<clever>
sphalerite: the -d in nix-collect-garbage will delete generations from profiles you have +w to, then it will do a normal gc, which affects things built by anybody
<clever>
/data/data/com.termux/files/nix/store/ac1pzls6p1s2z10svz5ah8wzz4z0qqfj-unpack-bootstrap-tools.sh: line 19: /data/data/com.termux/files/nix/store/9741plqqjkjnaxm7yzw5jh7x63kz9sn6-bootstrap-tools/lib/ld-*so.?: not found
<clever>
gchristensen: look inside the dir mentioned, do you have an ld-*.so?
<clever>
oh
<clever>
he left already
<clever>
thats what messed up my tab complete
lopsided98 has joined #nixos-aarch64
lopsided98_ has joined #nixos-aarch64
lopsided98 has quit [Ping timeout: 255 seconds]
<sphalerite>
clever: yes, that's what I said.
<sphalerite>
Dezgeg: booting the kernel directly using qemu's loading support rather than u-boot works
<sphalerite>
Hm, any way I can use PAE so I can get more than 3GB RAM in the single VM?
<Dezgeg>
that needs a kernel recompile to enable LPAE
<sphalerite>
oh
<sphalerite>
any reason it isn't enabled in the default kernel?
<Dezgeg>
it won't work on non-LPAE hardware then
<sphalerite>
oh, that's a pain
<sphalerite>
why is that?
<Dezgeg>
dunno, maybe it was too hard to implement dynamic switch of the page table formats
<sphalerite>
That just gave me a funny/silly idea
<sphalerite>
It should be possible in principle to create a kernel that works on machines with compeltely different instruction sets, shouldn't it?
<sphalerite>
maybe not in the fully general case, but depending on instruction encoding I bet one could make a kernel that supports both MIPS and x86 for instance.
<sphalerite>
Since there are so many ways of encoding NOPs, I'm sure there are some ways that collide with a useful jump
<Dezgeg>
for some two arches known in advance, probably been done before, but I think in general case it's too hard
<clever>
that reminds me of what i first did with nixos when starting out
<clever>
i had an SD card, that booted x86-64 and armv7l, with full nixos, sharing the same rootfs and store
<clever>
the key that let me avoid opcode collisions, was that each had a unique bootloader
<clever>
so i just had to duplicate /nix/var/nix/profiles/ and set nixos up to bind-mount either profiles.x86 or profiles.arm to profiles, then make sure both are gc roots
<clever>
so now nix-env is peruser and perarch
orivej has quit [Ping timeout: 256 seconds]
orivej has joined #nixos-aarch64
orivej has quit [Ping timeout: 240 seconds]
orivej has joined #nixos-aarch64
kahiru has quit [Remote host closed the connection]
<andi->
Anyone here motivated/with spare time to look into an aarch64 issue with Firefox60(+)? The package bump is currently in staging and is failing on aarch64. It first complains about sse2 not existing (enabling webrtc gets rid of that :/) and then fails for yet to investigate reasons.. The build output is visible at
<giaco_>
problem is I'm calling this command from bash outside bootstrap
<giaco_>
I am not even capable of printing the path of the shell executed by the bootstrap. Seems that glob is not working as at line 19 (LD_LIBRARY_PATH=$out/lib $LD_BINARY $out/bin/cp $out/bin/patchelf .) he is trying to launch "ld-*so.?"
<giaco_>
I am adding some test code before line 19 to check what is happening
<giaco_>
I cannot explain how is possible that the bootstrap busybox ash goes wild on bootstrap unpack-bootstrap-tools.sh. Statically linked shell + script, why it is not working for me?