#asahi · 2023-11-15 04:44
nicolas17today they released the KDK for macOS 13.6.2#asahi · 2023-11-15 04:43
nicolas17wtf is going on at Apple#asahi · 2023-11-15 04:43
marcanapparently.#asahi · 2023-11-15 04:43
nicolas17the "fix" was *only* making it not write the EAN data and everything else is still broken?#asahi · 2023-11-15 04:40
nicolas17:|#asahi · 2023-11-15 04:39
marcannicolas17: well that was a huge waste of time. 13.6.2 is still fucked. I'm not sure why I thought it wasn't. going to double check with macOS now but...#asahi · 2023-11-15 02:59
marcanah yeah#asahi · 2023-11-15 02:41
nicolas17I've been looking at bxdiff because it seems I can *partially* recover v1 from a v2 ipsw and a v1->v2 delta, which could be interesting for people who want to reverse engineer files from a v1 we have no download for (preinstalled OS versions)#asahi · 2023-11-15 02:35
nicolas17lzma.decompress(blk, format=lzma.FORMAT_XZ)#asahi · 2023-11-15 02:33
marcanworks for me#asahi · 2023-11-15 02:33
marcanoh ha#asahi · 2023-11-15 02:33
nicolas17Python 3.3 has an lzma module builtin#asahi · 2023-11-15 02:32
marcanthat works on linux, need to replace unxz with whatever thing on macos, probably the libcompression stuff I assume handles it like it does lzfse#asahi · 2023-11-15 02:32
marcanhttps://gist.github.com/marcan/3b7a596816cca9b26c1d108c95a3e730#asahi · 2023-11-15 02:32
nicolas17should've expected you'd be faster :P#asahi · 2023-11-15 02:31
marcansha1 matches. done.#asahi · 2023-11-15 02:30
marcannicolas17: I think I have it working#asahi · 2023-11-15 02:19
marcanbut also if this is in the compression framework, we already use that for lzfse#asahi · 2023-11-15 02:19
nicolas17the format is explained in the manpage :)#asahi · 2023-11-15 02:19
marcanis that available in recoveryOS?#asahi · 2023-11-15 02:19
nicolas17marcan: "man compression_tool" on your nearest macOS#asahi · 2023-11-15 02:18
marcanlet me hack something up to unchunk it#asahi · 2023-11-15 02:18
marcanyeah I think this is just xz archives stuck together with some headers#asahi · 2023-11-15 02:17
marcanI'm okay shipping unxz as a blob or whatever if it comes to that#asahi · 2023-11-15 02:17
marcanlooks like it's bog standard xz compression, except there's some weird chunking or something#asahi · 2023-11-15 02:16
marcancutting the file right there and using 7z x yields... the beginning of the DMG before an error#asahi · 2023-11-15 02:15
nicolas17(fun fact, the format is actually pbz, the x is a byte saying "xz compression" :D)#asahi · 2023-11-15 02:15
marcanno#asahi · 2023-11-15 02:15
nicolas17does the installer have a pbzx decoder already?#asahi · 2023-11-15 02:14
marcanhonestly we can probably just hardcode the pbzx start for this use case#asahi · 2023-11-15 02:14
marcanyes#asahi · 2023-11-15 02:14
nicolas17oops I just OOM'd my system with imhex#asahi · 2023-11-15 02:13
nicolas17but instead of trying to figure out what the flags mean, I will assume those fields are missing and assert that flags is 0#asahi · 2023-11-15 02:13
nicolas17I think the unknown field is flags and the flags on this particular file declare that diff_size and target_sha1 are missing#asahi · 2023-11-15 02:11
marcanlooking at https://github.com/ezhes/bxdiff50/blob/master/bxdiff50 and against the file it looks like for this "dumb" format the header is different (or rather: possibly the tail is part of controldata) and it should just be a single pbzx compressed extradata section#asahi · 2023-11-15 02:10
marcannp#asahi · 2023-11-15 02:07
nicolas17sorry >_< I was busy today dealing with Apple betas (they released macOS 14.2b3 in your sleep) followed by nonsense for a conference (see #asahi-offtopic)#asahi · 2023-11-15 01:58
marcanfwiw if it comes to shipping a precompiled blob helper or something I'm okay with that#asahi · 2023-11-15 01:58
marcannicolas17: how far did you get with that diff thing? looks like the firmware is the same so that's probably our only blocker for shipping this#asahi · 2023-11-15 01:56
marcanwith the upcoming driver update and overrides it will start but crash if you try to do certain things#asahi · 2023-11-15 01:56
marcanbut it also has complex shaders that need register spilling which is not implemented yet#asahi · 2023-11-15 01:56
marcanblender needs geometry shaders, which isn't released yet#asahi · 2023-11-15 01:55
marcanChaosPrincess: it will not work#asahi · 2023-11-14 22:31
holidayfor sure better than nothing :D#asahi · 2023-11-14 22:30
j`eyperectly usuable :p#asahi · 2023-11-14 22:25
holidayi think it's around 3 to 4 fps with the CPU running at 95-97% just by watching the default cube#asahi · 2023-11-14 22:25
holidayand the performance...well it started at least#asahi · 2023-11-14 22:25
holidayit seems like the version is still 3.6.4 #asahi · 2023-11-14 22:24
holidayok it started#asahi · 2023-11-14 22:23
j`eyLIBGL_ALWAYS_SOFTWARE=1 blender