<gchristensen>
my understanding is the idea is to bulk together big rebuilds to make it less annoying to contribute against master
<gchristensen>
for example it'd be pretty annoying if this merged and then ofborg failed to build everything in master due to timeouts
<gchristensen>
everything gtk related*
<infinisil>
gchristensen: Yeah that was also my understanding
<infinisil>
But then again, if too many things go to staging, it's hard to find breakages (because 99.9% of people use master)
<infinisil>
plus there's all those stdenv things on staging, which makes it even harder to find breakages
<gchristensen>
at any rate, I think the answer is it should go to staging
<infinisil>
I see
<MichaelRaskin>
Of course, people not using staging wouldn't be a problem if we had enough automated tests, which we won't have because NixOS tests are slow and threfore a pain to debug…
johanot has quit [Ping timeout: 240 seconds]
johanot has joined #nixos-dev
<infinisil>
Yeah.. :(
disasm has joined #nixos-dev
<infinisil>
MichaelRaskin: What do you think, should that PR go to staging or not?
<gchristensen>
I'd love to find a nicer way to do staging
<MichaelRaskin>
I think staging per se is OK, another approach would fix neither git nor our tests
<MichaelRaskin>
Rebuilding all GTK3 rev-deps is probably staging in my impression
<infinisil>
Alright, will switch the branch
<gchristensen>
thoughtpolice: 'round?
disasm has quit [Quit: WeeChat 2.0]
disasm has joined #nixos-dev
init_6 has quit []
drakonis_ has joined #nixos-dev
drakonis_ has quit [Read error: Connection reset by peer]
drakonis1 has joined #nixos-dev
drakonis has quit [Ping timeout: 245 seconds]
jtojnar_ has quit [Read error: Connection reset by peer]
drakonis1 has quit [Read error: Connection reset by peer]
drakonis1 has joined #nixos-dev
drakonis1 has quit [Read error: Connection reset by peer]
drakonis1 has joined #nixos-dev
drakonis1 has quit [Read error: Connection reset by peer]
drakonis1 has joined #nixos-dev
drakonis1 has quit [Read error: Connection reset by peer]
drakonis1 has joined #nixos-dev
drakonis1 has quit [Read error: Connection reset by peer]
<samueldr>
maybe a dep from git-annex built on something else with more cpu features, and ike missing the feature makes it fail tests
<MichaelRaskin>
As a person who has been banned from Google Search a few times (no kidding), I would also appreciate extra login options, GitHub probably beeing a reasonable default with the current development model
<infinisil>
+1 to that
jtojnar_ has joined #nixos-dev
<MichaelRaskin>
Although Hydra-only passwords that the user cannot change, only reset — also fine. We don't care about people with no password manager at all, I guess.
<MichaelRaskin>
I guess no incentive to leak such a DB, beyond the existing incentives to attack Hydra because stealable compute
<infinisil>
Idea for hydra: (ignoring building power needed) Build everything twice, then we can detect both reproducibility (if it succeeds) and transient failures (if it fails once)
<MichaelRaskin>
Maybe post-intensional-store…
<MichaelRaskin>
Also, build once but compare with ofBorg?
<MichaelRaskin>
(not perfect coverage, but might provide useful information when aggregated over a couple of months)
<samueldr>
ike is likely rocking a 6100 series opteron (if it is indeed still a dell r815 server) which "All models support: MMX, SSE, SSE2, SSE3, SSE4a, Enhanced 3DNow!, NX bit, AMD64, Cool'n'Quiet, AMD-V (SVM & Rapid Virtualization Indexing), HT-Assist"
<samueldr>
so not sure what feature could be lacking other than the later SSSE4 instructions I guess
<samueldr>
or whatever the right name for those :/
* samueldr
wonders how this could be verified
<infinisil>
MichaelRaskin: I think the intersection of { stuff ofborg built } and { stuff hydra built } is pretty small, considering that they don't "agree" on a nixpkgs version to build
<MichaelRaskin>
infinisil: you, disbeliever! (in staging)
<MichaelRaskin>
I mean, most packages survive whole _weeks_ without a rebuild on master
<infinisil>
Hmm
<MichaelRaskin>
So if ofBorg built something just before merge, then a post-merge state got into Hydra queue inside a couple of days, there is a large chance of intersection
<MichaelRaskin>
And if not, there is a large chance that reproducibility is not the issue anyone is currently interested in!
<infinisil>
Yeah I guess you're right
ekleog has quit [Quit: back soon]
ekleog has joined #nixos-dev
asymmetric has joined #nixos-dev
<gchristensen>
infinisil: you can restart jobs now