<infinisil>
sphalerite: Well people seem to like the idea of adding something to the installer, but I have never ever touched the installer
<infinisil>
Maybe it's not too hard to add though, I just don't know the right place right now
<infinisil>
And I'm really busy with university :)
<sphalerite>
fair enough
orivej has quit [Ping timeout: 264 seconds]
MichaelRaskin has joined #nixos-dev
<gchristensen>
niksnut: it looks like you might be preparing for a release. I heard a bug report on the 2.0 maint branch on new the installer not switching properly that I'm going too look at in a couuple hours
<__Sander__>
it seems to work without any other changes from master
<niksnut>
__Sander__: thanks, done
<__Sander__>
niksnut: cool
<__Sander__>
then we also have a "streamlined nix experience" on cygwin
<zybell_>
now add it to the cygwin installer
ma27 has quit [Ping timeout: 276 seconds]
contrapumpkin has joined #nixos-dev
orivej has quit [Ping timeout: 248 seconds]
orivej has joined #nixos-dev
__Sander__ has quit [Quit: Konversation terminated!]
orivej has quit [Ping timeout: 256 seconds]
orivej has joined #nixos-dev
ma27 has joined #nixos-dev
<gchristensen>
niksnut: w.r.t. the reversion of the multi-user Nix installer on Linux, you're right. what if we added a --multi-user flag which shunted it over to the multi-user flag if it is passed, but otherwise defaulted to single-user on Linux?
<gchristensen>
would that be an acceptable scenario to backport to 2.0-maintenance?
<gchristensen>
if not, I won't be crushed, but I'll be eager for a 2.1 :D
<thoughtpolice>
Oh no! I made a premature mistake
<thoughtpolice>
or, well, a mistake
<thoughtpolice>
gchristensen: I pulled in https://github.com/NixOS/nixpkgs/pull/39156 on 17.09 on accident, I misread the PR as a normal, bog standard update. However most stuff doesn't depend on oraclejdk -- does borg actually evaluate unfree things anyway?
<gchristensen>
evaluate yes
<thoughtpolice>
FWIW the set of things changed by oraclejdk is typically pretty minimal IIRC, so I think pulling them in anyway is mostly fine.
<thoughtpolice>
especially since they're unfree anything would rebuild anyway
<thoughtpolice>
gchristensen: thank you!
<gchristensen>
thank _you_ for merging :P I totally forgot to come back to that PR
<LnL>
gchristensen: I really like the idea of --multi-user
<gchristensen>
--daemon :)
<gchristensen>
(I'm trying to get away from "multi user" since I don't see that as the primary benefit)
<gchristensen>
I knew you would... you've asked for it before, sorry it took so long for me to actually do it
<Dezgeg>
isn't that the only benefit actually?
<gchristensen>
no, not owning your store is a huge benefit
<LnL>
gchristensen: even tho it's better for a number of reasons I think it's good to get new users on board
<Dezgeg>
for accidental deletions I guess? that's true, yes
<gchristensen>
and because a `pip` from nixpkgs by default installs stuff in to /nix/store/..
<gchristensen>
lots of tools try to modify their own install paths by default if they're able to
<LnL>
Dezgeg: there's no reason not to use the daemon if you're serious about nix
<LnL>
but having the installer create a daemon that runs as root is scary for new users
<LnL>
that's the most frequent thing I've received comments about when poking people to install it :)
<dtz>
having a way to uninstall might be good/important
<dtz>
or falling that, mentioning that multi-user installs may be difficult to remove (if that's true)
phreedom has quit [Remote host closed the connection]
<gchristensen>
that var used to be called "$PINCH_ME_IM_SILLY" and it makes no effort to fix broken permissions
<gchristensen>
it sounds tough
<gchristensen>
but worth doing if it isn't
<LnL>
that's one of the things I'm thinking of looking at when I have some extra free time
<gchristensen>
hot tip for hacking on the installer, build the binaryTarball _once_, then edit the produced files for the dev/test cycle, _then_ apply it to the pre-generated files and test. don't rebuild the binary tarball each time
<jtojnar>
chat and borg and darwin could probably stand the same
<MichaelRaskin>
Of all the process bottlenecks, non-authors-PRs do not seem to be a noticeable problem for Nixpkgs
<gchristensen>
haha yeah
<LnL>
yeah, that solves none of our actual struggles
<gchristensen>
honestly not sure what they could do to solve our struggles
<LnL>
true
<MichaelRaskin>
We-ell
<MichaelRaskin>
Allow CI checks to return «no opinion» that doesn't break «all checks passed»?
<MichaelRaskin>
Make check details more discoverable for non-committers?
<MichaelRaskin>
Support line-by-line comments on files even if the default representation of the file is a «rendered» one?
<gchristensen>
can't disagree with any of this
zybell_ has quit [Ping timeout: 265 seconds]
<MichaelRaskin>
Maybe even support «check is planned» notification, so that «ofborg overloaded» and «ofborg down» could look differently…
ma27 has quit [Ping timeout: 265 seconds]
<dtz>
grahamc: just used the multi-user installer on my beefy centos-based server!! Everything worked great, TYVM! :D
<dtz>
well I had some minor problems convincing CentOS that user namespaces are okay but that's not the installer's job :)
<MichaelRaskin>
Well, unless you want to follow the old Chrome for Windows installer logic…
<MichaelRaskin>
(As far as I remember, just visting the Chrome homepage using IE6 automatically downloaded and started Chrome installer, so the user suddenly got the «Install Chrome? Yes/No» dialog)
<MichaelRaskin>
Might be harder on fully up to date CentOS, though
<gchristensen>
dtz: w00t!
obadz has quit [Quit: WeeChat 2.0]
<thoughtpolice>
Hmm, I should try it on my RHEL machine!
obadz has joined #nixos-dev
obadz has quit [Quit: WeeChat 2.0]
obadz has joined #nixos-dev
MichaelRaskin has quit [Ping timeout: 248 seconds]