Re: charset_to_utf8 assertion of remaining src_size
We're using FreeBSD 10.1
Thanks
On Thu, May 21, 2015 at 4:56 PM, Xin Li <delphij@delphij.net> wrote:
-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA512
On 05/21/15 15:47, Cassidy Larson wrote:
I've been seeing this same error as well now:
Fatal: master: service(indexer-worker): child 20979 killed with signal 6 (core not dumped - set service indexer-worker { drop_priv_before_exec=yes }) Panic: file charset-iconv.c: line 132 (charset_to_utf8): assertion failed: (*src_size - pos <= CHARSET_MAX_PENDING_BUF_SIZE)
Quick question -- Are you using FreeBSD (10.0+)?
Cheers,
Xin LI <delphij@delphij.net> https://www.delphij.net/ FreeBSD - The Power to Serve! Live free or die -----BEGIN PGP SIGNATURE----- Version: GnuPG v2.1.4 (FreeBSD)
iQIcBAEBCgAGBQJVXmKfAAoJEJW2GBstM+nspmoP/0CKS0g8ka/4uPF+jaq7+DD6 1PxCufdz/p528rclbXvEK3XE++sULihNVtO24TgXMOIJZB6jCfbfo+O17N9ZS7g9 Vaz0r0URU28joAeDX/NR1yzYpHdo277RhXiPlKuyep+fEgi6Szm/mNaR+aCtYrge 8Qm2U2eaHFQk/2qpyvC4ZlFzXSVIR+AIMSh87YqJSWSVd3PjNDLxZLDeUaQyygG5 1lW3lsy0yDOTISI58rXavOi7qa4qKo0ySDOMLct3GrdXzL5S8N/Pc3RnRTP5p9dY 22cGoH3XloijWAef93qHSuPFuC9Yiko4g/FZAtfuQg6mCETvhPDfhXBgRf3IFYAr vuMOyBX+3TrQQI+zz3aolou10WH+E+PPbJFFZcNq4sRGQYtSXBA4RmDDKU4NgwoB gIzS43GjPidJTxdzEC+eedmS6Rt5wvyFaDK1h5+x/mgnfEie6E9Hfk8IBqRHZMUW HixCmBnC3p1GN+R8rbPxjuycP06bB7YXLJ28o06gM2agrGv5vu83j+WqO2g6wJIP hRKnyCVVPUeAnkLpS+7nXZwW5R/j8M5y7V8rbiYCX4q+zWCJZl/ac1G12zwva5k1 MwyjciDP951soMM2RYsQQSE5RxRKMC03kqGt+qnBx3jBR7afzhOmcO9gcKQJEY5R qjwnZ0voZtV7u2Ptcv1Q =hWt9 -----END PGP SIGNATURE-----
On 05/22/15 13:18, Cassidy Larson wrote:
We're using FreeBSD 10.1
I see. Yes that's the same problem I have seen.
There is a behavioral difference (I think it's a FreeBSD bug) between FreeBSD's iconv(3) and GNU implementation, and there is arguably a bug with Dovecot that iconv(3) state should be reset for each multipart part. The two together would trigger the problem when the previous part was ended with UTF-7 shift sequence. Timo have applied a fix (hg: 18721:9809f68aaa36) for Dovecot which should mitigates this issue at Dovecot side a few days ago.
I have created a reproduction program and discussed the issue with our iconv(3) developer, Tijl Coosemans <tijl@>, who have created a patch to correct this FreeBSD iconv(3) bug (not yet committed against -HEAD but I'd expect it happening soon).
Cheers,
Xin LI <delphij@delphij.net> https://www.delphij.net/ FreeBSD - The Power to Serve! Live free or die
On 22 May 2015, at 17:44, Xin Li <delphij@delphij.net> wrote:
On 05/22/15 13:18, Cassidy Larson wrote:
We're using FreeBSD 10.1
I see. Yes that's the same problem I have seen.
There is a behavioral difference (I think it's a FreeBSD bug) between FreeBSD's iconv(3) and GNU implementation, and there is arguably a bug with Dovecot that iconv(3) state should be reset for each multipart part. The two together would trigger the problem when the previous part was ended with UTF-7 shift sequence. Timo have applied a fix (hg: 18721:9809f68aaa36) for Dovecot which should mitigates this issue at Dovecot side a few days ago.
Could you or somebody help with getting this included as a unit test to Dovecot? I've attached a patch that I guess might cause the crash.
participants (3)
-
Cassidy Larson
-
Timo Sirainen
-
Xin Li