00:30
<
gchristensen >
Cstrahan can you confirm it is fixed?
00:34
<
cstrahan >
gchristensen: yep, the arithmetic error went away -- thanks!
00:48
<
gchristensen >
Great!
01:16
LnL has quit [Ping timeout: 255 seconds]
01:17
LnL has joined #nixos-aarch64
02:39
orivej has quit [Ping timeout: 265 seconds]
10:25
orivej has joined #nixos-aarch64
11:10
<
Mic92 >
cstrahan: there is already a github repo for rock64
11:11
<
Mic92 >
working fine, but should be upstreamed
11:11
<
Mic92 >
I also have a rock64.
11:11
<
LnL >
Mic92: btw, about the rustc test failures
11:12
<
LnL >
some of those look like they could be caused by the overlayfs
11:12
<
Mic92 >
LnL: interesting.
11:12
<
Mic92 >
do they use hardlinks?
11:13
<
Mic92 >
gchristensen: could you make /tmp a tmpfs some time?
11:14
<
Dezgeg >
it used to build on the box fine, what are the failures?
11:14
<
LnL >
just a suspicion, I might be wrong
11:14
buovjaga has joined #nixos-aarch64
11:15
<
Mic92 >
probably they all have the same root cause
11:15
<
Dezgeg >
there's the #!/bin/sh issue at least
11:15
<
Dezgeg >
try without sandbox
11:16
<
Dezgeg >
(/build/rustc-1.24.0-src/src/etc/cat-and-grep.sh: line 64: syntax error: support for $((arith)) is disabled)
11:16
<
LnL >
yeah, did you fix the getopts stuff?
11:16
<
Dezgeg >
not sure what's the status on that
11:19
<
Dezgeg >
cstrahan: did you append console=ttyS1,115200n8 or replace the other serial console= args? I think the kernel only supports one serial device at a time, so if the ttyAMA0 is earlier in the cmdline it might ignore the later ones
11:19
<
LnL >
that doesn't work in sh
11:21
<
Mic92 >
if that would the problem, It would also occure on x86_64
11:21
<
Mic92 >
I can replace the particular shebang with a bash one though.
11:21
<
Dezgeg >
x86 might be running a different version of nix
11:21
<
Dezgeg >
just try it without sandbox (for now)
11:22
<
LnL >
depends on the sandbox-paths
11:23
<
Mic92 >
Dezgeg: I can't do that on aarch64-build-box
11:24
<
Dezgeg >
--option build-use-sandbox false, no?
11:24
<
Dezgeg >
it worked for me at least
11:25
<
Dezgeg >
(on different stuff)
11:25
<
Mic92 >
I though this option is ignored, if NIX_DAEMON is running
11:26
<
LnL >
not if you're a trusted user
11:26
<
LnL >
and you need a trusted user for distributed builds
11:28
<
Mic92 >
seems to work
11:35
orivej has quit [Ping timeout: 256 seconds]
11:40
orivej has joined #nixos-aarch64
11:49
orivej has quit [Ping timeout: 240 seconds]
11:50
<
LnL >
the --option or the rust build?
12:04
<
Mic92 >
disabling sandbox.
12:04
<
Mic92 >
Build is still running
12:15
buovjaga has quit [Remote host closed the connection]
12:18
<
LnL >
btw, I tested the parallel build for 1.24.0 and that looks a lot better
12:19
<
LnL >
might help a lot for the aarch64 build
12:24
<
gchristensen >
Mic92 perhaps you could send a PR?
12:25
<
gchristensen >
Also
12:25
<
gchristensen >
, what for?
15:52
orivej has joined #nixos-aarch64
16:03
orivej has quit [Ping timeout: 248 seconds]
16:09
<
Mic92 >
LnL: without sandbox it builds fine, so this is not related to overlayfs
16:10
<
LnL >
hmm, gchristensen the hydra builder has sandboxing enabled right?
16:11
<
Mic92 >
LnL: it might be a broken shebang though.
16:12
<
Mic92 >
also I don't know why this only happens on aarch64
16:12
<
Dezgeg >
no, this #!/bin/sh in nixUnstable really needs just to be fixed
16:13
<
Mic92 >
this is what my next build tries
16:13
<
Dezgeg >
I mean fixing nix
16:13
<
Dezgeg >
not patching tons of packages that used to work just fine previously
16:13
<
gchristensen >
yeah...
16:14
<
LnL >
did something change with 1.12+?
16:14
<
gchristensen >
yeah it uses a janky bash now
16:14
<
gchristensen >
or janky non-bash
16:14
<
LnL >
oh the busybox thing?
16:14
<
Mic92 >
gchristensen: so I have to put bash into buildInputs?
16:15
<
gchristensen >
I don't know
16:15
<
LnL >
meh, people if people use sh they should do so correctly
16:15
<
Mic92 >
of course they should, this is step two
16:16
<
Mic92 >
the alternative would be to use usr/bin/env bash
16:16
<
Mic92 >
This is what go does
16:16
<
LnL >
yeah, just patch the script to make it use bash
16:16
<
Mic92 >
I can make this an upstream patch then.
16:18
<
Mic92 >
When did this SHELL=/bin/dash change?
16:25
<
Mic92 >
LnL: according to shellcheck the script does not contain bashism, just our sh is too limited
16:26
<
LnL >
huh, I'm very sure $(( ... )) is a bashism
16:30
<
LnL >
oh hmm, I'm thinking of if (( $i )); then ...
19:25
<
dtz >
TBH I really wasn't sure until I set to trying to chase down failures I was seeing xD
21:44
orivej has joined #nixos-aarch64
22:56
<
sphalerite_ >
I'm trying to get an armv7-capable aarch64 machine to build armv7 stuff with nix, but I'm getting "bad system call" at some point. I want to use gdb to work out where exactly this is happening, but I'm not sure how to control it to get the right process. Does anyone know how to make gdb follow a specific fork but not all forks?
23:35
orivej has quit [Ping timeout: 248 seconds]