<infinisil>
Consider this: A fuse filesystem for youtube videos
<infinisil>
Downloads videos on-demand using youtube-dl
<supersandro2000>
httpfs?
<infinisil>
Hmm not exactl
<samueldr>
infinisil: fun with a thumbnail view
<infinisil>
The idea with this would be that the downloading and extraction of youtube videos would be handled in the filesystem, so the user doesn't need to
<infinisil>
Which also allows integration with many other things, without extra work
<samueldr>
I mean, with a file browser that *produces* thumbnails
<samueldr>
because it'll need to download all videos
<infinisil>
I could point nginx to it, and I'd have an alternate youtube frontend
<samueldr>
(it *could* seek if the fs was good enough)
<infinisil>
I guess you can't list files in such an fs
<samueldr>
infinisil: consider channels and playlists
<infinisil>
Oh yes to that
<samueldr>
they'd be listable
<infinisil>
I guess you'd want to have `youtube/{by-id,by-channel,by-playlist}`
<infinisil>
by-id not being listable
<samueldr>
how to make a channel directory? maybe some trickery by ending a directory with a valid channel ID?
<samueldr>
because you can't list all existing channels either!
<infinisil>
Oh but no by-channel and by-playlist wouldn't be listable either
<infinisil>
But the entries within it would
<samueldr>
so mkdir "Samuel Dionne-Riel UCrMUS_fve7Fpj1DBXJTNhww"
<samueldr>
then with some state save which "subscriptions" were made
<samueldr>
(you could also add nothing in front)
<infinisil>
Wait what would mkdir ... do?
<samueldr>
infinisil: wrong channel I think
<samueldr>
(I thought it was -chat)
<infinisil>
Oh damn!
<infinisil>
I thought so too lol
supersandro2000 has quit [Disconnected by services]
supersandro2000 has joined #nixos-dev
bennofs__ has joined #nixos-dev
jonringer has quit [Ping timeout: 260 seconds]
<hexa->
gchristensen: hail hydra
bennofs_ has quit [Ping timeout: 240 seconds]
rajivr has joined #nixos-dev
<jtojnar>
latest version of pulseffects no longer supports pulseaudio (only pipewire), do you think it is okay to upgrade the pulseeffects attribute to the latest version and have the pulseaudio version under pulseeffects-legacy, even though pulseaudio is still the default?
justanotheruser has quit [Ping timeout: 272 seconds]
<{^_^}>
#112591 (by ymeister, 3 days ago, open): pulseeffects: 4.8.4 -> 5.0.0, pulseeffects-legacy: init at 4.8.4
<jtojnar>
on one hand, it is nice to have the unqualified attribute pointing to the latest version
<jtojnar>
on the other, people who still have not switched to pipewire will have to manually switch to pulseeffects-legacy
<worldofpeace>
Jan Tojnar: so pulseaudio != pipewire-pulse?
<worldofpeace>
or oh it's just pipewire server but the pulseaudio plugin
<jtojnar>
pipewire-pulse mimics the pulseaudio server API
<jtojnar>
but pulseeffects switched to native pipewire
<jtojnar>
so it cannot work with pulseaudio any more
<worldofpeace>
... they did put as in a weird position. I just tried to logic my way out of this issue and it didn't really work. I guess what you've suggested is the only real way to do it because of how they choose to isolate the legacy version
<worldofpeace>
I wonder if 4.8.4 vs 5.0.0 was a massive rewrite
<worldofpeace>
then it would make sense to leave pulseeffects the project as legacy, and then make a new project rebranded to support pipewire. even the name feels weird now
<jtojnar>
well, the other option would be having pulseeffects5 and pulseeffects attributes
<jtojnar>
the changes were not that big, most of the processing was handled by GStreamer so they only needed to swap some pipeline elements
<infinisil>
How about deprecating pkgs.pulseaudio, replacing it with pkgs.pulseaudio-legacy and pkgs.pulseaudio-pipewire
<worldofpeace>
Jan Tojnar: really, then maybe it would've been possible to have just supported both for a time
<jtojnar>
not worth the effort from what I heard
<jtojnar>
there was a bit of server specific code as well (though not as much one would think)
<samueldr>
because of the issue with hydra... hard to say
<danderson>
I seem to have missed the Issue With Hydra. Guess I'll go back in history
<samueldr>
I mean, the data is in hydra, nowhere else, short of building the `tested` jobset
<danderson>
right, I mean I wasn't aware that hydra was having a bad time more generally
<danderson>
I guess the DB connection error could have been a clue, in hindsight.
<siraben>
reading that thread be like o.O
<supersandro2000>
Could we make substituteInPlace fail if a pattern wasn't found? I think that would make updating a bit easier and you would notice earlier that something does not match up.
<{^_^}>
nix#4463 (by jtojnar, 3 weeks ago, open): nix-env: Use pname and version attrs in -q
<andi->
that ties nix to nixpkgs. Not sure if we want that. It already has a bunch of coupling that makes using anything *not* nixpkgs harder.
kcalvinalvin has quit [Quit: ZNC 1.7.4 - https://znc.in]
kcalvinalvin has joined #nixos-dev
orivej has joined #nixos-dev
<jtojnar>
it does not have to be nixpkgs-specific – other repos can use pname and version attributes too
<jtojnar>
or they do not have too, in which case it will work as before
cole-h has joined #nixos-dev
AlwaysLivid has quit [Ping timeout: 268 seconds]
kraem has quit [Ping timeout: 246 seconds]
kraem has joined #nixos-dev
rajivr has quit [Quit: Connection closed for inactivity]
kraem has quit [Read error: Connection reset by peer]
kraem has joined #nixos-dev
kcalvinalvin has quit [Quit: ZNC 1.7.4 - https://znc.in]
kcalvinalvin has joined #nixos-dev
justanotheruser has joined #nixos-dev
cole-h has quit [Ping timeout: 272 seconds]
<rmcgibbo[m]>
(moving discussion from -chat)
<rmcgibbo[m]>
supersandro2000 : i think you've already written that code in bash in your checks repo, but we'd just have to re-arrange it a little bit to put it into the bot so that i could call it after the build is finished in the same way i call hammering?
<supersandro2000>
using the build log is actually a better idea
<supersandro2000>
that way the build log is also available when the package was already build
<rmcgibbo[m]>
That's also the place to put the "ran 0 python tests" check too.
<supersandro2000>
yeah. Then I wouldn't need to run nixpkgs-review for python things
<supersandro2000>
not sure how to check the outputs with that
<supersandro2000>
also the stale substitueInPlace could go there
<rmcgibbo[m]>
If the structure is that it's a binary that takes as input the /nix/store directory/directories produced by a single input drv after the build finishes, it can internally run `nix log {store-dir}` to get the logs and then grep them, and also look for /lib and /include that are part of the same store directory, right?
<supersandro2000>
I think we need to add the path of nixpkgs
<supersandro2000>
nix log -f nixpkgs wget2
<supersandro2000>
how do we get the store directory?
<supersandro2000>
I can try to hack something up for review-checks and then we can port it cleanly to hammering
<supersandro2000>
eventually I can just remove most of the stuff from my script and it just runs hammering
<rmcgibbo[m]>
How about I start by making the "outer interface" of how the binary gets called and what kind of output it should produce, since I know how my bot works, and then once that's there you can fill in the details?
<rmcgibbo[m]>
I think it's better that it's separate from hammering since hammering works fine pre-build and this only will work post-build.
<supersandro2000>
yeah true
justanotheruser has quit [Ping timeout: 264 seconds]
<jtojnar>
I think it is fine to add build mode for hammering
<jtojnar>
already had this check in mind (qualified by include directory getting big)
<jtojnar>
we also want to add dev output when there are propagated-build-inputs (which is in theory possible to do in eval checks by traversing the inputs transitively but checking $out/nix-support/propagated-build-inputs is probably saner)
WilliButz has quit [Quit: bye]
<supersandro2000>
I am just hacking that together in shell and then we go from there
<rmcgibbo[m]>
:+1:
WilliButz has joined #nixos-dev
<rmcgibbo[m]>
i unfortunately am wasting my day trying to upgrade Cbc (mixed integer programming), abseil-cpp, google or-tools which all have failed to build in multiple creative and diverse ways.
<supersandro2000>
yesterday I started with wget2
<supersandro2000>
it has ssl support now
<supersandro2000>
rmcgibbo[m]: do you have the logs from nixpkgs-reviews available?
<supersandro2000>
I just fixed that already build packages are missing the logs and I would just reuse that
<rmcgibbo[m]>
Not sure I understand the question. There's still a bug in r-rmcgibbo that it fails to find to upload the logs to gist.github.com if the drv fails because one of its dependencies (like src or fetchpatch) failed. I haven't fixed that bug yet.
<supersandro2000>
I don't mean this bug
<supersandro2000>
I mean when you get a bunch of lines like when doing the second run
<supersandro2000>
nixbuild log of '/nix/store/g3328kr658gkvc2230dv6ay7gw6ac94j-perl5.32.0-wakeonlan-0.41.drv' is not available
<rmcgibbo[m]>
Once we have the logic written out in shell or whatever we can figure out the best way to put it into hammering so that it fits in nicely.
justanotheruser has joined #nixos-dev
pmy has quit [Ping timeout: 272 seconds]
pmy has joined #nixos-dev
justanotheruser has quit [Ping timeout: 260 seconds]
red[evilred] has joined #nixos-dev
<red[evilred]>
Did we end up finding a 21.05 release manager?
<supersandro2000>
I don't think so or it wasn't announced yet
<red[evilred]>
Same, I just wondered if it ended up being discussed somewhere else
<supersandro2000>
I think jonringer talked with someone about it but I can't fill you in with the details
justanotheruser has joined #nixos-dev
__monty__ has quit [Quit: leaving]
<supersandro2000>
jtojnar rmcgibbo[m] I've just opened https://github.com/NixOS/nixpkgs/pull/113062 with some examples more example paths so that people better know when they should add outputs
<{^_^}>
#113062 (by SuperSandro2000, 57 seconds ago, open): doc: update multiple output section with more example paths