<{^_^}>
#34409 (by steveeJ, 34 weeks ago, merged): qemu-vm: fix script syntax and VM args
pie___ has joined #nixos-dev
pie__ has quit [Ping timeout: 260 seconds]
pie___ has quit [Read error: Connection reset by peer]
pie___ has joined #nixos-dev
<Mic92>
srhb: I guess so
<Mic92>
too bad this was not catched by ofborg
<srhb>
Mic92: Yeah! Oh well, things to consider for improvement. :-) Though i think evaluation of release-combined is pretty slow.
ciil has quit [Quit: Lost terminal]
ciil has joined #nixos-dev
orivej has joined #nixos-dev
<steveeJ>
srhb: you pinged me. About the qemu-vm revert?
<srhb>
steveeJ: Yup, just wanted to hear if you had a proper solution in the pipeline before I reverted, but I did it already. :)
<steveeJ>
I read the updates on the PR. I will try to pair up with aszlig if possible and get this sorted out properly. There's a lot of redundancy that can be reduced around the qemu-vm expressions. The Perl test driver also has conditionals for checking the disk interface. Ideally we combine all of that in Mix somewhere
<steveeJ>
s/Mix/Nix
<steveeJ>
I don't have more than what's in the PR. I should run all nixos/tests next time ;-)
<aszlig>
steveeJ: mhm, at some point i'd rewrite all that stuff into using a list in nix or maybe even an attrset (the latter would make it easier to override)
<aszlig>
unfortunately qemu has some arguments that are order-dependant, so not sure whether an attrset would work
<aszlig>
otoh... if we directly pass options as a qdict an attrset would work
<aszlig>
actually... i'm wrong... the arglist is parsed into a qdict anyway, so it shouldn't be order dependent
<aszlig>
so i'd opt for qemu -readconfig, which reads an ini-style config
<aszlig>
it's then just a matter of generators.toINI
<aszlig>
we could even go so far and automatically generate nixos options for *all* the qemu options
<aszlig>
all the options are introspectable and have types, eg: QEMU_OPT_BOOL QEMU_OPT_BPS QEMU_OPT_IOPS QEMU_OPT_NUMBER QEMU_OPT_SIZE QEMU_OPT_STRING QEMU_OPT_THROTTLE
<niksnut>
that's undesirable due to maintainability and module system bloat
<aszlig>
niksnut: maintainability shouldn't be an issue, because it's just one command that directly spits out module options... module system bloat on the other hand is
<niksnut>
aszlig: it tightly couples the nixos module to a particular version of qemu
<niksnut>
also, somebody needs to run that command (and figure out how to do so)
<aszlig>
qemu/update.sh =)
orivej has quit [Ping timeout: 252 seconds]
phreedom has quit [Ping timeout: 256 seconds]
phreedom_ has joined #nixos-dev
phreedom_ has quit [Ping timeout: 256 seconds]
phreedom has joined #nixos-dev
<gchristensen>
during a remote build if the nix-build client disconnects should the remote build be canceled?
<gchristensen>
right now I'm showing that didn't happen with a kernel build when my laptop crashed
orivej has joined #nixos-dev
globin has quit [Remote host closed the connection]
orivej has quit [Ping timeout: 260 seconds]
globin has joined #nixos-dev
<sphalerite>
gchristensen: maybe the remote sshd is still running and hoping the connection might come back?
<gchristensen>
ah maybe so
<sphalerite>
as far as I've observed the builds did generally stop when the connection was lost ,although it might not happen immediately
<sphalerite>
very frustrating!
<steveeJ>
aszlig: no doubt that'd work much more reliably than what we currently have. it does sound a lot like rebuilding libvirt to a certain extent. has it been considered to integrate that into the tests?
globin has quit [Quit: o/]
globin has joined #nixos-dev
pie_ has joined #nixos-dev
makefu has quit [Ping timeout: 268 seconds]
pie___ has quit [Ping timeout: 252 seconds]
pie_ has quit [Read error: Connection timed out]
pie_ has joined #nixos-dev
makefu has joined #nixos-dev
orivej has joined #nixos-dev
aminechikhaoui has quit [Ping timeout: 240 seconds]
<domenkozar>
gchristensen: aha, 2.1.2 breaks macos since it wants to source daemon profile
<gchristensen>
aye
<gchristensen>
I thought it might
<domenkozar>
total used free shared buffers cached
<domenkozar>
Mem: 7479 1391 6088 179 170 939
<domenkozar>
so nix has 6G of memory
<domenkozar>
and with -j1 it runs out?
<domenkozar>
and it needs to unpack 4GB for the whole substituter package
<domenkozar>
something is really strange here
makefu has joined #nixos-dev
makefu has quit [Client Quit]
MichaelRaskin has joined #nixos-dev
<Mic92>
domenkozar: I thought this was in an old version where it unpacked stuff to memory first.
<Mic92>
ah, you already on the new one
<Mic92>
I did not read this far
<Mic92>
would be interesting to get a stacktrace for that, one can use gdb for that.
xeji has joined #nixos-dev
<aszlig>
steveeJ: well, i think for the actual point of fixing vmWithBootLoader, we should go for a more minimal approach so it can be backported to stable
<aszlig>
steveeJ: after that we can refactor that stuff seperately
<infinisil>
I'd appreciate if somebody could merge #46453, I'd rather not merge my own non-trivial PR's
<domenkozar>
Mic92: if only I could reproduce locally
xeji has quit [Quit: WeeChat 2.1]
<steveeJ>
aszlig: agreed. I will take a look on the weekend if you haven't done so until then
goibhniu has quit [Ping timeout: 276 seconds]
lopsided98 has quit [Quit: Disconnected]
lopsided98 has joined #nixos-dev
<gchristensen>
NixOS people -- how often are we blocked in hydra by aarch64 having a lot of builds waiting?
<samueldr>
(when making my reports for 18.09, I don't often see more jobs waiting on aarch64 than x86_64-linux; darwin more often seems to be lagging behind)
<samueldr>
though stable may not see the same usage