Closed kirillrdy closed 2 months ago
@kirillrdy I don't really care about automated formatting, but if you prefer a certain kind of formatting when you make the PR, just make it a separate commit and I'll accept it.
How do you think they get the latest URL? I want to automate this.
i'll ignore auto-formatting for now.
re URLs, if you have uptodate macos ? https://gist.github.com/b0gdanw/26f528dda5abea7ce6fe633ca2ef1063
Just to be clear, by automated formatting, I mean enforcement at a CI/commit-hook level.
Maybe some of these InstallAssistant packages are available on archive.org as a backup, or may eventually be. We could use those if they are.
Maybe some of these InstallAssistant packages are available on archive.org as a backup, or may eventually be. We could use those if they are.
an other ( more painful alternative ) is to boot into RecoveryImage and download latest installer using their GUI :-)
The CI machine is having hardware issues, and I'm not at home to fix it. So I'm having a bit of trouble verifying this the way I'd like to. Is it possible you could verify it and tell me if it works as expected up to 5 times with nix build --rebuild
? @kirillrdy
The CI machine is having hardware issues, and I'm not at home to fix it. So I'm having a bit of trouble verifying this the way I'd like to. Is it possible you could verify it and tell me if it works as expected up to 5 times with
nix build --rebuild
? @kirillrdy
I've built this at least twice on two different machines, but will do more and will report
ok i've done 5 rebuilds
[kirillvr@tsutenkaku:~/NixThePlanet]$ nix build .#macos-ventura-image --rebuild
error: derivation '/nix/store/cy0cmdbkw3zdki5ksdlrjx8jfl3hk61i-mac_hdd_ng.qcow2.drv' may not be deterministic: output '/nix/store/0z2pwrvpxi3j8gvscfdnmmkbi59bxdf4-mac_hdd_ng.qcow2' differs
I assume old image was also non deterministic.
no urgency to merge this, take your time
Yeah derivations aren't deterministic, and don't need to be. One of Nix's strengths is allowing you to model multiple kinds of reproducibility, and cache non-deterministic behavior, whilst also having the option of using content-addressed models as well. In this case all that matters is whether the derivations succeed and produce $out or not, and if that worked for you then I'm happy to merge it here.
CI is now running, and it's succeeding on the repeatability test now https://hercules-ci.com/github/MatthewCroughan/NixThePlanet/jobs/477/effects/effects.macos-repeatability-test/log
I managed to fix the CI machine remotely by coercing it into using
boot = {
tmp = {
useTmpfs = true;
tmpfsSize = "90%";
};
}
Whenever the virtual machine was too slow at writing/reading I would get
mac_hdd_ng.qcow2> Installation cannot proceed
mac_hdd_ng.qcow2> because the installer is
mac_hdd_ng.qcow2> damaged.
mac_hdd_ng.qcow2> Redownload the installer and try again.
Using the 64GB of memory via tmpfs allows the CI to run smoothly, whereas the disk(s) which are on multi-device bcachefs lead to the above error.
first link from https://mrmacintosh.com/macos-ventura-13-full-installer-database-download-directly-from-apple/
@MatthewCroughan what do you think about nixfmt-rfc-style ?