<LnL>
bisect points me to 0186286433cccbf6faf288c22df03e63e46dd94d for the 17.09 failure
la_putin has joined #nixos-dev
<gchristensen>
w00t rust is on to stage2.
ma27 has quit [Ping timeout: 256 seconds]
goibhniu has quit [Ping timeout: 264 seconds]
jtojnar has quit [Ping timeout: 268 seconds]
orivej has quit [Ping timeout: 240 seconds]
orivej has joined #nixos-dev
mbrgm_ has joined #nixos-dev
mbrgm has quit [Ping timeout: 268 seconds]
mbrgm_ is now known as mbrgm
la_putin has quit [Remote host closed the connection]
la_putin has joined #nixos-dev
orivej has quit [Ping timeout: 246 seconds]
jtojnar has joined #nixos-dev
la_putin has quit [Read error: Connection reset by peer]
la_putin has joined #nixos-dev
yegortimoshenko has quit [Remote host closed the connection]
yegortimoshenko has joined #nixos-dev
Dezgeg has quit [Ping timeout: 276 seconds]
la_putin has quit [Quit: Konversation terminated!]
el_putin has joined #nixos-dev
el_putin has quit [Read error: Connection reset by peer]
el_putin has joined #nixos-dev
el_putin has quit [Remote host closed the connection]
el_putin has joined #nixos-dev
ma27 has joined #nixos-dev
orivej has joined #nixos-dev
jtojnar_ has joined #nixos-dev
jtojnar has quit [Ping timeout: 268 seconds]
jtojnar_ is now known as jtojnar
JosW has joined #nixos-dev
jtojnar has quit [Ping timeout: 240 seconds]
orivej has quit [Ping timeout: 248 seconds]
ma27 has quit [Ping timeout: 276 seconds]
ma27 has joined #nixos-dev
el_putin has quit [Read error: Connection reset by peer]
el_putin has joined #nixos-dev
el_putin has quit [Read error: Connection reset by peer]
el_putin has joined #nixos-dev
Dezgeg has joined #nixos-dev
Dezgeg has quit [Client Quit]
<Mic92>
gchristensen: you are currently testing the rust pr?
el_putin has quit [Ping timeout: 264 seconds]
Dezgeg has joined #nixos-dev
pie_ has quit [Ping timeout: 240 seconds]
la_putin has joined #nixos-dev
jtojnar_ has joined #nixos-dev
jtojnar_ is now known as jtojnar
ma27 has quit [Ping timeout: 256 seconds]
<gchristensen>
I'm not, Mic92
<MichaelRaskin>
Is testing borgable?
ma27 has joined #nixos-dev
<gchristensen>
should be :)
<LnL>
oh, I commented on the wrong pr?
<MichaelRaskin>
I was not sure _which_ PR is discussed
<MichaelRaskin>
LnL: probably you picked the correct one
<gchristensen>
I'm not sure either :)
<LnL>
what's the attribute, don't see it exposed anywhere
<gchristensen>
are we talking cargo-vendor?
<LnL>
yeah I see cargoVendor = ... in the source
<LnL>
but it's not in rust or rustPlatform
<gchristensen>
I don't know :/ so that rust build last night timed out after 10hrs
<LnL>
I tested with enableParallelBuilding again, but that didn't seem to make a difference in time for me
<gchristensen>
:/
<gchristensen>
like it or not, rust is about to be a channel-blocker :P it'd be good to sort out
<LnL>
not sure why, I got it down to ~30min last time I tried that
<LnL>
not rebuilding llvm every time would also help
orivej has joined #nixos-dev
el_putin has joined #nixos-dev
la_putin has quit [Read error: Connection reset by peer]
<LnL>
oh! the vendored llvm might be exactly why it's broken on 17.09
ciil has joined #nixos-dev
zarel has joined #nixos-dev
pie_ has joined #nixos-dev
ma27 has quit [Quit: WeeChat 2.0]
ma27 has joined #nixos-dev
vcunat has joined #nixos-dev
<gchristensen>
I'm thinking build sampling might be a thing I could do in a horribly hacky way, of running nix-build -j0 and find the first one in the list of things to build which refers to a named attr in nixpkgs. can anyone imagine a way to do this? or do this better?
<gchristensen>
the best way I can think of would involve generating the package DB which includes attr names and out paths, and doing a matchup
<vcunat>
gchristensen: missing context - what exactly you want?
<gchristensen>
I want to let people just ask ofborg to build without having to worry about the exact thing it should be building. ex: in this PR there is a single correct thing ofborg should attempt: https://github.com/NixOS/nixpkgs/pull/34094
<vcunat>
well, it's easy to get the list of changed attributes from the rebuild-amount script
<gchristensen>
but it must be cross-architecture compatible (ie: by attribute name, and not by .drv) and preferably without having to pre-calculate the per-arch .drv's for each architecture
<vcunat>
but that list won't help you much for larger rebuilds
<vcunat>
so, it could be like - if the list is short enough, just pick two random attributes :-)
<gchristensen>
true! :D
<vcunat>
or the rebuild-amount part could just comment with the list, if it's short enough
<vcunat>
BTW, I recently added actual --print option to the script
<gchristensen>
I remember that. I don't actually use that script anymore, though the change looks good
<LnL>
I do :)
<gchristensen>
if the list is long I could do some drv parsing to get closer to the root of the change
<gchristensen>
btw LnL I finally added some automatic tests to the builder :') it was hard
<LnL>
nice!
<vcunat>
well, you can do `nix-store -q --graph foo.drv`
<vcunat>
I'm not sure if there's a convenient tool to "subtract" two dot-graphs
<MichaelRaskin>
vcunat: well, you could list the vertices, sort and diff.
<vcunat>
it might be best to (1) find common vertices (store names), (2) remove them from the second graph, (3) look at vertices that depend on nothing (on the augmented graph)
<gchristensen>
I think the depend-on-nothing version is more on the right track than nix-diff. is this something you'd be interested in taking a crack at? :)
<gchristensen>
bash-or-whatever is perfectly fine...
<vcunat>
well, not today
<vcunat>
And I'm perfectly OK if someone else does it ;-) but I think I could manage during the next week.
<vcunat>
Related: I'd actually love to have grouping by the root of failure for Hydra's newly-failed list.
<gchristensen>
that would be very cool :)
<vcunat>
When you have very many of them, it's hard to know if everything is caused by the same error or not.
<vcunat>
I always have to resort to sampling a few of them...
zarel has quit [Quit: Leaving]
<LnL>
that would be super nice
<vcunat>
To be clear, I don't expect to be capable of writing in perl soon.
<gchristensen>
well, let's see where ofborg leads. I'd love it if somebody were to implement this script as described by vcunat, even if it takes several days =)
<vcunat>
whoever implements this, I expect it would best fit into <nixpkgs/maintainers/scripts/> intially?
<gchristensen>
yeah I agree
<LnL>
I should probably also add some of my aliases there
{^_^} has quit [Remote host closed the connection]
{^_^} has joined #nixos-dev
<vcunat>
I think we miss some co-organization in creating a toolset for "common" tasks
<vcunat>
(nix-related ones)
<gchristensen>
I agree
<MichaelRaskin>
9999
<MichaelRaskin>
«I think we miss some co-organization» is true, the rest is details
<vcunat>
:-)
<LnL>
aaah, if I have to rebuild llvm one more time...
<vcunat>
That's the bug with ld detecting bogus macos versions?
<vcunat>
Oh, great, so I can stop caring about those errors.
<LnL>
yep, I tested that with a /tmp path and it seemed to work
<LnL>
don't know for sure ofcorse since it changed the store path :p
<vcunat>
Aha, rustc is also regressing on darwin.
<LnL>
that's an intermittent failing test, thought I disabled it with the last update
<LnL>
I'm looking at the 17.09 failure + trying to get rid of the vendored llvm build
<vcunat>
Then I suppose I'll just wait for more builds to finish (2.5k darwin jobs in queue), check the amount of newly-failing, and merge that commit to master.
<vcunat>
Probably tomorrow.
<vcunat>
It would be nice if niksnut (or anyone) could unblock the darwin builders, as half of them is stuck with days-old builds.
<LnL>
sounds good, then we have the cmake fix for libclang in master
<LnL>
:/
<vcunat>
yes
<gchristensen>
vcunat: copumpkin and I both can do that, I can help with that in ~20min
<vcunat>
Oh, I didn't know that. Maybe we should have a list somewhere :-)
<vcunat>
They're stuck for days, so 20 minutes is perfectly fine ;-)
<gchristensen>
that could work, but the page is hard to update
<vcunat>
I don't expect this information would change a lot.
<vcunat>
But certainly better some place than none.
<samueldr>
for organizational stuff, on the wiki, it may be better to namespace it, since the wiki is otherwise built with the end-user in mind; e.g. https://nixos.wiki/wiki/Org:Hydra for the nixos organization hydra specific informations (if wanted)
<gchristensen>
yeah, I sort of wish there was a place for operational docs
ma27 has quit [Quit: WeeChat 2.0]
goibhniu has joined #nixos-dev
taktoa has joined #nixos-dev
infinisil has quit [Quit: ZNC 1.6.5 - http://znc.in]
vcunat has quit [Quit: Leaving.]
infinisil has joined #nixos-dev
<gchristensen>
aw vcunat is gone
JosW has quit [Quit: Konversation terminated!]
<LnL>
any ideas what could cause 'SSE2 was not detected as available on an x86 platform'?
<Mic92>
broken compiling test in autotools?
<Mic92>
(broken for other reasons)
<LnL>
it's a rust test, but it woks fine with the default llvm build: run-pass/sse2.rs ... ok
<LnL>
I'll do a test with the same cmake flags instead of what we use by default