Hi Steven. On 27.01.2014 21:22, Steven Chamberlain wrote: > I could reproduce it 8 times out of 100 here on kfreebsd-amd64. > > The race happens after getting Connection Refused trying to contact the > fcgi-responder which has just exit (intentionally). > > The parent does a wait4() syscall, and if this returns the pid of the > process that just exit, a new listen is set up on 10000/tcp and lighttpd > forks a new fcgi-responder. The parent waits with select() for it to be > ready, then sends the FCGI request which is successful: is this a libc portability issue then? All in all this does not really sound like a problem in the lighttpd code base, does it? -- with kind regards, Arno Töll IRC: daemonkeeper on Freenode/OFTC GnuPG Key-ID: 0x9D80F36D
Attachment:
signature.asc
Description: OpenPGP digital signature