[jasonpgignac]

#!/bin/bash

# Sets remote login (SSH) on or off.

systemsetup -setremotelogin on

[jasonpgignac]

Well, you wouldn’t be able to do this as a true ‘payload free pkg’, since you don’t want to run the script right away – you want it to run on first boot. So, the easiest thing to do, at least in my experience, is to just have a pkg, that copies the script to the system, and then copies in a launchd job onto the system. Then, make sure the script logs itself, and that if there is no errors, that it deletes itself, and the launchd job at the end.

[jasonpgignac]

Just as an added note – we actually keep a hash of system types versus model name – this gives us some flexibility if we, for instance, want to have some scripts only apply to XServes, some to Laptops, etc. Otherwise, teh methodology is pretty much the same, except instead of just grepping for ‘Book’ we strip out the model, and look it up in a hash. YMMV

[jasonpgignac]

Just as an interesting side note to this –

At my company (and perhaps others would have input on whether they consider this a best practice?) I try to keep the registration key file(s) in a seperate package. This gives two benefits:

1) If I have a software with many individual licenses, I can keep a folder full of license pkgs, and just push the one needed for a particular client
2) If the key changes (for instance, we just went through a merger, and now ahve to use the new companies license keys for Adobe and Microsoft), then there is no need to rebuild the outside Pkg.

[jasonpgignac]

I can certainly see the point of having a guaranteed sort order, but this seems more like a ‘best practice’ than a programmatic limitation. I can more see your point about avoiding extraneous files (the examples you mentioned could be easily overcome by, say, testing to see if the given items are folders, but I can see the case in which an errant folder could end up there as well.

The XX-String model is certainly a good option. I wonder if adding a file extension to the folder name, or something might be intuitive? I’d say, in fact, just making bundle file, which could offer other advantages as well, but this might be a bit of impediment for the new user, I suppose. But, I wonder if maybe just having the folder named, say “01-Fake Update.instadmg” would do a good job of offering maximum flexibility in naming, while simultaneously making it clear what the folder is, semantically?

It also, I think, might be a good planning mechanism for a future state, where we want to develop a proprietary format? Imagine, for instnac,e a future condition in which apple releases two versions of it’s updates, a .pkg file, and a .instadmg file that’s pre-edited to work optimally with the instadmg process (I’m thinking, here, of all the packages I’ve hade to edit ever-so-slightly to make them work with instadmg). As instadmg is developing into a bonafide standard, I’m just thinking we may as well leave ourselves room to develop future standards?

Just my two scents, quite possibly I’m dead wrong

[jasonpgignac]

I’m not an awk programmer, but I think the problem is in line 667:

/bin/ls -A1 “$SELECTED_FOLDER” | /usr/bin/awk “/^[[:digit:]]+$/” | while read ORDERED_FOLDER

This is the line that lists out a given folder (in this case, the base updates folder, or the custom pkg folder), and feeds into, basically, a foreach loop. The voodoo that I’ll admit to not fully understanding, but that looks look the problem, is the awk in the middle. As an example, I made a folder, awktest, with some test folders inside of it. Look at this output test:


SAOTXML-5281YK0:awktest ugignja$ /bin/ls -A1 ./
01
02
03-Hello
04-Update 3


So you can see the folder structure I created. Now, run through this awk applet:


SAOTXML-5281YK0:awktest ugignja$ /bin/ls -A1 ./ | /usr/bin/awk "/^[[:digit:]]+$/"
01
02
SAOTXML-5281YK0:awktest ugignja$


So, basically, that awk section’s entire purpose is to winnow out any folders that contain characters that are not digits, it seems to me. *IF* you left spaces out your folder names, to relieve any escape issues (I don’t know if there is any, I’m just being safe, here), then I THINK (though I haven’t tested it, this is just from reading source code) you could strip out the awk part of the command on that line, and be able to use any folder name you like.

Developers, perhaps you could offer some insight into the purpose of restricting folder names? I would think it would be simpler to just have documentation say that folders need to be set up to sort in the order they need to be installed, and leave naming conventions to the technician? Just a thought, like I said, possibly I’m missing something?

[jasonpgignac]

My bad, sorry to get everyone excited – I should have known better than to believe Adobe was doing something the standard way 😉

[jasonpgignac]

Actually, and this is second hand info, but I’ve been told that CS4 will have a repackager downloadable from Adobe, that will allow you to wrap the installers in a PKG. I can’t confirm that, though.

[jasonpgignac]

All that the package does is create a user – if you went through the wizard, and got in the OS, you’d (presumably, if you have done it irhgt) see the user already exists. There is another package – clearReg I think it’s called – that removes the registration wizard for you. You need to install it. It’s available in the file section, here

[jasonpgignac]

This would, of course, depend entirely on what kind of wireless you are setting up. If ther’s 802.1x involved, for instance, it probably wouldn’t work, because the cert would be computer specific, right?

[jasonpgignac]

In the meantime, it probably shouldn’t be too hard to hack out their post-install script?

[jasonpgignac]

My pleasure – I imagine you could shell script circles around me 😀

[jasonpgignac]

Great product! I revised it slightly, by just gathering it into a single App bundle, so you don’t have to make sure the shell script is in the same folder, if that makes it easier for anyone – I know I can barely keep track of one thing, two is beyond my ability ;). Here, if anyone wants it.

[jasonpgignac]

Just to check that I understand – you’re writing a launch daemon or something that starts softwareupdate, as root, right? Are you waiting long enough on first boot to let the network card initialize? Or putting in a test to see if there is a net address yet?

[jasonpgignac]

It looks your log was for an install to the root volume. Could be your package only works on the system volume? I’d try mounting a system in Target mode, and then installing your package, see if it works then? Coul dbe that your install is working flawlessly but that it ain’t installing into InstaDMG – rather you keep installing into your host system.

[Author]

Posts