daemonize Change Log
Changes in version 1.7.8:
- Fixed various compiler and cross-linking issues. Thanks to @derekstraka for the patch.
Changes in version 1.7.7:
- CFLAGS and LDFLAGS not passed through from
Makefile. Fixes GitHub Issue #20 https://github.com/bmc/daemonize/issues/20. Thanks to Andrew Stormont (andyjstormont /at/ gmail.com) for the patch.
Changes in version 1.7.6:
- Fixed potential memory allocation issues. See GitHub Issue #15 https://github.com/bmc/daemonize/issues/15 and pull request #16 https://github.com/bmc/daemonize/pull/16. Thanks to Frederic Descamps (frederic.descamps /at/ percona.com) for the fix.
Changes in version 1.7.5:
- Added support for out-of-tree builds, per Issue #12. Thanks to @andy-js
Changes in version 1.7.4:
acconfig.hpatch from Andreas Müller (schnitzeltony at googlemail.com) that cleans up
- Moved version stamp into a header.
Changes in version 1.7.3:
- Merged patch from Ewan Mellor (ewan.mellor at citrix.com) that adds an
appropriate definition for
Changes in version 1.7.2:
- Fixed bad reference to
daemon.c, causing compilation problems (Should have been
Changes in version 1.7.1:
- Added emulator for flock(2) file locking call, emulated in terms of lockf(3), for systems (like Solaris) that don't have flock. Addresses Issue #3. Tested on Solaris 11 (x86).
Changes in version 1.7.0:
- Now supports a
-Eparameter, which allows setting one or more environment variables, to be passed to the daemon. Based on a suggestion from Michel Gouget michel /at/ gouget.org.
Changes in version 1.6.2:
- Local copy of daemon(3) now errors out (with a message to syslog(3)) if it cannot open /dev/null. Thanks to Folkert van Heusden folkert.mobiel /at/ gmail.com for noticing the problem.
Changes in version 1.6.1:
- Makefile now generates daemonize.html, an HTML-ized version of the man page.
- Fixed usage message, which incorrectly said "stderr", instead of "stdout" in the description of the "-o" option. Thanks to Sam Pierson for the patch.
Changes in version 1.6:
Rolled in changes submitted by Ken D. Farnen, (kenf /at/ 14Cubed.COM):
On slackware, all of the rc.???? scripts run directly from init, and init doesn't set up much in the way of an environment, whereas login stuffs loads of stuff into the environment. Daemonize now ensures that a minimum environment is passed to the exec-ed daemon (IFS and PATH).
The -u flag now puts USER, HOME and LOGNAME into the environment, for the benefit of programs that rely on them. The lack of these environment variables seems to have deleterious effects on MythTV, for example.
On Slackware Linux, access to video and audio device entries is controlled by secondary group membership. (e.g., A user who needs access to a dvb device needs to be in group "video", and a user who needs access to dsp devices needs to be in group "audio".), This is typically done by the login code. Daemonize now initializes those group memberships, as well, if the initgroups() function is present.
Changes in version 1.5.6:
- Accepted and applied a patch from Gary T. Giesen (giesen /at/ snickers.org) that allows the daemonize Makefile to play well with Linux RPMs.
Changes in version 1.5.5:
- Changes in 1.5.4 made it impossible to create PID files in directories not owned by the user specified with -u. Prior to 1.5.4, doing so was possible. It is now possible again, and the functionality has been corrected slightly to ensure that the proper PID is written and that the PID file has mode 0644.
Changes in version 1.5.4:
If a user name is specified via the -u option, the output files specified by the -o and -e options are now opened after daemonize switches users, not before (as happened in previous versions). This change ensures that the output files end up being owned by that user, not by root. This change has no effect if -u is not used.
Rearranged and simplified some of the code in daemonize.c
Added some documentation to the code.
Changes in version 1.5.3:
- Minor licensing change. (configure.in still emitted a message indicated that daemonize is GPL-licensed. It's BSD-licensed.)
- Updated copyright information.
Changes in version 1.5.2:
Two changes suggested and supplied by Yakov Lerner <iler.ml gmail.com>:
Added a "-l lockfile" option, permitting daemonize to use a lockfile to ensure that only one instance of the daemon runs at a time.
Now makes additional checks to ensure that the program specified to daemonize is a regular file, not a directory.
Changes in version 1.5.1:
- Now released under BSD-style license, not GNU Public License.
Changes in version 1.5:
daemonize can now optionally redirect the daemon's standard output and standard error to files
configure now supports a local basename() emulator.
Changes in version 1.4:
Version of daemon(3) shipped with daemonize attempted to redirect stdin, stdout and stderr to /dev/null, by closing all file descriptors (using sysconf(_SC_OPEN_MAX) to determine the maximum file descriptor). This method wasn't working, and apparently the forked daemon process was reacquiring (or still had) valid stdin, stdout and stderr file descriptors. I made some changes to daemon.c to be more consistent with how the native FreeBSD and Linux versions behave.
This bug may have contributed to a problem on AIX. According to Ron Avriel <ron /at/ cti2 /dot/ com>, when he telnets to an AIX 5.2 system and uses daemonize create a daemon, he can't log out of the telnet session (i.e., the logout operation hangs) until he kills the daemon. That observation is consistent with the theory that the daemon process is still holding open the controlling terminal.
daemonize now installs with 0755 permissions, instead of 0700.
Changes in version 1.3:
- The "-c" option now actually works. Duh.
Changes in version 1.2:
- The external names in getopt.c are now prefixed with "x_" (e.g., x_getopt(), x_opterr, etc.). The function's name has been changed to avoid type conflicts with the native getopt() on the host operating system. (This is known to happen on Solaris 8, for instance.) Thanks to Michael Fuchs <mf /at/ mfox /dot/ de> for pointing out the problem on Solaris.
Changes in version 1.1:
Now uses local getopt.c unconditionally, instead of defaulting to whatever's on the system. This gets around a problem when installing on systems with GNU getopt (e.g., Linux). Unless POSIXLY_CORRECT is set in the environment, GNU getopt scans the entire command line, which means it'll complain about "unknown options" that are specified in the command being passed to daemonize. One solution is to have daemonize set POSIXLY_CORRECT right before calling getopt() (and then restore it before running the program-to-daemonize, in case it might affect that program's behavior). However, it's just simpler to use the old, public domain getopt routine.
daemonize now verifies that the program actually exists before daemonizing. (It didn't check before.)