| 8 Oct 2024 |
 jade_ | In reply to @benjamin:computer.surgery
(unless the host is also a mac, and then you're back where you started) kinda. at least VM hosts only gather physical dust in general | 23:25:10 |
| jade_ | no virtual dust | 23:25:13 |
 just1602 | So we need some folks to throw money at lix.so one of the core maintainer can convert one of their closet into a apple server rack ? | 23:26:03 |
 KFears (tragedy arc) | In reply to @jade_:matrix.org
like, i think it probably is if you do it on apple hardware, and i don't advocate doing crimes, but i also don't like not having CI I think it's legal to run MacOS VMs on MacOS OS and hardware, and I think mirroring code onto Github and using its CI isn't a breach of TOS | 23:26:06 |
| jade_ | https://khronokernel.com/macos/2023/08/08/AS-VM.html oh. there is a 2 VM limit too. | 23:26:08 |
| KFears (tragedy arc) | But Apple situation is difficult for sure | 23:26:14 |
| jade_ | In reply to @kfears:matrix.org
I think it's legal to run MacOS VMs on MacOS OS and hardware, and I think mirroring code onto Github and using its CI isn't a breach of TOS the latter is iirc a breach of their TOS, but i am not 100% certain as to whether mirroring it is sufficient to be considered to be using github | 23:26:39 |
| KFears (tragedy arc) | Any chance we are gonna have levels of support where MacOS is not tier 1 for sure? | 23:26:40 |
 benjamin | In reply to@jade_:matrix.org
kinda. at least VM hosts only gather physical dust in general true :) being able to restore the VM to a snapshot every time the notarization keys get screwed up would actually be a huge improvement | 23:26:57 |
| KFears (tragedy arc) | In reply to @jade_:matrix.org
the latter is iirc a breach of their TOS, but i am not 100% certain as to whether mirroring it is sufficient to be considered to be using github I guess one could contact Github as an org official to clarify that point? | 23:27:10 |
| jade_ | i also just hate using gha so much | 23:27:31 |
| KFears (tragedy arc) | Who doesn't? | 23:27:39 |
| KFears (tragedy arc) | MacOS CI is so ridiculously expensive, though, it might be cheaper by a few factors to run GPU CI | 23:28:05 |
| just1602 | In reply to @kfears:matrix.org
Who doesn't? People who were forced to use Jenkins there whole career? | 23:28:26 |
| jade_ | In reply to @kfears:matrix.org
MacOS CI is so ridiculously expensive, though, it might be cheaper by a few factors to run GPU CI we have macOS CI but it's just useless for the installer because it can't actually run the thing to test it, which kind of defeats the purpose | 23:28:39 |
| jade_ | thus, yeah. if you operationalize OSX-KVM or another solution to do easy snapshotting of macOS VMs, that would be an awesomely useful dev tool for the installer and would make your job way easier | 23:31:08 |
| KFears (tragedy arc) | In reply to @jade_:matrix.org
we have macOS CI but it's just useless for the installer because it can't actually run the thing to test it, which kind of defeats the purpose
- How impossible is it to run MacOS VM tests for installer?
- How impossible is it to run MacOS tests in something like a FHS chroot?
| 23:31:19 |
| jade_ |
- well the issue you have is that there isn't like, a vagrant image for macOS or something, and that macOS has a bit of weirdness with needing a user logged in at the console to have certain functionality (possibly required for installer? unsure!) work properly even over ssh. the troubles you would have to work through are chiefly that the image is huge and that creating the image in the first place is manual, but might be able to be automated with a virtual keyboard and maybe some OCR.
- do you mean like, a chroot on macOS? chroot exists on macOS but it is missing a bunch of critical functionality and is banned by SIP. we don't care about SIP for CI because we can turn it off. but you still won't have bind mounts and enough of our installer is doing weird things to the base system that i don't think that chroot would be sufficiently representative compared to booting a throwaway VM snapshot.
booting throwaway VM snapshots is also not that slow, especially if zfs is implicated.
| 23:34:39 |
| jade_ | *
- well the issue you have is that there isn't like, a vagrant image for macOS or something, and that macOS has a bit of weirdness with needing a user logged in at the console to have certain functionality (possibly required for installer? unsure!) work properly even over ssh. the troubles you would have to work through are chiefly that the image is huge and that creating the image in the first place is manual, but might be able to be automated with a virtual keyboard and maybe some OCR.
- do you mean like, a chroot on macOS? chroot exists on macOS but it is missing a bunch of critical functionality and is banned by SIP. we don't care about SIP for CI because we can turn it off. but you still won't have bind mounts and enough of our installer is doing weird things to the base system that i don't think that chroot would be sufficiently representative compared to booting a throwaway VM snapshot.
booting throwaway VM snapshots is also not that slow, especially if zfs or anything else copy on write is implicated.
| 23:34:51 |
| jade_ | In reply to @just1602:systemli.org
So we need some folks to throw money at lix.so one of the core maintainer can convert one of their closet into a apple server rack ? we have apple CI today that is an old m1 mac in a lix developer's closet :P | 23:36:31 |
| jade_ | my point here is that, if we had the software to run the CI, the hardware could be found | 23:38:50 |
| jade_ | hardware is merely a money problem and we can find money for hardware | 23:39:08 |
| jade_ | or just find old hardware | 23:39:23 |
| KFears (tragedy arc) | In reply to @jade_:matrix.org
- well the issue you have is that there isn't like, a vagrant image for macOS or something, and that macOS has a bit of weirdness with needing a user logged in at the console to have certain functionality (possibly required for installer? unsure!) work properly even over ssh. the troubles you would have to work through are chiefly that the image is huge and that creating the image in the first place is manual, but might be able to be automated with a virtual keyboard and maybe some OCR.
- do you mean like, a chroot on macOS? chroot exists on macOS but it is missing a bunch of critical functionality and is banned by SIP. we don't care about SIP for CI because we can turn it off. but you still won't have bind mounts and enough of our installer is doing weird things to the base system that i don't think that chroot would be sufficiently representative compared to booting a throwaway VM snapshot.
booting throwaway VM snapshots is also not that slow, especially if zfs or anything else copy on write is implicated.
Well, my understanding is that chroot or VMs is about the only ways to test installer | 23:40:13 |
| jade_ | and i think that somewhat amusingly the most useful way to solve this would comply with apple EULA | 23:40:15 |
| jade_ | since we would want to be able to do it on M1 and you can perfectly legally have M1 VMs for software development | 23:40:33 |
| KFears (tragedy arc) | Which kinda then makes lix-installer not very supported at all on MacOS | 23:41:04 |
| jade_ | but i suspect that any qemu based solution would generalize to both m1 and not m1. | 23:41:04 |
| jade_ | In reply to @kfears:matrix.org
Which kinda then makes lix-installer not very supported at all on MacOS well, sort of. it just makes it very brittle | 23:41:32 |
