<samueldr>
in relation to this discussion, does any nix or nixos CLI tool use an argument which *has* to be used like -Xyz where -X is the argument and yz is the value
<gchristensen>
don't think so samueldr
<samueldr>
this makes things easier then
lassulus_ has joined #nixos-dev
lassulus has quit [Ping timeout: 260 seconds]
lassulus_ is now known as lassulus
<thoughtpolice>
So I pretty much hate Discourse instantly already.
<thoughtpolice>
Is it possible to remove the "safety switch" on Discourse that prevents horrible, no good, useless noobs like myself from...... writing a post containing more than 2 hyperlinks in it?
Jackneillll has joined #nixos-dev
Jackneilll has quit [Ping timeout: 240 seconds]
<niksnut>
thoughtpolice: I guess that's an anti-spam thing?
<gchristensen>
and of course, if you can pwn https://nixos.org/nix/install, you can pwn the hash we put in that page, so it really doesn't buy you anything
goibhniu has quit [Ping timeout: 260 seconds]
goibhniu has joined #nixos-dev
jtojnar has joined #nixos-dev
<manveru>
thoughtpolice: i think zimbatm can raise your trust level for that
<zimbatm>
thoughtpolice: where you prevented from posting at all?
<zimbatm>
there is also a spam scanner that's using Akismet
<zimbatm>
thoughtpolice: I raised you to level 1, let me know if it's still preventing you to post
<manveru>
under admin/site_settings/category/posting
__Sander__ has quit [Quit: Konversation terminated!]
page has joined #nixos-dev
<thoughtpolice>
zimbatm: It's not preventing my posting, it's just preventing including anymore than two hyperlinks in a post. I had 3, so I just removed one (it was redundant).
<thoughtpolice>
Which is a very bad, pointless anti-spam restriction, if that's its intention as niksnut suggested. If you already have Akismet I'm not sure how much this buys us honestly, and it mostly seems to be a pointless restriction that would just frustrate people.
<thoughtpolice>
Or maybe it was intentionally designed to frustrate people and nothing more, no anti-spam about it. That's also a possibility :P
<gchristensen>
I think it goes away as you post more
<gchristensen>
but if you're a brand new user and post a bunch of links, that seems weird, right?
<thoughtpolice>
... 3 links? Really? I was just referencing GitHub issues. Also you could, maybe, maaaaybe make a weak arguement that prevalent/regular developers will register early, get past this annoyance quickly (because their "new user" status has nothing to do with their actual experience), and thus make it more effective for newcomers to be "filtered" in case their posts are "bad".
<thoughtpolice>
Honestly it's kind of mind boggling to me that the internet has become so damn hostile that even hyperlinking -- quite literally, a foundational aspect of the WWW -- is even viewed suspiciously. What a hell we live in.
<gchristensen>
:(
<cbarrett>
Thanks Google
<cbarrett>
:)
<thoughtpolice>
But anyway, I don't buy that argument of course! Let the people post hyperlinks, I say!!
<niksnut>
thoughtpolice: spam basically destroyed the old wiki, so being careful about allowing hyperlinks seems reasonable to me
<thoughtpolice>
I don't buy that for a second. Forums are not wikis, first off, so the dynamic of shitposting/spamming is totally different. You see a forum index every day, and get alerted on topics. Nobody watches "Changed History" on a wiki page. It's hard to miss. Second, you can make a number of easier restrictions like "Require GitHub OAuth", which would likely be a pretty good start to help leverage GitHub's anti-spam.
<gchristensen>
well hopefully it doesn't cause many actual problems in actual use over time, I'm sorry you got bit by it.
<simpson>
At scale, with enough users, people watch the history of their pet pages. This happens on WP and some Wikias.
<niksnut>
thoughtpolice: of course forums can be destroyed by spam, see Usenet
<thoughtpolice>
What does usenet prove at all lol. All I'm saying is vaguely justifiable "anti-spam" features that are actually user-hostile just piss people off and don't actually solve anything as far as I can see. Is there literally any support for the fact that implementing 2-link posting restrictions meaningfully denies spammers? On Discourse? For a community of our size?
<gchristensen>
I don't think I've seen many people have issues with the limitations new users are under?
<gchristensen>
I'm happy to defer the thinking and problem solving of anti-spam to Discourse, even if it isn't totally what I'd do
<thoughtpolice>
Yeah but come on -- we defer the thinking and problem solving of a lot of problems to a lot of people, and they still fuck it up and we have to deal with it. We are, quite literally, human package managers. I don't buy this either, you're always responsible for your own home.
<thoughtpolice>
Look, if the basic message here is "thoughtpolice, sorry that happened, it sucks, it's fixed for you, but nobody cares enough to fix it because it's not really a problem", that's fine. But we don't have to pretend user-hostile anti-spam features with seemingly little realistic basis are, like, a massive cornerstone of our community's security.
<gchristensen>
well, the people who have worked to set up the discourse and are working to improve the Nix community have good intentions and I would like to ask you to keep that in mind, and approach them with good intentions. your hostility is quite abusive I think, and I think not welcome
Sonarpulse has quit [Ping timeout: 276 seconds]
<thoughtpolice>
Trust me, I led the development of a large open source project before, so "expressing any minor amount of displeasure" being considered "hostile" from afar is par for the course, for me. I don't think it's hostile to suggest meaningless restrictions are dumb. I'm not suggesting discourse is bad, I'm suggesting that it is annoying in a particular way, and your arguments don't cut it.
<thoughtpolice>
In fact, I quite literally just said if, the reality is, that I'm just not that important -- that's fine! I'm ok with that. I do a lot of stuff around Nixpkgs I don't always like. That's part of the deal.
<thoughtpolice>
In fact between GG and Discourse, I like Discourse for a couple reasons (including GG's godawful account/SMTP management with forwards). But I'm not going to, like, pretend "oh wow i'm super glad this annoying thing was totally in my way, in case, uhhhhhh, reasons happened"
<gchristensen>
ok, great :)
<gchristensen>
that does suck, I am sorry it happened to you, I think it isn't a big problem
<simpson>
It's a massive problem and the Discourse should be reconfigured immediately, but societal norms tell us that we can sweep anything under the rug and we should look for somebody to blame.
<thoughtpolice>
To be fair though, I'll admit my tone is not always particularly easy to read (and not everyone in the community is a born english speaker, to make that harder). So, not like I'm _really_ mad or anything, but yeah I can get how it can come off wrong.
<gchristensen>
<3 I understand
<simpson>
The configuration of the Discourse, of any of our management software, dictates privileges for users, which in turn dictates the shape of our organization, which in turn (Conway's Law) dictates what kind of code we produce.
<simpson>
So badly-configured Discourse leads to bad stuff in nixpkgs.
<simpson>
Anyway, just don't blame anybody (or blame me, I don't give a fuck) and reconfigure it so that trusted contributors aren't subject to the spam filter.
* simpson
failed conflict-avoidance class
<gchristensen>
my understanding is this limit goes away once you've posted a few times
<copumpkin>
niksnut: you have anything planned for my "content-addressable" binary cache issue or is it esoteric enough that I should just fix it on my side and we assume that nobody else will encounter it? Otherwise I'll just scan my S3 bucket (only a few tens of thousands of .narinfos :P) and remove CA from anything with non-empty references and close the ticket, but it doesn't seem ideal if it'll bite other people too
Sonarpulse has joined #nixos-dev
page has quit [Ping timeout: 240 seconds]
<thoughtpolice>
singletons 2.4.1 is broke again on ghc 8.4.3 :(
<thoughtpolice>
peti: For minor 'trivial' fixes like this (using `dontCheck` in configuration-ghc-8.4.x.nix in the mean time, submitting a singletons bug) are you ok if i push to master?
<peti>
thoughtpolice: Sure, no problem.
<thoughtpolice>
peti: Criterion also needs a touchup (bump to 1.4.1.x for 8.4, bounds-related for `base-compat`) which I can take care of too
<peti>
thoughtpolice: Yes, that would be nice. Just hack away. The more people contribute fixes, the better for everyone.
<niksnut>
copumpkin: yeah, might be best to fix your .narinfo files
<niksnut>
I guess we could ignore the CA field, but it seems potentially risky to ignore security-related fields
<contrapumpkin>
niksnut: so "if has CA field and non-empty References, delete CA field" would be sensible logic here?
<niksnut>
although, otoh, probably nothing bad can happen here
<niksnut>
contrapumpkin: the problem is that we *don't* want to ignore non-empty references in the future
<contrapumpkin>
would the "solution" be to outlaw all references from FO derivations?
<contrapumpkin>
not a huge fan but I could sort of see that making sense
<contrapumpkin>
or are you just planning to change the hashing scheme to factor them in?
<contrapumpkin>
so about fixing my narinfo files, what would you do instead of what I said?
<niksnut>
what specifically? you said many things :-)
<contrapumpkin>
[14:24:59] <contrapumpkin>niksnut: so "if has CA field and non-empty References, delete CA field" would be sensible logic here?
<contrapumpkin>
sorry :)
<contrapumpkin>
I'm just trying to get my binary cache not to break my clients
<contrapumpkin>
doesn't seem like emptying th references is sound
<niksnut>
do you mean in nix or in a script to fix your existing .narinfo files?
<contrapumpkin>
my one-off script that I'll run against my S3 bucket
<contrapumpkin>
just trying to do the least shitty thing that will fix the bug
<niksnut>
in that case, yes
<contrapumpkin>
and anything created by nix 2.0.2 and above won't have the CA field in there, right?
<contrapumpkin>
so I will only have to do this once
<niksnut>
yes
<contrapumpkin>
cool, thanks
<LnL>
jtojnar: around?
<jtojnar>
LnL: yep
<LnL>
ok, so from what I can tell the issue is reproducible with your example (needs some small tweaks) but it's only a problem when meson is used
<jtojnar>
LnL: could you also try with gdk-pixbuf master?
<LnL>
sure, don't think I tried master
<jtojnar>
master dropped autotools support so they might have fixed it
<jtojnar>
LnL: also are you aware that the presence of shared-mime-info during build enables the detection?
<jtojnar>
so they should not be there
<LnL>
no, I don't know anything about pixbuf :)
<LnL>
doesn't work on master, autogen.sh was removed and it doesn't build without that
<jtojnar>
LnL: yep, they dropped autotools on meson
<jtojnar>
s/on/for
<LnL>
make: *** No rule to make target 'stamp-gdk-pixbuf-marshal.h'
<jtojnar>
LnL: yep, the test is built for autotools, you can add `|| true` to the make things
<LnL>
oh whoops, didn
<LnL>
't disable enough
MichaelRaskin has joined #nixos-dev
<gchristensen>
pierron: given `pkgs = import <nixpkgs> {};` is there a way to get a pkgs' which has included a new overlay?
<gchristensen>
use case is I have a very specific pkgs available to me, which I didn't import, and I need to access one or two packages from inside an overlay
<gchristensen>
if not, I will have to make this overlay not an overlay :(
<LnL>
jtojnar: still the same problem
<jtojnar>
LnL: does it build without shared-mime-info?
<LnL>
oh hold on, this is without the patchs since one of the others didn't apply
<jtojnar>
LnL: the patches should be only needed on Linux
<jtojnar>
LnL: could I get an access to the darwin box? I would like to debug it further
<LnL>
not sure what you're referring to then
<LnL>
I can't portforwarding of my isp is weird and doesn't work properly
<jtojnar>
LnL: shared-mime-info should not be in buildInputs
<jtojnar>
otherwise it will be built with sniffing
<LnL>
it isn't, unless something propagates it
<jtojnar>
maybe ask ebassi on #gnome-hackers on irc.gnome.org
<jtojnar>
I got to run
<LnL>
but maybe there's something global that gets detected?
<LnL>
jtojnar: anyway, we can bump it to 2.36.12 if we keep using autotools for now
<jtojnar>
we should probably solve this now, otherwise the bug will go to 2.38
<gchristensen>
I should erase anything personal from my mac so people can debug on it