id=53891&edit=1