From: Per Jessen Date: August 8 2012 4:11pm Subject: RE: UDF writing to unix socket - segfaults? List-Archive: http://lists.mysql.com/mysql/227972 Message-Id: MIME-Version: 1.0 Content-Type: text/plain; charset=utf-8 Content-Transfer-Encoding: quoted-printable Martin Gainty wrote: > this is pure speculation unless we can get ahold of the source code > for your specific version of glibc and determine what the maximum > sizes are .. otherwise anything I suggest would be speculative..lets > take a look at > http://fossies.org/dox/glibc-2.16.0/sysdeps_2mach_2hurd_2sendto_8c_sour= ce.html >=20 > ssize_t > 29 __sendto (int fd, > 30 const void *buf, > 31 size_t n, > 32 int flags, > 33 const struct sockaddr_un *addr, > 34 socklen_t addr_len) > Here his maximum buffer length for sending is unsigned int > specifically size_t so.. > are both send and receive entities IPv4 or both entities are IPv6 > if thats the case.. > can you send > 0 Bytes > 2 bytes > 4 bytes > 8 bytes > at what maximum length of buffer does the segfault occur > ? Interesting that you should pick on the length - I think that might be it. When I switched to using a UDP socket, sendto() did complain about= the length. (error 90, message too long). Hmm, that's probably a UDP packet-size restriction, whereas no such restriction exist for a unix socket (I presume). Thanks for the idea! --=20 Per Jessen, Z=C3=BCrich (24.4=C2=B0C)