Bug #165

crash on startup on SIGPIPE

Added by Anonymous over 9 years ago. Updated almost 7 years ago.

Status:FixedStart date:
Priority:NormalDue date:
Assignee:-% Done:

0%

Category:core
Target version:-
Missing in 1.5.x:

Description

lighttpd receives SIGPIPE if it tries to output warning about invalid config directive. if lighttpd is started from output capturing program like initlog, then lighttpd receives SIGPIPE, because lighttpd has daemonized() and after that initlog closes connected STDERR and exits itself.

i believe moving daemonize afterwards the log_error_open() could prevent this problem.


stat64("/etc/localtime", {st_mode=S_IFREG|0644, st_size=814, ...}) = 0
write(2, "2005-06-29 15:54:19: (server.c.633) WARNING: unknown config-key: index-file.extensions (ignored) \n", 98) = -1 EPIPE (Broken pipe)
--- SIGPIPE (Broken pipe) @ 0 (0) ---
Process 26854 detached

to reproduce put this into lighttpd/1.3.14 config:


index-file.extensions = (
   "index.html",
)

and start lighttpd from initlog:


strace -s5000 -ff nice -n 0 initlog -c  "lighttpd -f /etc/lighttpd/lighttpd.conf" 

-- Elan Ruusamäe <glen

lighttpd-1.4.15-stderr.patch Magnifier (1.07 KB) spillgroup, 2007-07-11 14:48

History

#1 Updated by spillgroup over 7 years ago

We faced the SIGPIPE problem too and use the attached patch to circumvent the problem. By the time Lighty spawned its child process it shouldn't be writing to stderr anymore, so we call log_error_open() directly AFTER daemonize()'ing.

#2 Updated by glen almost 7 years ago

  • Status changed from New to Fixed
  • Resolution set to fixed

Changeset r2049 commited

Also available in: Atom