<drakonis>
when did nix-bundle become available as a plugin/
<drakonis>
?
<drakonis>
rather, why didnt i hear about this?
cjpbirkb1 has joined #nixos-chat
<energizer>
when using computers, i often get an error like 'Permission denied with no context about *which process* was trying to do *what* when it failed on *which file* *at what time* because the file had *these permissions* but needed *those permissions*
<energizer>
so i have to spend 20 minutes figuring it out
<energizer>
can somebody fix this error reporting plz
cjpbirkbeck has quit [Ping timeout: 260 seconds]
<energizer>
error: "Permission denied", gee thanks for the info buddy
<energizer>
that's exactly one level better than "error: there was an error" and somehow that is the standard level of diagnostic information we live with. can't even imagine how many collective years are spent debugging these
rajivr has joined #nixos-chat
cjpbirkbeck has joined #nixos-chat
cjpbirkb1 has quit [Ping timeout: 246 seconds]
waleee-cl has quit [Quit: Connection closed for inactivity]
<MichaelRaskin>
energizer: for any group of people small enough to coordinate efficienty, fixing this is more effort than learning strace debugging. So…
<energizer>
MichaelRaskin: so no tools ever get better, qed?
<MichaelRaskin>
strace does!
<MichaelRaskin>
What you ask for being done _well_ requires literally annotating every line of code that could syscall in every application with the information about severity level of the error.
<MichaelRaskin>
A ton of effort with extremely little payoff for everyone who could do it (because these people have already been forced to learn strace, maybe ltrace, gdb, and more)
<MichaelRaskin>
And this is a poorly-scaling effort
<MichaelRaskin>
Not a ton of effort once to fix everything in a large ecosystem
<energizer>
i dont think the work needs to be done at application level
<energizer>
it could be built into languages
<energizer>
and logging frameworks
<MichaelRaskin>
Well, stacktraces are already there in many languages, and sometimes you actually need a real stacktrace, but then it turns out that they are hard to read to users
<MichaelRaskin>
So if you can comfortably read the autogenerated explanations, you are likely to be able to read strace comfortably. And if you need humans to think about annotations, this is too costly
ixxie has joined #nixos-chat
cole-h has quit [Quit: Goodbye]
<eyJhb>
strace is generally i nice thing!
<eyJhb>
ashkitten: regarding not keeping build files/backuping them up, etc. at some point you don't really need to backup git repos as well...
<MichaelRaskin>
Its testimonials claim that its output is in some extreme cases more understandable than the source code, and I even believe it!
<eyJhb>
E.g. Cloning google/google-ctf. I don't need a backup of that :p
<ashkitten>
eyJhb: that assumes i only have git repos that have been pushed
<eyJhb>
Well, I know that is true. But yet again, this is in a case of CTF. So obfuscated sourcecode
<ashkitten>
which is demonstrably untrue
<eyJhb>
ashkitten: Yeah, and that is a thing :p But a scratch place for other repos maybe
<ashkitten>
maybe yeah
<eyJhb>
Or be better than all humans on planet earth and always commit!
<ashkitten>
i should prolly keep those in $HOME/tmp
<ashkitten>
lolno
<eyJhb>
:D
<ashkitten>
hey at least i know how to use git add -p
<MichaelRaskin>
eyJhb: I personally know one person who has rebound TextAdept Save command to also invoke fossil commit
<ashkitten>
wow
<eyJhb>
MichaelRaskin: THat however does not sound THAT nice
<ashkitten>
that's hardcore
<ashkitten>
that person sounds like they could fuck me up
<MichaelRaskin>
If there is something worth a non-auto commit message, they of course do a manual commit on top
<ashkitten>
that person could probably bench press 10 of me
<MichaelRaskin>
Nah
<ashkitten>
lol
<eyJhb>
2 of me MichaelRaskin ?
<ashkitten>
eyJhb: fun fact: for things i forget to commit in a timely manner in my nixos-config, i use git add -p and sometimes manually edit the diffs to create a discrete set of changes that can be reverted
<eyJhb>
I am so tired of this fly...
<ashkitten>
not that i ever bother reverting them...
<ashkitten>
idk why i bother when i see other people with personal repos and the commit messages are garbo
<ashkitten>
i figure if i'm ever mentally healthy enough to participate in capitalism i can just show my commit log to the interviewer and they'll burst into tears, roll out the red carpet, and appoint me queen of the company
<MichaelRaskin>
Now that I think of it, a full-time commit-annotator position for a mid-size team might actually be a good idea. Readable history _and_ a person with an idea what the hell is going on, what's not to like.
<ashkitten>
heh
<ashkitten>
so i'd have to look at other people's crappy code all day and make sense of it enough to write useful commit messages
<MichaelRaskin>
I would assume that the company who has any chance to think in such terms (as opposed to getting rid of their entire QA) would have OK code, just with useless comments and completely random trash in commit messages
<ashkitten>
so i should work for my gf's company
<ashkitten>
(note: any company she's currently at)
<eyJhb>
ashkitten: waaay too readable commits!
<eyJhb>
Also, <rant> why does every google CTF challenge rely on Google stuff and is unable to work without it?! </rant>
<eyJhb>
joepie91: ^ Opening and closing tags, just saying
<{^_^}>
#97383 (by cole-h, 1 week ago, open): "/dev/fd/62: No such file or directory" in init script on nixos-unstable(-small)
<cole-h>
Where this started happening
<eyJhb>
ANd it is a reboot issue thing
<eyJhb>
NIIICE :D
<cole-h>
Yep. `build-vm` to the rescue
<eyJhb>
Put it into a script then, and run the bisect automatically?
<ivan>
maybe something stopped swallowing stderr early
<cole-h>
Wish I had a beefy server to build on :(
<infinisil>
cole-h: You can get one, they're not that expensive if you just use them for a bit!
<eyJhb>
I really do wish, that we could have some community server for "some people" that could be used for building stuff
<adisbladis>
We have that for aarch64
<cole-h>
I mean, I could use nixbuild.net... But the only server is in the EU, atm.
<eyJhb>
cole-h: EU? The best place in the world you mean
<adisbladis>
cole-h: I've used packet.net spot instances sometimes
<eyJhb>
adisbladis: why not x86_64? :p
<adisbladis>
eyJhb: Because no one has sponsored it yet :P
<eyJhb>
adisbladis: I am not sure I can get AAU to sponsor it
<eyJhb>
They don't like NixOS either. They love ansible.
<eyJhb>
Shitty thing
<eyJhb>
adisbladis: 0.5 USD / Hour! Am I rich or something?! :(
<eyJhb>
If I could pay in LEGOs or Duplos maybe
<adisbladis>
eyJhb: You rarely need it for more than an hour or two :P
<adisbladis>
stdenv rebuilds are painful
<eyJhb>
You need to add in the additional hours that gets put on, because I forget to destroy it
<adisbladis>
But as a data point I used the c3.medium.x86 instance type and rebuilt the entire go set in about an hour
<adisbladis>
Oh, actully it was c2.medium.x86
<eyJhb>
You get quite a lot for the price
<infinisil>
I'm just thinking about backup retention schemes
<eyJhb>
Hosh posh, backup, no one needs them infinisil
<infinisil>
And there's different approaches to take, with different use case
<infinisil>
s
<infinisil>
I do need them!
<infinisil>
E.g. if you want to protect against failing devices, you just need to keep the latest version of your data somewhere else, and update this as regularly as possible
<infinisil>
No need to keep multiple versions of your data for that
<infinisil>
But if you want to protect against accidentally removing files you didn't want to remove, only then you need a scheme like "keep every days data for the last week, keep every weeks data for the last month, etc."
<infinisil>
And a third usecase is archiving, where the goal is to keep data for a longer time. For that you just need a single "Make an archive every month, and keep those for X years (or forever)"
lejonet has quit [Ping timeout: 240 seconds]
<infinisil>
So for my backup tool, I'd allow an arbitrary combination of the three
<infinisil>
Maybe there's also other use cases I'm missing right now, which would need a different retention policy
lejonet has joined #nixos-chat
<infinisil>
Also, you probably want to retain the data on different machines for the different use cases. E.g. the backups should be on another disk and a remote one (3-2-1 rule). The removal-protection data can be stored just once on a local disk. And the archives could go to some very cold storage
<joepie91>
props to whoever built that for making it work properly without JS
evalexpr_ is now known as evalexpr
<samueldr>
joepie91: hi
<joepie91>
samueldr: oh, is that your work?
<samueldr>
implementation mine, design someone who's not here external to the nixos community
<joepie91>
aha :)
<philipp[m]>
Looks like search.nixos.org has broken css.
<samueldr>
oh lol
<gchristensen>
and status.nixos.org :)
<joepie91>
nice work
<gchristensen>
really great work
<philipp[m]>
Or is it a browser plugin of mine?
<philipp[m]>
The age old question...
<samueldr>
they just imported the nixos.org css
<philipp[m]>
But the main site looks mighty fine now.
<samueldr>
casualty of the move, they'll eventually be touched up to use the new design
<samueldr>
(being fixed)
<gchristensen>
great work, samueldr and team
<joepie91>
samueldr: one nitpick: it seems the angular background for page headings has a white background color; making that transparent would make the site work a lot better with Dark Reader :P
<philipp[m]>
Yeah, not really breaking anything important anyway :D
<joepie91>
samueldr: oh, same for the angular design bits on the homepage it seems
<samueldr>
joepie91: if you have a proposal to fix that I'm all ears, but it's problematic because of the the way it needs to be implemented
<samueldr>
for the homepage one at least
<joepie91>
how come?
<samueldr>
the titles *could* see that reversed and have the colour part of the image, but doing so would force us to use multiple images for different background colours (something part of the mockups of upcoming redesigned pages)
<samueldr>
for the homepage, because it actually slips on top of the existing horizontal box
<samueldr>
it has to work for quite wide displays
evalexpr has joined #nixos-chat
<joepie91>
samueldr: hm. brain is refusing to parse this atm, maybe better to discuss another time :P
<samueldr>
there are different limitations that made it so adding a white background to cover crimes up was needed
<samueldr>
though this implementation surely is not the *only* way to do it, and things can change if possible or needed
<samueldr>
sometimes other people come up with different strategies to implement something
<joepie91>
samueldr: yeah I can probably think of something when my brain is not refusing to cooperate
<samueldr>
on the homepage, the bottom angle has to cover part of the actual DOM element that has a blue background too, since the angle ends up going upwards on top of the element
<samueldr>
how does darkreader works?
<samueldr>
does it find any background color rules for white and makes it dark?
<drakonis>
it does, yes.
<samueldr>
oooh
<samueldr>
I think I see an option
<drakonis>
it looks somewhat weird with the new page
<samueldr>
instead of baking the white colour in the image, bake it into the pseudo-element I guess
<joepie91>
<samueldr> how does darkreader works?
<joepie91>
dark magic
<samueldr>
haha
<joepie91>
:P
<joepie91>
it does a lot more than just background color rules
<drakonis>
it looks so weird on the new website
<joepie91>
and it gets it right like 95% of the time, just images that contain part of the background are a problem
<samueldr>
yeah, but in this particular instance, is that enough for that?
<drakonis>
no
<samueldr>
hm?
<drakonis>
parts of the website are blue
<samueldr>
yeah well
<drakonis>
i mean
<samueldr>
I meant about the particular issue joepie91 talked about
<drakonis>
oh yeah
<samueldr>
I guess trading where the white is set for that background element might work
<joepie91>
also yeah baking it into a wrapper element should work
<joepie91>
anything that's CSS, it can analyze
<drakonis>
the light blue chunk looks very strange under darkreader :(
<drakonis>
specifically the block for the search bar
<drakonis>
there's a black left to right strip where it should be white
<drakonis>
but why am i saying this, it doesnt help
__monty__ has quit [Quit: leaving]
<joepie91>
samueldr: there doesn't seem to be any preview metadata for the new site? judging from Twitter's failure to render a preview box
<samueldr>
oh, I alway forget that exists
<samueldr>
though I don't think it had beforehand?
<joepie91>
not sure what changed, but previously Twitter rendered at least some box
<joepie91>
now it just renders nothing
<joepie91>
samueldr: oh and another small suggestion: https://nixos.org/features.html has very very little whitespace between the columns, that should probably be increased, now it kinda feels like they blur into each other
<samueldr>
it's not the final content, so your suggestion will probably be never implemented and the actual content instead be used :)
<samueldr>
but I understand
<joepie91>
the spacing between examples on the frontpage looks a lot better
<joepie91>
samueldr: well this is a general design thing, no? not a content thing
<joepie91>
as in, general text column spacing
* samueldr
looks
<samueldr>
since you said the home page is fine
<joepie91>
I suspect the homepage uses a different grid system
<samueldr>
it should be using the same
<samueldr>
but what I think happens is that it gets two times the spacing
<joepie91>
yeah that looks correct
<samueldr>
the built-in gutter (1 time) from the columns, and an additionnal gutter spacing
<joepie91>
on features.html it's only right-spacing
<joepie91>
on homepage it's right and left (and top and bottom) spacing
<samueldr>
joepie91: two times right
<samueldr>
so it's three gutters between even
<samueldr>
the <li> have the built-in spacing of the columns
<joepie91>
oh you're right!
<joepie91>
so actually both places have right-spacing
<joepie91>
but the content additionally has left/right/top/bottom spacing on the frontpage
<joepie91>
but on features.html it does not
<joepie91>
the .clickable-whole provides the additional spacing on the frontpage it seems
<joepie91>
and .whynix has its own special padding rules
<samueldr>
yeah, that's been used to target, but it's not built-in to that component
<samueldr>
right now it's the homepage which has unique styles