| 10 Mar 2025 |
 ElvishJerricco | is that what EFI PEs are? | 17:58:37 |
 aloisw | Or maybe they mandate relocalizable binaries there. | 17:58:54 |
 raitobezarius | i don't remember everything in the spec but it was non a trivial simplification | 17:58:56 |
| raitobezarius | the only bad consequences I remember is a neutralization of the microkernel performance wrt to IO ports | 17:59:15 |
| raitobezarius | anyway, reporter should just provide more HW information for its x86_64 stuff | 17:59:48 |
| raitobezarius | if he's booting on Emerald Rapids, it becomes plausible | 17:59:59 |
| raitobezarius | * if they're booting on Emerald Rapids, it becomes plausible | 18:00:03 |
| ElvishJerricco | what's special about emerald rapids here? | 18:00:15 |
| raitobezarius | well, among all CPU architectures that people have been booting, I think most people have been up to 3rd Xeon Scalable generation or 4th Xeon Scalable gen | 18:00:57 |
| raitobezarius | 5th is too recent | 18:00:59 |
| raitobezarius | I booted baremetal NixOS on 2th gen & 3rd gen, but never on 4th & 5th yet so far | 18:01:37 |
| raitobezarius | and I think many server operators on NixOS already booted on AMD Genoa | 18:01:54 |
| ElvishJerricco | ok so the point is that it's new enough hardware that it might have funkiness we haven't tested against? | 18:02:02 |
| raitobezarius | (and we have good experience with x86_64-v3) | 18:02:04 |
| raitobezarius | if it was a kernel crash, I could also say that it's old enough HW that the kernel breaks because I have many of those on my desk | 18:02:49 |
| raitobezarius | (VPD page sizes are completely fucked up on SCSI since NixOS 21.05) | 18:03:09 |
| raitobezarius | (still not fixed in any kernel) | 18:03:17 |
| raitobezarius | but also if they say I'm booting with coreboot, it's yet another set of plausible I guess (?) | 18:03:52 |
| aloisw | In reply to @aloisw:julia0815.de
Or maybe they mandate relocalizable binaries there. Ok it seems that maybe PE is always relocalizable. | 18:04:51 |
| raitobezarius | except on RISC-V | 18:05:08 |
| raitobezarius | * except on RISC-V for COFF | 18:05:14 |
| aloisw | I knew that stuff better 2 years ago. | 18:05:35 |
| ElvishJerricco | oh great. The user practically ignored my request for HW info and decided to bisect nixpkgs instead of considering the systemd issue I linked | 18:07:11 |
| ElvishJerricco | guess we'll see them again in a day or two | 18:07:34 |
| raitobezarius | (lol) | 18:08:07 |
| raitobezarius | i think it's a great experience to bisect baremetal HW | 18:08:15 |
| raitobezarius | I totally did not put my little brother to this task for the last 2 weeks | 18:08:28 |
| raitobezarius | they would have a better chance just reverting the systemdp atch | 18:09:05 |
| raitobezarius | now we are talking about bisections, i'm going to post this: http://kernel.org/pub/linux/kernel/people/paulmck/Answers/Heisenbugs/QQAbisection.html | 18:09:18 |
| ElvishJerricco | raitobezarius: Do you know if it would be possible to make qemu emulate an x86_64 system with this sort of issue? | 18:11:43 |
