<{^_^}>
eadwu's karma got increased to like 1, I think
<cole-h>
siraben: I wonder how it should've been caught if nothing in Nixpkgs is using stdenv.lib anymore...
<cole-h>
;)
<siraben>
lol, true.
<cole-h>
Sounds like you need to PR nix-darwin to remove stdenv.lib next :D
<hexa->
error: attribute 'allowAliases' missing, at /etc/nixos/nixpkgs/pkgs/stdenv/generic/default.nix:156:16
<cole-h>
(That's the location of the lib "alias")
<cole-h>
s/lib/stdenv.lib/
<cole-h>
wait
<cole-h>
waaaaaaaaaaaaaaait
* cole-h
facepalms
<cole-h>
sorry hexa-
<abathur>
could use brains from anyone w/ a good grasp of nix-daemon user/group uid/gid issues; macOS users have been landing in recovery during Big Sur updates and I *think* the cause is user migration it is doing. It goes OK if I delete the nixbld users, or if I rename them to _nixbld* and give them uids in the 200-400 range
<abathur>
in one sense the fix seems fairly straightforward, but it also seems a little _bold_ for default installs to eat up 32 UIDs in the ~200-UID range available for role accounts
<siraben>
cole-h: wow, thanks
<ajs124>
allowAliases defaults to true everywhere else in nixpkgs, though
<ajs124>
ah, no. not everywhere, sorry.
<ajs124>
1/9 is true.
<cole-h>
Well
<cole-h>
If it's unset
<cole-h>
I feel it's fair to assume false
<ajs124>
Don't get me wrong, we build all our systems with allowAliases = false, but my impression has always been that they are enabled by default
<cole-h>
OK, I see what you're saying.
<cole-h>
`rg config.allowAliases or` -> most are `or true`.
<cole-h>
btw, is stdenv.lib slated for deprecation (as the comment says) or *removal* in 21.11?
<cole-h>
cuz it's already kinda deprecated by virtue of having a builtins.trace saying so... :P
<cole-h>
cc siraben ^
<siraben>
cole-h: removal
<cole-h>
Well, if somebody decides to PR the condition to `or true`, they should probably fixup that comment, too :D
maljub01 has quit [Quit: maljub01]
maljub01 has joined #nixos-dev
jonringer has quit [Remote host closed the connection]
<cole-h>
FYI, I did ^
supersandro2000 has quit [Ping timeout: 265 seconds]
<siraben>
I'll keep the PR open until that is fixed then.
pmy_ has quit [Read error: Connection reset by peer]
pmy_ has joined #nixos-dev
<yorick>
I'm afraid it'd coderot before anyone fixes this
<yorick>
(unless you want to take a shot)
<siraben>
Do packages that have makeWrapper in buildInputs cross-compile correctly atm?
<yorick>
yes
<siraben>
Then I should make the change in the other direction and put makeWrapper in buildInputs?
<adisbladis>
I don't think so
<yorick>
um, I think the channel morally disagrees :D
<gchristensen>
siraben: not that I know of
<yorick>
all the packages that have makeWrapper in nativeBuildInputs are currently broken, but everyone doing serious cross-compiling has a bunch of patches on top of their nixpkgs fork anyways
<siraben>
So the solution is to fix makeWrapper then, hm.
<yorick>
but fixing makeWrapper would be ideal
<siraben>
working with cross-compilation would easier if we had cached cross-compilers...
<yorick>
cross-compilation would be easier if it were tested on hydra at all
<yorick>
how does it manage not to show up in the cache then?
<gchristensen>
perhaps you are not using the same cross builds it is producing
<yorick>
also, hrm, that's a lot of evaluation errors since 4 days ago
<adisbladis>
I think probably because it's a separate jobset, so it doesn't match nixos channels very well
<gchristensen>
,evalerrors = If you see a jobset suddenly started having a lot of evaluation errors around February 1, we only started remembering evaluation errors around then. The errors have probably "always" been there, but we didn't record them.
<{^_^}>
evalerrors defined
<gchristensen>
,evalerrors yorick :P
<{^_^}>
yorick :P: If you see a jobset suddenly started having a lot of evaluation errors around February 1, we only started remembering evaluation errors around then. The errors have probably "always" been there, but we didn't record them.
<yorick>
aha!
<gchristensen>
,evalerrors = If you see a jobset suddenly started having a lot of evaluation errors around February 1, we only started remembering evaluation errors around then. The errors have probably "always" been there, but we didn't save them for every evaluation: we'd only keep the evaluation errors for the most recent evaluation.
<{^_^}>
evalerrors redefined, was defined as If you see a jobset suddenly started having a lot of evaluation errors around February 1, we only started remembering evaluation errors around then. The errors have probably "always" been there, but we didn't record them.
kini has quit [Remote host closed the connection]
<siraben>
yorick: how would you recommend to fix makeWrapper?
kini has joined #nixos-dev
<yorick>
you could try targetPackages.runtimeShell, I think that should work with your use
<yorick>
but it has the side-effect of probably breaking if it's in buildInputs
<siraben>
probably breaking even when compiling to A from A?
<yorick>
maybe someone more clever than me knows the way
<yorick>
(sorry)
<rnhmjoj>
siraben: did you see my reply on issue #110736?
Raito_Bezarius has quit [Ping timeout: 260 seconds]
pmy_ has joined #nixos-dev
<yorick>
thanks!
dhess has quit [Remote host closed the connection]
dhess has joined #nixos-dev
toddgamblin has joined #nixos-dev
pmy_ has quit [Read error: Connection reset by peer]
<eyJhb>
x-post from #nixos, regarding the /nix/store. Is there any method to harden /nix/store ? ie. I thought about disallowing the ability to list the content of /nix/store, for anyone except trusted users
pmy_ has joined #nixos-dev
<zimbatm>
eyJhb: it wouldn't really help as the store paths are also recorded in /nix/var/nix/db/db.sqlite
<eyJhb>
zimbatm: But is access to that file required for anyone except trusted users, etc. ?
<zimbatm>
not sure
<eyJhb>
Ie. does php/apache/whatever really need access to that
<zimbatm>
not them but it might be needed to the non-trusted nix users
<qyliss>
could be group-readable by them
<zimbatm>
s/to/by
<zimbatm>
true
<eyJhb>
Just thinking, a small change like that could be "huge" :)
<zimbatm>
another vector is to follow all the store paths starting from /run/current-system
<eyJhb>
Damn it, yeah
<zimbatm>
sorry for being contrarian...
<eyJhb>
How dare you zimbatm. Poking holes in everything :D
<jtojnar>
Mic92: *FlagsArray should just never be used in Nix
<eyJhb>
back to the drawing board I guess... Or rather, exam, but would be cool!
pmy_ has quit [Read error: Connection reset by peer]
orivej has joined #nixos-dev
<zimbatm>
ok, to your side, hiding the /nix/store would still be useful to some extent
<zimbatm>
assuming that the db is not needed by the client
<zimbatm>
it could prevent a developer from accidentally leaking a secret during build
<zimbatm>
let's say they nix-build a repo that uses git-crypt secrets, referenced by the nix code
<zimbatm>
btw, these ideas have been discussed in the past. We really need to group them somewhere so they are easy to find.
<eyJhb>
supersandro2000: Well, then the comment still stands. And not sure about doCheck = false; when there are no tests, is there any agreed upon convention about that? Not really something I would put into a update commit
<Mic92>
zimbatm: Performance is also a hold-back. You need to avoid double caching and metadata lookup is a lot more expensive.
<supersandro2000>
I wouldn't find anything so I just don't attempt to search something
<supersandro2000>
also python tests are usually whatever and only find bugs in their code or things not working in the sandbox. import check is usually enough.
<Mic92>
jtojnar: mhm. Why did we not get bit by it earlier?
<eyJhb>
There is a lot of usual, idk, usual in this, seems like it is based on nothing, with no discussions, etc.? What is the reason to remove it for every python module, even if it does nothing at this point?
<jtojnar>
Mic92: we are getting bitten by this all the time
<Mic92>
jtojnar: I mean it worked reasonable well for buildGoPackage. One needs to use one string
<jtojnar>
but it is hard to notice since some programs just ignore invalid arguments
<jtojnar>
Mic92: yeah, it works by chance if you have just a single argument
<jtojnar>
which is probably the most common case
{^_^} has quit [Remote host closed the connection]
{^_^} has joined #nixos-dev
<supersandro2000>
eyJhb: redirect your discussion at @jonringer. I just copied it from him and he knows what he is doing.
<eyJhb>
supersandro2000: It would be nice when you are sending these automated messages out, so that it can be seen why it should be done like so. Which seems like a pretty easy step before spreading that test method wide and far :)
<eyJhb>
jonringer is only on Matrix, right?
<supersandro2000>
or discord
cole-h has joined #nixos-dev
<supersandro2000>
I probably knew it at some point and just forgot it.
<eyJhb>
I will link anything if I find it :)
<eyJhb>
Ohh frick, this is my first big Matrix room. I think my server might die.
<qyliss>
jonringer is often on IRC
pmy_ has quit [Read error: Connection reset by peer]
<qyliss>
but not always
<eyJhb>
Couldn't find them in #nix on Matrix, so trying #nixos :| Not sure if that is wise.
pmy_ has joined #nixos-dev
<eyJhb>
Well crud, I cannot join #nixos because error, and I cannot leave #nix because of error
<qyliss>
eyJhb: if I were you I'd just ,tell them here
<eyJhb>
qyliss: best way to do that while they're offline?
<zimbatm>
Mic92: doesn't FUSE allow to forward some of the syscalls to the underlying OS?
<qyliss>
eyJhb: the bot has a ,tell command
<qyliss>
,help tell/
<qyliss>
,help tell
<qyliss>
i don't actually know how you ask it for help
<qyliss>
but it works like this:
<qyliss>
,tell eyJhb hello
<eyJhb>
qyliss: the bot is DOWN for maintanence :p
<qyliss>
oh :(
<cole-h>
haha
<qyliss>
then never mind :(
<eyJhb>
infinisil and gchristensen is doing surgery on it in -chat :p
<cole-h>
The amqp message queue is being migrated to a different host atm :P
<eyJhb>
Guess I will just continue reading up on exams til then :p
pmy_ has quit [Read error: Connection reset by peer]
orivej has quit [Ping timeout: 240 seconds]
<V>
eyJhb: use MemoServ
<eyJhb>
Ahh true!
<qyliss>
omg how did I not know about MemoServ
<V>
IDK
<V>
I know all the obscure IRC facts
<hexa->
also consider memoservs ability to deliver memos as emails
{^_^} has quit [Remote host closed the connection]
{^_^} has joined #nixos-dev
pmy_ has joined #nixos-dev
<infinisil>
,tell
<{^_^}>
Use `,tell john Remember to do the laundry` to send this to john next time he's talking in this channel
<infinisil>
eyJhb: Back online :)
<eyJhb>
Perfect, thanks infinisil !
<eyJhb>
( and gchristensen )
<Valodim[m]>
eyJhb: is your name the beginning of a base64 encoded jwt token?
<eyJhb>
`{"alg`, so yeah, pretty much!
pmy_ has quit [Read error: Connection reset by peer]
<gchristensen>
lol
<gchristensen>
incredible
<eyJhb>
gchristensen: Sup? :p
pmy_ has joined #nixos-dev
cole-h has quit [Ping timeout: 240 seconds]
rajivr has quit [Quit: Connection closed for inactivity]
pmy_ has quit [Read error: Connection reset by peer]
pmy_ has joined #nixos-dev
xwvvvvwx has joined #nixos-dev
xwvvvvwx has quit [Remote host closed the connection]
xwvvvvwx has joined #nixos-dev
pmy_ has quit [Read error: Connection reset by peer]
pmy_ has joined #nixos-dev
pmy_ has quit [Read error: Connection reset by peer]
orivej has joined #nixos-dev
pmy_ has joined #nixos-dev
pmy_ has quit [Read error: Connection reset by peer]
<supersandro2000>
I have zero tech modifications in my body
jonringer has joined #nixos-dev
pmy_ has quit [Read error: Connection reset by peer]
pmy_ has joined #nixos-dev
pmy_ has quit [Read error: Connection reset by peer]
pmy_ has joined #nixos-dev
<Mic92>
supersandro2000: could you surive without your computer so?
<Mic92>
mydra looks cool!
mkaito has joined #nixos-dev
mkaito has joined #nixos-dev
mkaito has quit [Changing host]
pmy_ has quit [Ping timeout: 256 seconds]
pmy_ has joined #nixos-dev
<jtojnar>
do you remember a discussion of what is expected of maintainers?
<jtojnar>
I am pretty sure it was discussed at some point but cannot find anything
<jtojnar>
well, maybe it was more of a discussion of personal interpretations of the “maintainer” term
<eyJhb>
jonringer: AHA! Got you. I had a "discussion" with supersandro2000 regarding setting `doCheck = false;" on Python packages that have no tests. Do we have a issue or anything on why this is done?