<Tv`>i think i'm missing something from my understanding; i would have expected guix pull and whatever that other command was that i've already forgotten to have brought my /gnu/store to contain "the latest stuff" already
<Tv`>and if the config.scm is the same, it shouldn't ask for more
<mange>"guix pull" is like "apt update". It doesn't actually update your system/profile, it just updates the package definitions. Then you have to "guix system reconfigure" or "guix package -u" to upgrade your system/profile respectively.
<Tv`>guix package -u is the one i meant and forgot
<Tv`>couldn't have done reconfigure before, because i didn't have a config file to use
<mange>Then "guix package -u" only updates your user profile, not your system. If you're doing "guix system container" then it's likely that doesn't share many packages in common with your profile.
<fps>i tried to guix pull occasionally over the last weeks and it always failed..
<mange>fps: That's... strange. I wouldn't expect guix pull to need python. Can you confirm that the exact command you are running is "guix pull"? Can you also tell me what version of guix you're running by running "guix describe"?
<mange>It wouldn't be too old, just misconfigured. I would still hope that it would be able to build everything it needs locally, but it probably does make sense to get it to fetch substitutes. Check out "(guix) Substitutes" in the manual.
<fps>mange: this is a guixsd system where i never touched the configuration manually. i will check the manual though for the new susbstitute url
<Tv`>mange: it seems like the vm image had ci.guix.info preauthorized, as far as i can tell
<fps>mange: after authorizing it it seems to use ci.guix.info without explicit --substitute-urls..
<fps>mange: guix pull worked now.. i still would have expected it to work without that though..
<Tv`>mine still wants to build a bunch, including the qemu that fails.. it's downloading some but apparently can't find substitutes for all
<Tv`>that guix pull is currently running, we'll see what happens after it's done (slow internet here)
<fps>btw: does every terminal in guixsd totally break on reverse history searches for anyone else. too?
<fps>[break as in similar to when you setup your $PS1 wronly and redraws are all over the place]
<mange>fps: I would have expected guix pull to work without substitutes, too. Substitutes just helped us to work around the fact that it wasn't building on your machine. I don't know why it wasn't building on your machine.
<fps>oh and i think it happens *after* running some guix commands. my hinch would be the progress bar implementation screwing up the terminal
<mange>Tv`: It can take the build machines a while to catch up when packages definitions are changed. Sometimes when you "guix pull" then try to reconfigure it will mean that you end up building some stuff yourself. The easiest solution is to just wait. Alternatively, you can pick an older commit and "guix pull --commit=" to it (in the hopes that the build machines have already built those packages).
<fps>hmm, interesting. during "guix package -u" bash segfaulted
<fps>and during "sudo -E guix system reconfigure /etc/config.scm" make segfaulted..
<rekado>there’s nothing special happening when using Libreboot. You just let the on-chip GRUB decrypt the disk and pass execution on to the on-disk GRUB. You still need to install GRUB.
<raghavgururajan>How?? For full disk encryption, there is only one partition for me. If I don't pass --no-bootloader, I am getting error that no efi partition found at the end of guix system init command.
<rekado>you probably don’t want to install in EFI mode then.
<rekado>what does your configuration file’s bootloader field look like?
<lfam>"Traditional" distros like Debian and Red Hat update the system by changing the installed software "in place", where the software is deployed. If the process is interrupted for some reason, the system is left in an inconsistent state that may or may not work. The package management tools on those distros cannot recover from such a state and the systems are basically broken at that point
<Elon_Satoshi>Ah. But Guix installs software somewhere else, in its own environment, and then simply changes the symlinks
<lfam>Guix (and Nix) prepare the updated software in another part of the system. They create a set of symlinks to the updated versions in what you could consider a "staging" area. In the final step, a single symlink is updated, and this single symlink is what makes the updated system available for use. We call these sets of symlinks "profiles"
<lfam>By reducing the change to a single step, it's possible to reason about the state of the system in a transactional way. Either it was updated, or not. There is no possibility of an in-between state
<lfam>And then that one symlink's history can be tracked, and this gives us the ability to keep a history of previous states and change between them freely
<Elon_Satoshi>When I hear the word transaction, I think money. What does transaction mean in this context?
<lfam>It means the change was either performed or not. Similar to how your bank would like to know whether or not the money is in your account, and not allow the money to exist in two different accounts at once
<lfam>Anyways, the point for Guix and Nix is that we aim to eliminate a common failure mode of older GNU / Linux distros
<mikadozero>I am getting an error when building qemu-minimal-2-10-2 during the check phase. The log the error mentions for tests/qom-test: FAIL can not allocate memory broken pipe. The system has 1GB of memory.