Inetd makes some (at least for 0.99.6) broken assumptions. Linux allows the situation to occur when you select, are told there is a new connection and before you do the accept the connection vanishes. Set the socket nonblocking before the accept and back to blocking just afterwards. In addition you may need to check the SIGCLD handling if you get 'Transport endpoint not connected'. Alan
| Rafael J. Wysocki | Re: 2.6.27-rc1 mtrr fixes do not work |
| Denis Cheng | [PATCH] trivial: the memset operation on a automatic array variable should be opti... |
| Andrea Arcangeli | Re: [PATCH 02 of 11] get_task_mm |
| Andrew Morton | Re: [GIT *] Allow request_firmware() to be satisfied from in-kernel, use it in mor... |
git: | |
| Johannes Schindelin | Re: git-clean buglet |
| Mark Levedahl | Re: mingw, windows, crlf/lf, and git |
| Jon Smirl | ! [rejected] master -> master (non-fast forward) |
| Sam Vilain | [RFC] Authenticate push via PGP signature, not SSH |
| Kevin Neff | Patching a SSH 'Weakness' |
| Frank Bax | QEMU crashes |
| Julien TOUCHE | setting up ssh tunnel/vpn |
| Darrin Chandler | Re: SMTP flood + spamdb |
| Jussi Kivilinna | [PATCH net-next-2.6 v4 1/2] net_sched: add size table functions |
| Eric Dumazet | [PATCH] fs: pipe/sockets/anon dentries should not have a parent |
| Patrick McHardy | Re: pkt_sched: add DRR scheduler |
| Matheos Worku | 2.6.24 BUG: soft lockup - CPU#X |
