To be honest, if the user has a non-existent or otherwise invalid $HOME, they have bigger problems, and a lot more than the Nix installer is going to fail. We don't have to do 5 different checks to validate $HOME.

I've made a rewrite of script at this stage. But it is in local git, I hit a PR buffer.

I look at it like this:
To make it fully robust and portable.
That on any environment it going to work.

Users on some systems can not have HOME, plethora of users can have single HOME, there are a lot of occasions.

If that are required conditions, they need to be checked, and installer needs to fallback/rollback. It is how I see it.

Better to check for default, expected conditions, that fit 95%, and then get reports that 5% had errors, and work with them, to make it cover ->100%.
Then to assume 95% is all there is, and get 5% of unknown side-effects, and that 5% going to be frustrated by first Nix acquaintance. And big number from that 5% probably was some nonstandard specialists that could be important for project.

This is simple checks that does need no maintenance, and I am ready to pour time and wrote them.

I aim to make installer work in unknown Linux conditions. So it will work on any distribution/environment. And also work on already running servers. This is a bar I set for myself.

One time I done secure system that created a lot of users, for every client of the company.
And their home was: /scp folder with additional jailing for users.
/scp was HOME to big number of users.

@ edolstra
Sorry, I am tired. I can sound not appreciative.
I just try to make installer an ask you to say, is root install is possible.

It should ease the migration of available admins/systems to Nix.

Thank you for everything.

Please, let me unveil some shell-fu on Linux installation.

I am not making 4 000+ shell script here. Thou I've could make much more portable, advanced and safe install, using shellfire shell framework. But it going to have huge entrance level to modify it for anyone else.
And even 4 000+ shell script is not any performance hick-up for today systems, words from shellfire developer itself.

But the aim here is that - sh - already installed on any even the most bare system, and compatible with itself (almost).
sh is everywhere, even on Windows now, but I not try to make it for Windows now.

So it is bad, to do requirements checks and report to the user the status before doing work on his setup?

I am trying to do unattended Linux install on any distribution and setup. The more portable, polished and user friendly it will be - the bigger fame, community and project you will get.

Since Nix is perfectly fitting to do portable script (it installs separate infrastructure in /nix and HOME/.nix-profile). It is very real goal.

I just see, that you developed initial version of the script, and since than, - contribution to it was sporadic. And there is a lot can be done in regard to improve user experience and automation of install.

I am trying to send a little by little pull requests to get chain of process going.

@Anton-Latukha actually, the only thing that matters is that you can write the requisite files into $HOME. It doesn't matter where $HOME is or who owns it, just that you can write, and only when you need to do the write.

So I think the only useful test is to check if $HOME is defined, and then let errors during installation just happen.

@wmertens

1. Thank you on input.

2. Doesn't HOME must have also 'x', to be able to open it:

$ mkdir /tmp/test_folder
$ chmod 0200 /tmp/test_folder
$ touch /tmp/test_folder/file
touch: cannot touch '/tmp/test_folder/file': Permission denied

$ chmod 0300 /tmp/test_folder
$ echo "Hello, I'm here!" >> /tmp/test_folder/file
$ cat /tmp/test_folder/file
Hello, I'm here!

3. I though to do all requirement checks at start.
   And if they pass - do changes on the system.
   It is by design more clean installer this way.

If we allow errors just happen as script go now. Someone adds something before the line that traps error. And now we have uncontrolled side-effect.

As I started in "install-nix: Full progress of Linux installation #1565". All checks go before doing any actions.

4. That is why also I do additional checks. Does HOME is folder or link to folder, because it can be file.

5. And check if HOME owned by user. Because it can be a community folder with wide access rights.

And when people would come with bugreport of esoteric case, when HOME is not owned by user, - it can be modified to also permit that legitimate but esoteric use.

I can agree that last point is debatable. But, check, does HOME is folder or soft-link to folder, and folder also needing 'x' to write anything to folder - I'm solid on that.

Ok.

I close this.