#
#asahi-alt · 2026-02-16 05:51
leioand on downtimes chip away at the package list that was compiled with armv8.5-a via some grepping of /var/db/pkg/*/*/CFLAGS
#asahi-alt · 2026-02-16 05:51
ink1 day 7 hours 24 minutes
#asahi-alt · 2026-02-16 05:50
leiopersonally I would just gradually migrate and rebuild specific things if something looks broken like ffmpeg did; the wrong armv8.5-a wasn't that great
#asahi-alt · 2026-02-16 05:49
inki should probably run it through genlop first i want to see how long it thinks it will take
#asahi-alt · 2026-02-16 05:49
leioI said on top of your normal world line to get stuff like "--ask" (-a) or such ;)
#asahi-alt · 2026-02-16 05:49
inkso the update actually doesn't matter then if it just recompiles everything from scratch anyways?
#asahi-alt · 2026-02-16 05:48
inkoh
#asahi-alt · 2026-02-16 05:48
leiosure, shouldn't matter, -e makes it think nothing is installed for dep resolution
#asahi-alt · 2026-02-16 05:48
inkor literally just emerge -e @world
#asahi-alt · 2026-02-16 05:48
inkwith update in it still?
#asahi-alt · 2026-02-16 05:47
leio-e flag to your normal @world line added
#asahi-alt · 2026-02-16 05:47
leiomeh, why do we have vfp flags unmasked on arm64
#asahi-alt · 2026-02-16 05:47
inkwhat would i type in to recompile everything
#asahi-alt · 2026-02-16 05:45
leiosharing would be kind of needed if we did 32-bit multilib on arm64, but we do not
#asahi-alt · 2026-02-16 05:44
leioand that's the problem with sharing CPU_FLAGS_ARM in a nutshell
#asahi-alt · 2026-02-16 05:44
inki see
#asahi-alt · 2026-02-16 05:44
inkoh
#asahi-alt · 2026-02-16 05:44
leioarm32 things, profiles/desc/cpu_flags_arm.desc document these, I don't recall what tools output those for user convenience
#asahi-alt · 2026-02-16 05:44
inki assume for armv4 through armv8, and if m1 is 8.4 why would those tags not be relevant
#asahi-alt · 2026-02-16 05:43
inkwhat are v4 through v8
#asahi-alt · 2026-02-16 05:42
leio(when I say many, it's not many in count; not many packages overall have hand-written assembly to be using CPU_FLAGS_*)
#asahi-alt · 2026-02-16 05:42
leioso it is masked to not break, on arm64 packages should just unconditionally enable 64-bit neon on arm64 without flags, and many do; some don't and needs sorting.
#asahi-alt · 2026-02-16 05:41
inkwtf
#asahi-alt · 2026-02-16 05:41
leiosome ebuilds that react to CPU_FLAGS_ARM=neon are for 32-bit neon and will fail to compile with 64-bit neon
#asahi-alt · 2026-02-16 05:41
leiobut it is masked on arm64, except some things that unmask it
#asahi-alt · 2026-02-16 05:40
leiothat's a whole other mess from this. But it was there for you.
#asahi-alt · 2026-02-16 05:36
inkeither
#asahi-alt · 2026-02-16 05:36
inkwait neon isn't even on my list of flags
#asahi-alt · 2026-02-16 05:35
JamesCalligeros[m]now that there are multiple desktop-class arm64 platforms, problems like this are becoming more difficult to ignore
#asahi-alt · 2026-02-16 05:35
JamesCalligeros[m]like all non-x86 platforms, arm64 has been considered an afterthought for tinkerers playing with grown up toys until relatively recently
#asahi-alt · 2026-02-16 05:34
JamesCalligeros[m]historical reasons
#asahi-alt · 2026-02-16 05:32
ink*way
#asahi-alt · 2026-02-16 05:32
inkwhy is it not already that wa
#asahi-alt · 2026-02-16 05:32
leio(yes, separate USE_EXPAND)
#asahi-alt · 2026-02-16 05:31
leioI don't understand why the tool invents those thumb and co things on arm64 in the first place
#asahi-alt · 2026-02-16 05:31
leioI think we know what's going to happen, just... need to do the work :(
#asahi-alt · 2026-02-16 05:31
JamesCalligeros[m]are we getting our own USE_EXPAND?
#asahi-alt · 2026-02-16 05:30
JamesCalligeros[m]no point fixing cpuid2cpuflags until we know whats going to happen long term with arm64
#asahi-alt · 2026-02-16 05:30
leioit's not that critical as those extras are just masked on arm64 anyways afair
#asahi-alt · 2026-02-16 05:30
JamesCalligeros[m]whats the point when the whole CPU_FLAGS_ARM situation is fucked anyway
#asahi-alt · 2026-02-16 05:30
leioI wonder if I could nerdsnipe chaos_princess or someone to look at cpuid2cpuflags instead of talking to the tools maintainer/author or filing bugs :P
#asahi-alt · 2026-02-16 05:29
leioso I guess that means we filter out some to start with and are left with "neon aes sha1 sha2 crc32 asimd asimddp asimdfhm asimdhp"
#asahi-alt · 2026-02-16 05:27
inkssbs sb paca pacg dcpodp flagm2 frint
#asahi-alt · 2026-02-16 05:27
inkfp asimd evtstrm aes pmull sha1 sha2 crc32 atomics fphp asimdhp cpuid asimdrdm jscvt fcma lrcpc dcpop sha3 asimddp sha512 asimdfhm dit uscat ilrcpc flagm
#asahi-alt · 2026-02-16 05:27
leioif it's just a single line picked out like that, yes
#asahi-alt · 2026-02-16 05:27
inkunless that's a problem
#asahi-alt · 2026-02-16 05:26
inki mean i could just copy it here
#asahi-alt · 2026-02-16 05:26
inkfor lscpu
#asahi-alt · 2026-02-16 05:26
leioyes, Flags
#asahi-alt · 2026-02-16 05:26
inkor the whole thing

plus ›