Re: another Linux/SE zombie problem

James Cardwell wrote: .. > If I define sesoctcp as the protocol, and start the sqlexecd daemon, > (either alone or with the 'nozombie' program) - zombie processes are > created after using dbaccess on the linux box - everything else works > fine, (remote connect from my desktop, and the Perl/DBI scripts - no > zombies are created via these connects) Same thing happens to me. Without nozombie, the zombies are sqlexecd processes. With nozombie, the zombies are sqlexec processes. Apparently since sqlexecd is just a gateway to sqlexec it reenables signal handling for SIGCHLD and does not perform the waitpid(). > If I define seipcpip as the protocol, and do not start sqlexecd, then > dbaccess on the linux box and Perl/DBI both work fine - and no zombie > processes are created. However, I cannot connect from my NT box by > defining seipcpip in SETNET32. > > It seems to me that using sesoctcp as a protocol is requried for > comapatability with later online utilities. Is there: RTFM. seipcpip uses unnamed pipes for communication. Unnamed pipes exist on the LOCAL FILESYSTEM. Therefore you can't connect at all from a remote machine. Methinks this is the kind of thing Informix would do to those of us who got "developer licenses." After all, you don't need remote connectivity to be a developer, right? Sigh. I'm waiting for my Oracle8/Linux prerelease CD to come in the mail. There has been an issue with missing Pro*C files there, but apparently Oracle has fixed it already. -- Cheers, Orly