How do you try out an application and keep $HOME clean afterwards

Question

As a new Gnome user, naturally I turned to evolution for an email client well integrated with Gnome. As it turned out, I couldn't find any love for it, so I removed the package. But the problem was that it left behind in several places stuff I no longer wanted.
I had prepared for just that. During the try-out of evolution, I had inotify monitor the files it laid its hands on:
$ inotifywait -mr -e CREATE -e MODIFY -e DELETE ~
The list I got from this:
create ~/.pki/nssdb/{cert9.db,key4.db,pkcs11.txu}
create ~/.config/evolution/*
create ~/.local/share/evolution/*
modify ~/.config/dconf/user
modify ~/.local/share/gvfs-metadata/home*
add password to ~/.local/share/keyrings/login.keyring (gnome-keyring)
download mails etc. to ~/.cache/evolution/
Obviously the cleanup was not as easy as simply removing files/folders, I had to:
clean up dconf database:
$ dconf reset -f /org/gnome/evolution/
$ dconf read /org/gnome/desktop/notifications/application-children
['gnome-network-panel', 'nautilus', 'evolution']
$ dconf write /org/gnome/desktop/notifications/application-children "['gnome-network-panel', 'nautilus']"
$ dconf reset -f /org/gnome/desktop/notifications/application/evolution/
clean up login password stored in gnome-keyring: install and run `seahorse', delete the entry "Evolution Data Source XXXXXXXXXX.XXXX.XX@hostname".
clean up data and config:
$ rm -rf ~/.cache/evolution/ ~/.local/share/evolution/ ~/.config/evolution/
I don't know about the NSS stuff, but since no other applications seem to use it (IIRC NSS is used by Chromium, which I don't use), I simply remove it all:
$ rm -rf ~/.pki/
gvfs metadata stuff I really don't know about; ignoring.
As you can see, It takes some long hard work to simply try out an application like evolution while keeping my $HOME intact.
I imagine I could have tried it out under another user (correctly named "tester"), or maybe used some kind of filesystem snapshot (which ext4 lacks afaik), or some kind of sandbox. Too much hassle maybe?
My question to you is how do you do a throw-away trace-free test-run of an application?
Last edited by lolilolicon (2013-07-10 04:48:20)

Answer 1

lolilolicon wrote:@firecat53, I'm not sure I understand (I don't have an idea of the tools you just mentioned), but I feel that virtual machines seem to be overkill/over-complicating? For trying out they may be very suitable, but for daily use they don't seem to be very convenient.
Heh, to me having to mess with dconf databases and moving $HOME seem way more complicated!
So if you just do a quick Arch install into a virtual machine (you can expedite it by mounting a shared directory and sharing/copying your /var/cache/pacman/pkg directory to avoid re-downloading all your installed packages). Configure it as you wish. Now create a shapshot of the virtual machine as it now exists. This is almost a one-click operation in Virtualbox. Now you have a 'saved state'. Boot up the VM, install the package you want to test...change things, break things, whatever. When you're done, shutdown the VM and select the option to 'Discard current state' or some such. Now the VM reverts back exactly to how you set it up originally. Keep it updated just like your regular system (or in whatever state you need for testing) and snapshot it after an update. You can discard old snapshots when you're comfortable with your current update.
The other other tools I mentioned are just tools that can deploy and configure  multiple virtual machines (for example for developers to be able  work collaboratively in identical environments). Probably overkill for what you need.
I think you will spend less time getting a suitable test VM configured then you will trying to worry about filesystem snapshots and monkeying with environment variables. Just my two cents
Scott
Last edited by firecat53 (2013-07-10 20:52:16)