stage-1: fix typo that breaks resume

[spencerjanssen]

$d should be $sd, this causes resume from hibernate to fail if
resumeDevice is not explicitly set in config. Introduced in commit:
'stage-1: Shut up warnings about swap devices that don't exist yet'

Resume stopped working after a recent update, and this commit fixes it for me. I tested this against current nixos-unstable, but I have not tested the change after rebasing on master.

[vcunat]

Ah, this is why resume didn't work for me on the systemd-219 branch. Picked it back 026c31d.

[copumpkin]

What would a VM test for this look like? Do we have some way to simulate a resume in qemu?

[vcunat]

It would be nice to have a typo-resilient language for such "scripts" (i. e. with static type checking).

[lucabrunox]

@vcunat yeah except it doesn't exist... e.g. you might still write getenv("fooz") instead of getenv("foo"), type checking won't save you.

[vcunat]

The need for outside-environment variables is rather rare – I see mostly need for local bindings (such as this mistyped iterator). The truth is I don't know about a suitable language+library for these kind of things.

[lucabrunox]

@vcunat so you have two ints, a and b, and you mistype one for the other :P

[copumpkin]

The fact that types don't catch everything isn't an argument against using
them. Further, most decent type systems allow you to construct more
interesting types than Int that carry more significant semantics, so that
you don't have to go using unadorned machine representations as parameters
to your functions, but can actually pass in actually meaningful values.

On Fri, Jul 31, 2015 at 10:03 AM, lethalman [email protected]
wrote:

@vcunat https://github.com/vcunat so you have two ints, a and b, and
you mistype one for the other.

—
Reply to this email directly or view it on GitHub
#9056 (comment).

[edolstra]

@vcunat Well, we can rewrite the stage-1 script in Perl :-)

[vcunat]

@lethalman: mistyping into something that is defined is much less likely, especially when one minimizes scope sizes, and adding a strong type-checking would decrease the probability even more.

Perl doesn't sound as a significant enough improvement in terms of finding errors before running, but maybe it's just because I know too little about it (or about a suitable lint-like tool for it).

[edolstra]

Perl (unlike Python) does actually complain about undefined variables, so it would have caught this error. Regarding better type checking, there are probably not that many options, given that the language interpreter has to fit inside an initrd... (Unless of course we'd want to go for a compiled language.)

[copumpkin]

For now I'd just settle for a test 😄

[vcunat]

(I'm sorry most of this thread is off topic.) I think we could have compilation for such things, as we don't allow in-place modification anyway. To stay conservative, we might still leave the source along each executable. But even then I don't see a clear choice of language.

[copumpkin]

To stay conservative, we might still leave the source along each executable.

cough NixOS/nix#553