- launchd-dev - lists.macosforge.org

Starting Apache: delay to when users/groups are fully loaded?
by Cabel Sasser 03 Dec '12

03 Dec '12

Hey Guys, We're launching our own httpd via a launchd file on startup. However, here's the result: Nov 30 17:26:54 posh com.panic.httpd[120]: httpd: bad group name web If we "launchctl start" this same file a few seconds later, it works fine! So it appears that httpd is trying to run before Open Directory or something has fully loaded all of the user/group data…? What's the best course of action for us? Is there a way to delay a launchd item, or wait, or anything? We're tired of manually starting apache every time the server reboots. :) (Attached find the .plist file in question.) Thanks, Cabel --- <?xml version="1.0" encoding="UTF-8"?> <!DOCTYPE plist PUBLIC "-//Apple Computer//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd"> <plist version="1.0"> <dict> <key>Disabled</key> <false/> <key>Label</key> <string>com.panic.httpd</string> <key>ProgramArguments</key> <array> <string>/usr/local/apache2/bin/httpd</string> <string>-D</string> <string>FOREGROUND</string> </array> <key>RunAtLoad</key> <true/> <key>KeepAlive</key> <false/> </dict> </plist>

3 2

launchd StartInterval and sleep
by Todd Heberlein 29 Oct '12

29 Oct '12

What is the expected behavior when using StartInterval to start a script every n seconds and then the machine goes to sleep? The reason I ask is that I had a StartInterval of 4 hours (14400 seconds), put my laptop to sleep overnight, and when I opened it, I expected the script to automatically execute because the interval expired during the sleep. But this didn't happen. Now experimenting with just a 5 minute interval, it seems that if the interval would be reached when the machine is asleep (1) the script doesn't fire when the machine wakes back up, and (2) the start interval starts again using the new wakeup time. For example, the next time the script fires is 5 minutes after wakeup. Is this the way it is supposed to work, or am I doing something wrong? Thanks, Todd

3 3

Job is terminated before the daemon
by Kentzo 19 Oct '12

19 Oct '12

I have a launchd on-demand job that is triggered on network access. The plist looks like this: <dict> <key>Label</key> <string>com.kulakov.vm.development</string> <key>ProgramArguments</key> <array>_omitted_</array> <key>inetdCompatibility</key> <dict> <key>Wait</key> <false/> </dict> <key>Sockets</key> <dict> <key>NetworkListener</key> <dict> <key>SockServiceName</key> <string>5906</string> <key>SockType</key> <string>stream</string> </dict> </dict> </dict> </plist> The program is essentially a bash script that looks like this: #!/bin/bash nc -w 300 some_ip 5907 At this point, I expect that when a packet comes to the 5906, tcp proxy using nc will be started and the job keep running until nc returns. Unfortunately the job is killed before nc returns. When new connection is made to 5906, the script is restarted. Is it intended to work like this? -- View this message in context: http://mac-os-forge.2317878.n4.nabble.com/Job-is-terminated-before-the-daem… Sent from the Launchd mailing list archive at Nabble.com.

1 1

Re: [launchd-dev] Facing problem with launch daemon on logout
by Quinn "The Eskimo!" 01 Jun '12

01 Jun '12

On 1 Jun 2012, at 09:15, Nandini Vijayashankar wrote: > Iam not able to get the output of the mentioned command > I get the error as "Invalid argument"(when /System/Library/LaunchDaemon is given) and "not an object file"(when /System/Library/LaunchDaemon/<plistfile> is given) Earlier you wrote: > I have a plist file in /System/Library/LaunchDaemon.This plist file invokes a shell script which starts my daemon process You have to point otool at the path to your daemon's executable, not at its launchd plist or the launchd directory. So, what executable does your shell script launch? And what does otool say if you run it against that? Also, as a third party vendor if you shouldn't be installing stuff in "/System/Library/LaunchDaemons". Rather, you should use "/Library/LaunchDaemons". We have a habit of blowing away all the stuff in the /System hierarchy when we update the OS, so if you want your daemon to persist across upgrades you need to put it in the /Library hierarchy. Share and Enjoy -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Relations, Developer Technical Support, Core OS/Hardware

1 0

SMJobBless returning error 4098
by Eric Gorr 30 May '12

30 May '12

I am trying to install a secure helper tool with SMJobBless. When it fails and before calling SMJobBless, I am calling SMJobRemove because I need to remove an older version of the tool - the removal succeeds. SMJobBless is returning an error code of 4098. The NSError object is only telling me that "The operation couldn't be completed. There was an error in the CodeSigning subsystem." If I rerun my code, the SMJobBless function works. I would assume this is because it was removed previously, but why would it have not worked the first time? I can then communicate with the tool and everything functions normally. Observing that everything is functioning normally, I believe I can be certain that I am meeting the five requirements for SMJobBless as described in the documentation. If I increment the version of my tool and try again, SMJobRemove will work, but, again, SMJobBless will return the error code 4098. If it matters, I am using OS X 10.7.3.

1 0

su/sudo behave differently for controlling a user’s launchd
by Sidney San Martín 28 May '12

28 May '12

I have an installer that loads an agent into launchd. A whiiiiile back, I read Quinn’s post about Mach bootstrap namespaces and launchd: http://lists.macosforge.org/pipermail/launchd-dev/2010-February/000741.html (In my case, I decided that loading it just for the current user was probably good enough.) Well, I just did an overhaul on the installer and noticed that loading the agent on 10.5 was failing with an error: launch_msg(): Socket is not connected Huh?! All seemed to be well in 10.6 and up. After bashing my head against it for a while, I discovered that a line like this in my installer: su "$USER" -c "launchctl load …" had, in the overhaul, changed to this: sudo -u "USER" launchctl load … Using su fixes the issue. I have no idea what the difference might be. I’m posting this in case anyone else runs into the same problem. Also, I wonder, does anyone has any insight into why they behave differently? Sidney

2 1

StartInterval's remaining time
by Jorge 22 May '12

22 May '12

For jobs scheduled to run N seconds via the StartInterval key, is there a way to know how many seconds remain until the next scheduled run?

1 0

Re: [launchd-dev] Facing problem with launch daemon on logout
by Quinn "The Eskimo!" 09 May '12

09 May '12

On 8 May 2012, at 08:07, Nandini Vijayashankar wrote: > If i restart the computer and login, my daemon is running. > After this if i logout and login again, daemon is exited with error code 143 > > If i explicitly load the job, logout and login, daemon is running. > > Problem is when the system is restarted and I logout and login. It's likely that your daemon is connecting to the window server, and thus killed when the window server dies on logout. Do the following: $ otool -L /path/to/your/daemon and check that all the frameworks you use are daemon safe [1]. Also, when you explicitly load the job, make sure you use "sudo". Without that your daemon ends up loading in the per-user context, which generally isn't what you want. Share and Enjoy -- Quinn "The Eskimo!" <http://www.apple.com/developer/> Apple Developer Relations, Developer Technical Support, Core OS/Hardware [1] Per the definition in Technote 2083 "Daemons and Agents". <https://developer.apple.com/library/mac/#technotes/tn2083/_index.html>

1 0

'launchctl remove' synchronicity
by Thomas Clement 07 May '12

07 May '12

Hello list, It seems "launchctl remove" and "launchctl unload" do not behave in the same way when unregistering a job from launchd. I can reload a job by doing "launchctl unload" / "launchctl load" and it works fine. However doing "launchctl remove" / "SMJobSubmit" makes launchd complain about "Conflict with job: job_label over Mach service: service_name". It seems that when SMJobSubmit is called and tries to load the job, the Mach services of the previous instance of the job are still registered into launchd. Adding a delay between "launchctl remove" and "SMJobSubmit" seems to solve the issue. Is this to be expected? Thomas

3 2

Facing problem with launch daemon on logout
by Nandini Vijayashankar 07 May '12

07 May '12

Hi, I have created a launch daemon process. I can start the process using: launchctl load <plistfile> But when I log out, this daemon is getting killed and I see the log message in the console as: com.apple.launchd : (<process name>) Exited with code : 143 When we logout, only launch agent process should be stopped. But why is the daemon process exiting or getting stopped? This issue iam facing only with Mac 10.7 (Lion). In Mac 10.5 and 10.6 daemon process keeps running even after logout. Any help for this is highly appreciated. Thanks

2 3