On Tue, Jan 16, 2007, Paul J Stevens <[EMAIL PROTECTED]> said: > Aaron Stone wrote: >> On Sat, Jan 13, 2007, Marc Dirix <[EMAIL PROTECTED]> said: >> >>> I favor a monolithic threading approach, thus having non-blocking IO. >>> Intelligence needs to be kept at the place where it is needed, inside >>> the thread. >> >> Yep, I'm going with non-blocking sockets and a read/write wrapper >> (ci_read, and ci_write, which we already have :-) that uses select for >> timeouts. > > Beware there are possibly still some traces of fprintf around in > timsieve.c and lmtp.c
There's also a stream printf used for a lot of imap output through gmime. I hope that gmime knows how to handle non-blocking sockets... The mandate that no functions may have signal side effects (unless they are fatal, SIGSEGV, duh ;-) is pretty serious. Aaron _______________________________________________ Dbmail-dev mailing list Dbmail-dev@dbmail.org http://twister.fastxs.net/mailman/listinfo/dbmail-dev
