0.14 Bevy Jam

[bitter sparrow]

think i like that better

actually that lets you see the current bg color but not the next one

how about this

when transitioning from screen A -> B:

1. spawn screen B underneath screen A, UI interactions disabled
2. fade the opacity of screen A over time
3. despawn screen A once it's fully transparent

as for actually implementing this idk if it's possible

[agile dirge]

too invasive for the template

[bitter sparrow]

yeah fair enough

[slender belfry]

I actually want to rename it to something containing the suffix System. Not because I have any preference here, but because that the name that many libraries expose, so I think there is some level of expectation there

[agile dirge]

honestly, my proposal is too big too

[bitter sparrow]

i feel like there's not really a standard. i've seen Step, Set, and System as suffixes, or even no suffix

sometimes Steps

[slender belfry]

Ah really? Very possible that I've just randomly mostly come accross the ones that do System, then

[bitter sparrow]

let me find some examples rq

https://docs.rs/bevy/latest/bevy/ecs/prelude/trait.SystemSet.html#implementors

dont see step here but i swear ive seen it used

[slender belfry]

I've never seen it personally, but I believe you

[bitter sparrow]

probably 3rd-party crates

[slender belfry]

Lemme check some big ones

[bitter sparrow]

there's this hehe https://docs.rs/avian2d/latest/avian2d/schedule/enum.PhysicsStepSet.html

[slender belfry]

Avian is Set

[bitter sparrow]

StepSet 😄

[slender belfry]

Yeah, but thats a step from physics 😛

[bitter sparrow]

sure

[slender belfry]

See https://docs.rs/avian3d/latest/avian3d/schedule/enum.PhysicsSet.html

Let's see LWIM

https://docs.rs/leafwing-input-manager/latest/leafwing_input_manager/plugin/enum.InputManagerSystem.html

System

[bitter sparrow]

https://docs.rs/bevy_asset_loader/latest/bevy_asset_loader/loading_state/struct.LoadingStateSet.html

[slender belfry]

bevy_mod_picking: Set

lightyear: SystemSet

[bitter sparrow]

https://github.com/search?q=repo%3AIyesGames%2Fiyes_progress systemset&type=code

iyes_progress uses Set and no-prefix

[slender belfry]

So I guess Set is the majority here?

I hate Set is already a data structure

[bitter sparrow]

in the ecosystem it seems yes, in bevy proper it seems System is the majority

[slender belfry]

But eh

[slender belfry]

Kinda interesting split

[bitter sparrow]

yeah i agree with that. i wanted to name something like StateSet

instead i went with StatePattern :p

[slender belfry]

How about we do SystemSet as a compromise?

[bitter sparrow]

at the very least i think a bevy issue to make the suffixes consistent is reasonable? or at least Systems -> System

[bitter sparrow]

too long imo, System is already pretty long

in terms of correctness though i get it. Event is a common prefix for events for example

[slender belfry]

Let's go with Set then? I don't like it much, but if that what the ecosystem mostly uses, I'd like to follow suit

[bitter sparrow]

seems like everyone in the bevy ecosystem was like.. SystemSet is too long. let's remove one of the words 😄

[bitter sparrow]

works for me

in which case it should be AppSet and not Set ofc

[slender belfry]

hehehe

[bitter sparrow]

or something else-Set

[prisma delta]

What do you want to do with the bad / malformed theme suggestions?

[bitter sparrow]

probably remove and let the user know so they can submit something else?

[slender belfry]

You mean the ones like
"Maybe do something with the new 3D rendering options"?

[prisma delta]

Yeah, or the ones with long explanations

[slender belfry]

Hmm. On one hand, it would be nice to have them be homogenous. On the other hand, they historically don't get any votes anyways

[bitter sparrow]

how about ping the author and let them fix it or lose? 😄

[slender belfry]

Like "hey, your suggestion is malformed, you will probably get no votes FYI"?

[prisma delta]

Yeah, that's probably the way

[slender belfry]

I like that that one doesn't delete anything. Seems friendly enough.

[slender belfry]

Now what do I vote 😄

[prisma delta]

No need I have decided lol

[slender belfry]

BTW I hope you're enjoying your vacation @prisma delta. I've seen you respond to stuff on GitHub and couldn't help but think "Stop working!" 😄

[prisma delta]

The lack of responsibility is the big one lol

But mucking around helps me be less overwhelmed when I get back

[slender belfry]

Oh I believe that

[bitter sparrow]

vote autoCAD like the rest of us obviously

[slender belfry]

I'm supremely intrigued by that theme

[prisma delta]

Some borderline ones:

Fubar beyond destruction
Worms-like (the artillery game genre)
Mechanic Syndrome (It works when you go to get help)
This Jam was brought to you by Technicolor, and shot on location in the Uncanny Valley.
Should I ping all of these?

[slender belfry]

Hrmm

[bitter sparrow]

the first one i just don't understand what it means

the other ones could be pinged with a quick explanation as to what should be changed

no explaining the theme, or too long / specific

[slender belfry]

The first one is really obscure. The second and third one, I get why they added the explanation. Number 3 should probably just be "It works when you go to get help"

Yeah, I'd ping them

[prisma delta]

kk

Actually, one of you want to handle this? You have my blessing to pester folks ❤️

Be nice about it, and explain that historically themes with these problems get almost no votes

[slender belfry]

Sure, on it 🙂

[slender belfry]

Would you write them publicly on #jam or dm them?

[bitter sparrow]

you could probably ping all of them in a single message?

[slender belfry]

Yeah, I'll do that

[bitter sparrow]

as long as the reasoning isn't super specific for their particular theme

o just noticed there's at least one double-suggestion

rewin has two suggestions

[slender belfry]

I'll post it here first and then you can tell me if that's fine 🙂

[bitter sparrow]

sounds good

i'll do a double-suggestion search

[prisma delta]

Publicly in #jam 🙂

Consider grouping the pings by problem

[bitter sparrow]

this may be helpful, did some copy/paste and text editing:

[slender belfry]

Hi birds :broovy: 
Us from the Jam working group have spotted some submitted themes that are phrased in a way that, in our experience, makes them unlikely to be picked.
In general, themes that are easy to understand, quick to read and need no explanation have done better than others.
These are your themes and you're free to leave them as-is, but we've got some suggestions for how to improve them if you want.
@ivanceras: Fubar beyond destruction -> Total Destruction
@physgun.: Worms-like (the artillery game genre) -> Artillery
@__lyynn: Fog (there are things to do with the new volumetric lighting system) -> Fog
@screamsinjank: Mechanic Syndrome (It works when you go to get help) -> It works when you go to get help)
@__rewin__: Time travel and multiverse (let's blow our minds) -> Time travel and multiverse
@frank_7674: Mechanics Mayhem: Focus on unique and unconventional game mechanics as the central theme. -> Mechanics Mayhem
@__rewin__: Everything in this game is procedurally generated and there are no assets -> Total Procedural Generation
@dreamertalin: This Jam was brought to you by Technicolor, and shot on location in the Uncanny Valley. -> Uncanny Valley
@floppydisck: Focusing on some of the new features would be nice, like the lighting and fog improvements -> Lighting

Additionally, the following themes are very close to another one already submitted. Feel free to edit your post if you want to change it to a new one.
@neo.blondie: time travel and multiverse
@onkoe: cad software :)
@stdpi: cad and construct seems fun

Lastly, the following users have submitted two themes! Do you prefer to keep your first or second one?
@__rewin__, @jetp250, @DokkaeCat

[bitter sparrow]

some python magic later, found all the double-posts

DokkaeCat, jetp250, rewin

[slender belfry]

thanks!

lemme edit the post

[bitter sparrow]

in you want -> if you want

i'm not sure about providing suggested replacement themes, i think a general "this is the issue with your theme" would be better

and grouping the user tags by issue

[slender belfry]

Alright, sec

[fallen stag]

missing a paren for screamsinjank

error: unbalanced parentheses, does not compile

[bitter sparrow]

good jam theme

[fallen stag]

Jam theme: Could not compile project due to one previous error

take it its free

[slender belfry]

Hi birds :broovy: 
Us from the Jam working group have spotted some theme submissions that are phrased in a way that, in our experience, makes them unlikely to be picked.
In general, themes that are easy to understand, quick to read and need no explanation have done better than others.
You may want to edit your post to improve your submission.

May be difficult to understand:
- @ivanceras: Fubar beyond destruction

Shouldn't include an explanation:
- @physgun.: Worms-like (the artillery game genre)
- @__lyynn: Fog (there are things to do with the new volumetric lighting system)
- @screamsinjank: Mechanic Syndrome (It works when you go to get help)
- @__rewin__: Time travel and multiverse (let's blow our minds)
- @frank_7674: Mechanics Mayhem: Focus on unique and unconventional game mechanics as the central theme.

A bit long:
- @__rewin__: Everything in this game is procedurally generated and there are no assets
- @dreamertalin: This Jam was brought to you by Technicolor, and shot on location in the Uncanny Valley. 
- @floppydisck: Focusing on some of the new features would be nice, like the lighting and fog improvements

Of course, these are your themes and you're free to leave them as-is :)

Additionally, the following themes are very similar to another one already submitted. 
Feel free to edit your post if you want to change it to a new one.

- @neo.blondie: time travel and multiverse
- @onkoe: cad software :)
- @stdpi: cad and construct seems fun

Lastly, the following users have submitted two themes! No worries, but could you please decide which one you like best?
- @__rewin__
- @jetp250
- @DokkaeCat

[bitter sparrow]

May be difficult to understand: @ivanceras

Shouldn't include an explanation: @physgun., @__lyynn, @screamsinjank, @frank_7674

Too long: @__rewin__, @dreamertalin, @floppydisck

Very similar to an earlier theme suggestion: @neo.blondie, @onkoe, @stdpi

Lastly, the following users have submitted two themes! Do you prefer to keep your first or second one? @__rewin__, @jetp250, @DokkaeCat

[slender belfry]

Better?

[bitter sparrow]

heh

[slender belfry]

Same time 😄

[bitter sparrow]

i think including the suggestions verbatim may not be needed?

[slender belfry]

I'd be happy if I got pinged with the submission

[bitter sparrow]

understandable

[slender belfry]

Otherwise I'd have no idea what I submitted tbh

But that might just be me being forgetful

[bitter sparrow]

btw all the multi-suggestions are exactly 2 of them, in case that helps improve wording

seems we split into issues pretty much the same way

Has or needs an explanation:
this might be confusingly worded though

[slender belfry]

Ah, lemme change it then

[slender belfry]

ironic

[bitter sparrow]

true lol

[slender belfry]

I took your wording

[slender belfry]

@prisma delta what do you think about this? 🙂

[bitter sparrow]

okay i like the current iteration personally

[slender belfry]

Yeah I'll post it like that

[bitter sparrow]

oh btw

you can link to this channel directly

[slender belfry]

well, I can't post it 😄

Thanks, Discord!

[bitter sparrow]

split it into multiple messages?

[slender belfry]

Sec

[bitter sparrow]

curious what the "blocked content" is lol. spam detection ig?

[slender belfry]

Yeah multiple messages = no blocked content

[bitter sparrow]

nice 👍

[slender belfry]

[bitter sparrow]

rewin pinged 3 times 💀

[agile dirge]

I hope this won't be necessary, but we should encourage people to delete wrong submissions and enter new ones

Right now it's not a problem, but I'd rather not have an (edited) theme win

[bitter sparrow]

technically you can see the edit date i believe if you hover over:

[agile dirge]

hm, that should be fine then

just remove edited post submission

yeah you can, all good then

[bitter sparrow]

thoughts on screen transitions:

Okay so after some more thought, my current stance is that this feature would clearly make the template's game better, but it's not clear that it would make the template's code better, which is what users will be interacting with.

Opting for wont-fix for Bevy Jam 5 + reconsider later, especially if Bevy adds some new features that make screen transition code simpler and easier to add/remove.

[agile dirge]

agreed

[bitter sparrow]

cool

simple pr: https://github.com/TheBevyFlock/bevy-template/pull/106

[bitter sparrow]

looked at other trait naming conventions

Plugin is 100% consistent :)

and there are so many plugins, so that's impressive

[slender belfry]

Huh, didn’t expect that!

[bitter sparrow]

events are always either Event or no suffix. which i think is fair, sometimes the suffix would make readability worse

[slender belfry]

Yay!

[slender belfry]

Agreed

[bitter sparrow]

no concrete states are provided by bevy proper

[slender belfry]

Idk about how to do Triggers

[bitter sparrow]

like OnXyz vs XyzEvent?

technically they both impl Event

[slender belfry]

They’re technically events, but I kinda want to make it obvious that they’re meant to be observed

[slender belfry]

Exactly

[bitter sparrow]

yeah tbh they should be different traits, or the functionality should be unified somehow

but that's been discussed

[slender belfry]

Yeah its typing is a little bit awkward right now

Has Bevy consistently named its observables OnFoo?

[bitter sparrow]

bundles are 100% consistent

[bitter sparrow]

well observables only just landed 😄 i think bevy proper only provides the component events?

[slender belfry]

I thought some internal things were ported? I guess that was just an open issue and not yet implemented

[slender belfry]

Schedules have no real convention. Neither do components

So that really only leaves system sets, huh

[bitter sparrow]

i consider components to be consistently no-suffix

since none of them use a Component suffix

checking the same for resources

im keeping track in a list :)

[bitter sparrow]

all good

[slender belfry]

Could you open an issue on Bevy for the inconsistent SystemSets?

[bitter sparrow]

i'll probably open an issue for all naming inconsistencies at once i think

[bitter sparrow]

huh for schedules it's just ExtractSchedule that's the odd one out?

[slender belfry]

Oh wow, forgot about that one. Yeah, that should have no suffix.

[bitter sparrow]

found a niche one https://docs.rs/bevy/latest/bevy/render/render_graph/trait.RenderLabel.html#implementors

CameraDriverLabel odd-one-out

https://docs.rs/bevy/latest/bevy/ui/struct.UiLayoutSystemRemovedComponentParam.html

what's that java bean factory meme

https://github.com/bevyengine/bevy/issues/14274

basically everything is fine except SystemSet and 3 stray type names

ik marker components are inconsistent too but didn't bother trying to find them all

[hoary flower]

for next jam, I'd love to experiment with something to help with theme voting ; ludum dare "tournament voting" process is so much fun

[bitter sparrow]

we've created 12/20 of the latest upstream bevy issues

[agile dirge]

https://github.com/TheBevyFlock/bevy-template/pull/107
https://github.com/TheBevyFlock/bevy-template/pull/108
https://github.com/TheBevyFlock/bevy-template/pull/109

[bleak grail]

could you add an entry for this template in https://github.com/bevyengine/bevy-assets/tree/main/Assets/Templates?

[slender belfry]

I was waiting for that until the README was done

[slender belfry]

Approved all 🙂

Working on the macOS CI right now

Afterwards, I'll start a hackmd for us to work on the readme

And then I'd say we're done?

[agile dirge]

I'm adding animation cause why not

[slender belfry]

that'd be cool 😄

[agile dirge]

i will need a bit of feedback on how to structure this properly

[slender belfry]

BTW @prisma delta could I get some clarification on the "officiality" of this template? I have understood the following, but I want to make sure I got it right.

• The template is not part of the Bevy org because of two main reasons
  • The Bevy foundation has not enough maintainer-power (what is the gender-neutral term for manpower?) to guarantee that it stays up-to-date
  • The Bevy foundation at this point does not wish to enforce official opinionated ways of structuring a game
• The template is thus in a second-party repo managed by a working group
• The template is "semi-official" in so far as that
  • It is officially endorsed as a good starting point for jams
  • It does not use third-party crates
  • We take care of incorporating feedback by the community

[prisma delta]

So, that was the intended state of things. Given how in-flux this is (needs user testing, needs broader feedback, needs engine features), I'd like to strike "It is officially endorsed as a good starting point for jams" from the list and change the itch.io page to link the minimal CI template + the Bevy Assets page for templates (to which this should be submitted)

[fallen stag]

(i like horsepower or FTE 'full-time-equivalent' as neutral terms)

[slender belfry]

Thanks 🙂

[prisma delta]

I'm very pleased with how this work has gone (lots discovered, and lots to steal!), but I don't think this is mature enough to recommend to people coming to Bevy for the first time

[slender belfry]

That's a bit of a bummer :/ Agreed that the template page should be linked, but I feel strongly that this should at least be preferred over the CI template for beginners. We took great care to remove the parts that clash with the immature parts of the engine, addressed the feedback we got and simplified a lot in the process. What would be missing to regain that status?

[agile dirge]

I honestly don't think template is a good "learning material", it's meant to showcase basic functionality, but that's inherently complex due to the amount.
What the template gives is the ability to have a functional game without learning all that was necessary, you can straight up ignore parts of it and play around with specific features.
Examples are the begginer level learning resources

Personally I've never downloaded a template and went "okay let's read all of this now"

[prisma delta]

User testing is the big one for me. I'd like to do a pass on it myself once I'm off vacation, but even the carefulest examination won't stand up to users mucking around and breaking things.

[agile dirge]

The bug testing pipeline:

1. Deep core developers make a feature
2. User interface developers find bugs in core features
3. Template makers find bugs in user interface
4. Users find bugs in templates

bonus if you find rustc issues

[slender belfry]

Alright, that's fair enough.

[prisma delta]

I do want to promote it unofficially in #jam, and get folks with a bit more experience trying it out and roasting the patterns 😄

In terms of engine features, I'm also very annoyed that we don't have basic input-action mappings for use in a vanilla Bevy template

[hidden pond]

I do like that it seems that 'second party' has become the way to describe our affiliation since I used it yesterday lol

[slender belfry]

That sounds good. Keep in mind that it's likely people who dislike a certain pattern will voice that while people who are okay with it will probably not write anything. Sorry, I notice I'm being a bit protective here of the project 😅

[slender belfry]

Yeah I stole that from you 😛

[slender belfry]

I really really really want LWIM upstreamed 🎮

[agile dirge]

We're also entering a stage where engine provides 2-3 solutions to more generic problems and then it's a matter of preference

[hidden pond]

Continue to steal from me in that regard, after all: "Good Artists Copy, Great Artists Steal."

[prisma delta]

BTW, for audio (especially on web), you really need to load the various sounds into memory ahead of time or you'll get nasty stuttering

[slender belfry]

Hmm, you mean we need to add an asset loading stage without bevy_asset_loader?

[hidden pond]

(But fwiw I feel like it's an apt discriptor of our 'status')

[prisma delta]

Yep 🙂

[slender belfry]

Alright, I'll note that in the issue. Thanks for the heads-up

Is it okay if the assets/template description reads like this?

The template brought to you by the Bevy Jam Working Group.
A great way to start working on a jam entry. Includes a basic project structure.

[prisma delta]

Yep, this is great

[dawn crag]

bevy_asset_loader is handy and I use it in 90% of my stuff, but implementing a loading state without is pretty trivial and not scary.

[prisma delta]

Especially with the new substates!

[slender belfry]

Ah, good to know! I never looked into the implementation and just kinda assumed it did some magic, haha

[hidden pond]

Also I presume that work will continue on this template after the Jam...would we still want to organize in this wg even afterwards or? This doesn't necessarily need to be answered now, but I thought I'd ask just so that I don't forget.

[prisma delta]

(things I want to steal for the examples...)

[slender belfry]

I'd like to rename the wg to just "Bevy Jam" and continue work in here, if that's fine for @prisma delta

[prisma delta]

Yeah, that's fine by me 🙂

[slender belfry]

I have a suggestion: when you take a peek at the template after your vacation, you could write down a list of stuff you'd like to upstream as examples. We can then turn that into issues on the template so that we can submit individual PRs for them.

[prisma delta]

Will do!

[slender belfry]

I have found that I kinda like adding examples to Bevy 😄

It's nice to be able to focus on just a single feature and showcase it in the best way possible

[prisma delta]

Yeah, it's really satisfying

[slender belfry]

Regarding the jam in general, I don't want this comment to go unnoticed. I've never participated in a Ludum Dare, but I assume tournament voting means voting in a series of diminishing theme candidates?

[hoary flower]

yes, I can write a summary of the ludum dare process when I've got the time if there's interest

[slender belfry]

Thanks 🙂

[agile dirge]

why do we have a tag named "main"?

it's causing issues

[slender belfry]

maybe from the release workflow?

Feel free to delete it

Moving the ducky looks supremely blurry on my macOS. Is this an issue on main on other screens as well?

@agile dirge would you mind starting the application and moving the duck around real quick?

[agile dirge]

One sec my cat brought a bird as a gift...

Is motion blur enabled by default?

[slender belfry]

Dunno how visible it is in the video

But the sprite is not looking "crisp" while moving

[slender belfry]

I think the effect is visible on fullscreen

[agile dirge]

i mean

if i freeze the video

any given frame the duck is crisp

so it's not an app issue

[slender belfry]

You're right

This may be because of the missing visual interpolation that was yeeted with the fixed timestep

After all, in every frame, the ducky will move by a very heterogenous amount

If so, I guess it's fine to leave it?

[agile dirge]

animation helps

when the sprite is changing it's less noticeable

[slender belfry]

That's fair

[agile dirge]

https://github.com/TheBevyFlock/bevy-template/pull/110

[slender belfry]

I can check it out after I fix the macOS stuff

It’s really annoying to debug since the OS just tells you "this is wrong, I do not like" without any indication about what exactly is missing

[agile dirge]

k

[slender belfry]

Have you seen that we need an asset loading stage after all?

[agile dirge]

yes

[agile dirge]

https://github.com/TheBevyFlock/bevy-template/pull/111

[slender belfry]

https://github.com/TheBevyFlock/bevy-template/pull/112

Will need to check again once it's merged, but works locally

[bitter sparrow]

it's from manually triggering the release workflow without pushing an actual tag

[slender belfry]

@bitter sparrow I addressed your comments

[bitter sparrow]

idk where you got the powershell syntax from

it's not in the code you linked to

probably a mistake though, i think it will error when run

[slender belfry]

Is that not what you're reffering to?

[bitter sparrow]

nope

[slender belfry]

Or am I misunderstanding something?

@dawn crag recognize the code? 😛

[bitter sparrow]

that's bash syntax

[dawn crag]

yeah there's no way that could work on a mac. pyr is talking about the specific line pointed to in the review

and I have no idea where that came from, not mine

[bitter sparrow]

the | Out-File ... is powershell

[slender belfry]

OOOOOOOOH

I see

Yeah no way 😄

[bitter sparrow]

hehe

must have been copy-pasted from the windows job then

[slender belfry]

haha, sorry, it was so far right on my screen that my mind just blinked it out of existence

[slender belfry]

exactly

there you go

[bitter sparrow]

im so annoying for adding suggestions that just add a period to the end of a comment 😄

will review some more

@slender belfry left some more comments

i think the "upload package to artifacts" step is accidentally missing, and the "upload to itch.io" job needs artifacts or it won't upload the mac release

@slender belfry dumped comments

btw unrelated but i just want to share that i think i've got a great entity spawning API using entity commands

resources required to spawn an entity are decoupled from the system that spawns the entity, and an entity ID is returned immediately, and the ergonomics are nice imo

[dawn crag]

I think in the past it was suggested to just use https://crates.io/crates/cargo-bundle for mac packaging but I never looked into it / don't remember who suggested that.

[bitter sparrow]

just had to write a few extension traits to add some methods missing from upstream bevy

mac is so extra

it's an executable, what more do you need 😩

metadata and a specific package format apparently

[dawn crag]

@bleak grail might have had opinions about universal binaries, dmg vs app, cargo-bundle, etc.

[slender belfry]

Didn’t know about that, gonna try it out later 🙂

[agile dirge]

Share

Im curious what you crafted

[bitter sparrow]

extension traits here: https://github.com/benfrankel/bevy_jam_template/blob/main/src/util/patch.rs
UI widgets here: https://github.com/benfrankel/bevy_jam_template/blob/main/src/ui/widget.rs
example spawning the title screen here: https://github.com/benfrankel/bevy_jam_template/blob/main/src/screen/title.rs

if i could make w/e change i want to bevy, i'd make spawn take an EntityCommand instead of a Bundle and then impl<T: Bundle> EntityCommand for T

then you could do both commands.spawn(widget::menu_button("Quit")) and commands.spawn(NodeBundle { ... }) instead of requiring spawn_with for the former

the other thing that would make ergonomics a little bit nicer is being able to use SystemParams in entity command functions

for now you can use a SystemState at the top of the function if needed

[bitter sparrow]

actually idk how this interacts with being used in a command instead of a system

[agile dirge]

Do you use it anywhere? Im mostly curious in that

Aqh, you added example

[bitter sparrow]

yea

[agile dirge]

Isnt that the same version you posted earlier?

[slender belfry]

@bitter sparrow I addressed everything on the macOS PR

[bitter sparrow]

hmmmmmm

i think i made it work better somehow but idr

[slender belfry]

ah wait I did not, sec

[bitter sparrow]

i think there was some edge case before

that made it annoying for a spawn function to call another spawn function

but maybe i'd resolved that already and forgot

[slender belfry]

Now 👍

[bitter sparrow]

almost perfect

[slender belfry]

(I'm so sick of macOS build pipelines, please release me from my suffering)

[bitter sparrow]

the thing is

[slender belfry]

kill me

[bitter sparrow]

when i release you from your suffering

that's out of the pan and into the oven

github actions will be your next code reviewer

[slender belfry]

Sounds cozy ngl

[slender belfry]

;-;

nvm, not cozy

[slender belfry]

addressed

[bitter sparrow]

commited one small thing myself and approved

(adding quotes in one place)

[slender belfry]

thanks

[bitter sparrow]

merged, you can go ahead and trigger a release

[slender belfry]

whelp, this will take a while now since it has to build for two targets

[bitter sparrow]

probably still faster than the windows job

tbh maybe it's possible to split it up into three jobs

two to build for each macos platform, one to combine them into the final macos package

so the build jobs can run in parallel

make it work then make it fast tho ofc

[slender belfry]

Feel free to do that, I'm not touching that beast any more than necessary 😄

[bitter sparrow]

oh yeah for sure. ive spent enough time on the release workflow that i feel like i know what i'm doing with the code atp

which is sad, i shouldn't understand how the release workflow works

[slender belfry]

That great, actually 🙂

[bitter sparrow]

CI is supposed to be unintelligible

[slender belfry]

You could try generating the yaml from a more esoteric language?

Some custom DSL maybe?

[bitter sparrow]

i'll use a golf lang

like pyth

https://esolangs.org/wiki/Pyth

[slender belfry]

TIL

[bitter sparrow]

Every function, variable and statement in Pyth is a single character. In addition, [...]

[slender belfry]

My vote goes to Piet, I bet you know that one

[bitter sparrow]

yup

funny concept, no clue how it works lol

[slender belfry]

Oh it's actually fairly simple.
You see, a codel in Pie̷t̵ ̸i̴s̶ ̶l̸i̵ke an i̵͇͘m̷̘̈ḁ̸͋ḡ̵̱e̷̦'s ṕ̶̫i̵̳̍x̷̠̐ẻ̷̡l̸̛̞.
S͍̆ome Pi̵̢ët pr̴oǫràms a̴r̵̻e̶̳̗ ̷̗̈upscálȩd, ̴͔̥m̴̫̱ë̴̢ȃņi̵̜͋nģ thať ạ c͖o̵̹d̷̞ęĺ mḭg͑h̶̢̲t̷̞̱ ̴̨ṉŏţ alẁäỵš ̷̢̒b̵̛ě ̵̨̢
ąý̴̨̧s̵͔̼ ̸̢̰b̵̛̜ę̴̫ ȅqưìṽ̟͕ä̵̰̥ĺ̶ěņt tȯ 1 pi̴̢̡x̶̨̨e̵̡̢l̸̢̛,̵̛̠ ̴̨̨b̵̨ȕt a çơḑę̸̧̩ľ̶͍̥ ̷̧̧ị̶̢͓ş ảlẃäý̶̧͕ş̴̼̯ ̷̜a sưbstitu͖̰t̴̢̛ę̵̧̨ ̴̧̣f̵̞̹ơ̸͉̠r̶̢̛ pixēls.

[bitter sparrow]

im sure it would make looking at our CI code much more enjoyable

the benefits end there though 😂

[slender belfry]

No one will be able to submit a PR to the CI

idk if that's a pro or con tho

https://github.com/TheBevyFlock/bevy-template/pull/114/files

Need a quick approve on something trivial

[bitter sparrow]

done

[slender belfry]

Guess what. It really is faster to build twice for macOS than once on Windows 😄

[bitter sparrow]

not surprised, considering windows takes 3x as long as the other jobs 😔

heh i see the exit button in the web build on itch.io

target = "wasm" should be target_family = "wasm" i believe

[bitter sparrow]

yeah yeah lol

been like this from the start

[prisma delta]

Oh I wonder if it's multiple logins or something?

[bitter sparrow]

hmm if it's about the-bevy-flock user, it should just be francois

[slender belfry]

It's the terrorism we've been sneaking into the template 👀

[bitter sparrow]

nah there's no way that would be an issue

oneliner https://github.com/TheBevyFlock/bevy-template/pull/115

[slender belfry]

YAY the macOS build on itch.io and the GitHub release works!!!!!!!

I AM FREE!!!

[bitter sparrow]

amazing

[slender belfry]

Approved, but was that the only place where we used the wrong thing?

[bitter sparrow]

ope there are more. i'll add to pr

[slender belfry]

👍

[bitter sparrow]

done

oh you already approved right

[slender belfry]

hehe

[bitter sparrow]

also it seems the manual release workflow created a v0.1.0 tag. that's better than creating a main tag so should be fine

[slender belfry]

Yep!

Alright, my next task is writing a nice README. Here's the collaborative hackmd: https://hackmd.io/@hohenheim/H1dRzMhDC/edit

I hope no Discord bot scrapes this link and posts crypto in it

[bitter sparrow]

spooky

[bitter sparrow]

btw we're currently using Esc to go from playing screen back to title screen... unfortunately Esc is bad on web because it cancels fullscreen

not a big deal but mentioning

makes perfect sense on native tho

[nova trail]

Hey all! I'm not sure where the best place for this is, but I just created a Notion project to help organize my team for the jam and I think other teams could benefit from it as well.

You should be able to duplicate this template and customize it for yourself using a button in the top right corner. https://hot-bolt-35e.notion.site/Bevy-Jam-Team-Planner-8ea225058e894660b463dedffa33ede6?pvs=4

Would anyone find this useful and if so, how could we get the word out?

[bitter sparrow]

this is the right place 😄

[prisma delta]

Linking it in #jam now and as the jam starts would be good

BTW, I'd really like a team-matching mechanism

Maybe we just do it manually?

[bitter sparrow]

how would team-matching work?

usually there are various requirements for a team, like at least one artist, or matching timezones, matching goals, etc.

[slender belfry]

People write in a designated chat with their role and experience and we match them?

[prisma delta]

That's the sort of requirements I'd like to figure out 🙂

But yeah, #jam-find-a-team always feels kinda unhelpful and chaotic to me

[bitter sparrow]

btw. funnily enough, previous jam i teamed with 1 other person and we had completely different timezones

[slender belfry]

Add a little checkmark emoji reaction to the people that were already matched

[slender belfry]

Oh right, that’s important as well

[bitter sparrow]

we convened twice a day when we were both awake, and then we'd swap working solo for the day

which actually surprisingly was an advantage

no merge conflicts or arguments during work time, so it can be more productive

[prisma delta]

Yeah, I don't think I'd want to be in the same time zone 🤔

[bitter sparrow]

automatically creating a thread on each post in #jam-find-a-team maybe?

some people don't like receiving dms, and creating a thread yourself is a barrier

some kind of template for posting info about yourself and what you're looking for in a team could be helpful in organizing the posts

[slender belfry]

Talking about time zones: I'm off for the night, see ya 👋

[bitter sparrow]

like "my timezone is xyz, i don't care about others' timezones", or maybe you do care

[nova trail]

I have a section in my notion template that covers team bios and I found that just having a bit more structure to people's introductions can do wonders for matchmaking:

Overview

[Enthusiastic one sentence introduction goes here!]

Interests

[favorite games and other interests go here]

Portfolio

[links to portfolio, github projects, notable social media accounts, etc go here]

Jam Availability

Time Zone: CST
[days and hours of availability during the jam go here]

Contact Info

Discord:
Github:

Well, perhaps it should take up less vertical space for the discord version 😅

[bitter sparrow]

Hi, I'm Pyrious! I love bevy blah blah blah...

Skills: Code, pixel art, etc.
Timezone: GMT+55.
Availability: I'll be busy but can contribute a few days.
Portfolio: https://pyrious.itch.io
Favorite games: Minecraft, Noita, etc.

contact info less relevant for matchmaking and more relevant after the team has been formed

other lines that could be included:

Goals: I want to submit a game!
Looking for: Similar timezones, someone with music skills, etc.

[bitter sparrow]

bumping this

[bitter sparrow]

@slender belfry am i reading this right that the MacOS package has assets in bevy-template.app/Contents/MacOS/assets/assets/?

[bitter sparrow]

working on release.yaml in my own template. will make a PR after everything looks good

[slender belfry]

At least not in the final package, no

[slender belfry]

Maybe you could speed up Windows by using LLD? Don't bother with cranelift though, it's really not there yet on Windows for anything but cargo check, which I'm not even sure it speeds up since that shouldn't produce any LLVM anyways

[bitter sparrow]

interesting.

[bitter sparrow]

i can give that a shot as well

[slender belfry]

Also https://github.com/Mozilla-Actions/sccache-action

That would also be nice on non-Windows targets

[slender belfry]

Added cargo generate support! Run cargo generate TheBevyFlock/bevy-template --branch cargo-generate

@hidden pond ^

I also removed licenses from that branch

In theory, we could ask the user if they want to keep it or not

But eh 🤷‍♂️

[bitter sparrow]

smart

nah it should 100% default to no licenses and user can add their own

[slender belfry]

@agile dirge I've left reviews on your PRs 👍

[bitter sparrow]

my github workflow-fu may be powerful enough now to combine all of the jobs into one matrix job. we will see

[bleak grail]

It was decided against in the official template to not introduce complex CI concepts

[slender belfry]

I have come to treat the CI template mentally as more of an extension of bevy/examples, so in that vein, that makes perfect sense.

[slender belfry]

I'd be interested in seeing that. I suspect that I will have no clue what’s going on, but I’m ready to be proven wrong 🙂

[slender belfry]

@bitter sparrow got a small CI improvement: https://github.com/TheBevyFlock/bevy-template/pull/121

[agile dirge]

updated Prs

[slender belfry]

reviewed both 👍

Only left one small suggestion each

[slender belfry]

Went ahead and also added cargo-cache and sccache to speed things further up.

[bitter sparrow]

ok, will check out once im done with matrix change

want to finish it while it's in my head

[bitter sparrow]

390 -> 318 lines isn't as great as i was expecting, but:

1. it builds macOS in parallel
2. my template is not using trunk yet, so i have to do those steps manually
3. had to split some steps into smaller steps to make each platform align so they can be merged
4. had to add "if platform == web" etc. checks to a lot of steps

still have to test that it works :p

thoughts on using variables instead of env for configuration?

aka: https://docs.github.com/en/actions/learn-github-actions/variables#creating-configuration-variables-for-an-environment

instead of values directly in the release.yaml file

i'm thinking probably not worth

would be useful if we had multiple workflows that need to access the same variables

[slender belfry]

Is that at the same place as where the itch.io secret goes?

If so, the user needs to go there anyways, so it might save them from having to touch the workflow file

[bitter sparrow]

it seems to be the same page, yes

the real reason i asked about using variables is because it'd let us remove the useless check-if-itch-is-configured job lmao

i did find a way to make that job more concise though

i really waited 15 minutes for CI to fail with this error message

mv: './tmp/bevy-jam-template-linux/bevy-jam-template/run' and './tmp/bevy-jam-template-linux/bevy-jam-template/run' are the same file

[bitter sparrow]

(i actually ended up fixing 3 errors from the one CI run i just thought this error was funny)

i discovered that rm -f xyz is not an error if xyz doesn't exist

but mv -f xyz xyz does error lol (whether or not xyz exists)

[agile dirge]

Prs updated

[slender belfry]

Approved both
Good work!

[agile dirge]

Animation is waiting @bitter sparrow approve

also a rebase, ill do that quickly

[hidden pond]

Thanks for doing that! Apologies for taking so long. IRL has been busier than normal lately >.>

[slender belfry]

No worries 🙂 Feel free to add options to the generate.toml, there’s a bunch of possibilities I didn’t implement. The user currently has to enter an itch.io name, but it would be nice if that could be left blank, which would disable that part of the CI. We could also add their name to the credits automatically.

[slender belfry]

https://github.com/TheBevyFlock/bevy-template/pull/123

Added asset credits into the game

BTW @agile dirge I just realized that I didn't even try the sounds out yet, I always had the game on mute.
Checked it out now and it's very very fun 😄 Love the footstep sounds combined with the music, it's so cute haha

Got a super tiny PR: https://github.com/TheBevyFlock/bevy-template/pull/124

[agile dirge]

that's covered in duck animation

cause i was checking module headers

[bitter sparrow]

oh yeah you were missing out

[slender belfry]

Ah, good. I'll close my PR then.

@bitter sparrow I addressed your comments

[agile dirge]

shouldn't cause a conflict though

[slender belfry]

👍

[bitter sparrow]

so wrt the remaining comment, i misread

but im still a bit confused

the other jobs run 1. scc-cache 2. cargo-cache 3. cargo build, in that order

the macos job has two cargo builds

[slender belfry]

Yep, and I think cargo cache reads the current target and caches that. But I'm not sure. That's why I ran it twice

Figured it can't hurt 🤷‍♂️

[bitter sparrow]

then cargo cache should run after cargo build in the other jobs?

[slender belfry]

What? no

[bitter sparrow]

current target would be empty before, though

[slender belfry]

It fills the target IIRC

[bitter sparrow]

hm ok

[slender belfry]

But I'm really really not sure. Summoning @lunar sail

In any case, I know it works now. It might just be running once in the macOS flow without doing anything, but eh

I figured if you can split the macOS job it won't matter

[bitter sparrow]

yeah

[slender belfry]

as then they will both also follow sccache -> dtolnay -> cargo-cache -> cargo build

[bitter sparrow]

oh it automerged lol i approved with 2 nits

all good just comment nits

hm

is it wrong to run wasm-bindgen after wasm-opt?

i'm getting an error during wasm-bindgen currently and wondering if this is the reason

[bitter sparrow]

left a few nits. largely looks good though so i'll approve after

also clippy complaining about an unused import

[bitter sparrow]

one comment, will approve after

[slender belfry]

done 👍

[slender belfry]

That definitely sounds wrong, yes

[bitter sparrow]

yeah it seems so

also wasm-bindgen outputs binary_bg.wasm right

weird

[slender belfry]

Fixed them in https://github.com/TheBevyFlock/bevy-template/pull/125

[bitter sparrow]

appreciate it

[slender belfry]

Triggering a release to populate caches

[bitter sparrow]

should populate on CI as well? is the cache shared between CI and CD?

i'm assuming yes

[slender belfry]

It is currently not, no

[bitter sparrow]

ic. they use different profiles actually so that makes sense

[slender belfry]

I think we can configure cargo-cache to reuse the CI cache in CD

But that will only work for Linux

[agile dirge]

https://github.com/TheBevyFlock/bevy-template/pull/110 updated

[bitter sparrow]

approved, committed a quick tiny fix

[agile dirge]

wait, u sure?

default transform sets Z = 0

[bitter sparrow]

scale = 1 should be the default

[agile dirge]

gah

[bitter sparrow]

for translation yea

[agile dirge]

right

[bitter sparrow]

hehe

[agile dirge]

but hear me out 👉 👈
we're representing infinitely thin shapes

/j

mind if i delete 0.1.0 tag and add it again, so this version goes on itch?

i wanna check audio stutters

[bitter sparrow]

release workflow should upload to itch regardless

[agile dirge]

i don't want a main tag

it gets prioritized over main branch

[bitter sparrow]

well you can delete main tag for sure

the release workflow should be creating a v0.1.0 tag now instead

[agile dirge]

oh, alright

ill trigger manually then

[bitter sparrow]

np to delete that as well, it'll just be recreated

[agile dirge]

ah, Jan is already running one

I'll test on trunk then

in the meantime

[bitter sparrow]

good to make sure trunk works

i should prob try out trunk on linux just to be sure. have never used it

[slender belfry]

Feel free to cancel and start a new one. Mine is just to populate caches.

[agile dirge]

the asset preloading landed earlier so it should be fine as long goes to itch

[slender belfry]

Oh wait, it's done anyways 👍

[agile dirge]

no stutters 👍

[slender belfry]

No stutters for me either 🎉

Only some visual stutters because of lack of interpolation, but that's fine

[agile dirge]

idk man, you've got some 144Hz eyes or something

I don't see any issues xD

[slender belfry]

No, just running it on a shitty laptop 😄

I also get no visual stutters on my desktop PC 🙂

Looking through the issues, I think the only thing missing is better readme: https://hackmd.io/@hohenheim/H1dRzMhDC/edit

Feel free to add brief TODOs and comments in there, I'll try and write everything out nicely later today

@lunar sail about to roast how wrong I was about how cargo-cache works 👀

[lunar sail]

Hi. Indeed cargo-cache will fill target and some other folders, so it should be used before cargo build.
But I didn't check out sccache yet so I'm not sure how the two interact or if sccache can fully replace caego-cache

[bitter sparrow]

just wanted to share while i'm going insane, i got the most blazingly-fast ubuntu-latest instance in my latest release attempt

3:19 to cargo build + wasm-bindgen + wasm-opt for web release

meanwhile windows is still going at 13:00+ 🤪

[lunar sail]

Hmm, at first look, sccache might be able to replace cargo-cache 🤔
cargo-cache is also a bit hacky as we had to work around the limitations of GitHub's weird caching system. Maybe sccache can also work more reliably

I don't have much time this weekend to look into it, but I'd recommend to try out just using sccache for a project and looking in the CI logs to see if it's sufficient

[slender belfry]

I'll test it right now 🙂

Running a release now with pre-built caches using both LWCC and sccache

Will then write a PR to remove LWCC and then trigger two releases again

[lunar sail]

Thanks! Curious to hear what you'll find :)

[slender belfry]

Got a trivial PR in the meantime: https://github.com/TheBevyFlock/bevy-template/pull/126

And here's the PR for removing cargo-cache: https://github.com/TheBevyFlock/bevy-template/pull/127

[slender belfry]

Looks like sccache is not caching anything 👀

[bitter sparrow]

mmm another way to do this would be let _ = app_exit, but that's kinda indirect

[slender belfry]

that's supremely ugly, plz not

[bitter sparrow]

:p

the cfg spam is ugly too but yeah. code would be prettier with bevy_mod_picking

[lunar sail]

It might only work for subsequent runs. You can try pushing an empty commit once this CI run has completed

[slender belfry]

I was using sccache + LWCC before

so it should not matter

[lunar sail]

Hmm.. then I'm not sure

[slender belfry]

Same. I don't really want to spend too much time on this since cargo-cache is working nicely already. I'll remove sccache for now instead

[bitter sparrow]

make a question issue about sccache probably

[lunar sail]

Once I have more time again I'll try to look at sccache in more detail

[slender belfry]

Feel free to use bevy-template as a guinea pig 😉

[slender belfry]

Aight

https://github.com/TheBevyFlock/bevy-template/pull/127

[bitter sparrow]

merged

[slender belfry]

hurray!

[bitter sparrow]

also my WIP matrix workflow finally succeeded (up to the itch.io upload where it failed but i know why)

idk if the packages are actually good yet but it's almost done

[slender belfry]

Oh great CI wizard, could you also merge main into cargo-generate whenever a PR is merged?

(And maybe create an issue if there's a merge conflict)

[bitter sparrow]

😰

[slender belfry]

Not a big deal if not, I'm just doing it by hand every now and then 😛

[bitter sparrow]

maybe but definitely not in the near future lol

[slender belfry]

👍

[bitter sparrow]

after the jam maybe

[slender belfry]

I think it should be simple though

But yeah, no hurry

[bitter sparrow]

so the workflow im working on has two matrix jobs

build and package

build produces binaries as artifacts, and package produces packages

so build produces 1 more artifact than package, because of the double macos

[slender belfry]

Makes sense

[slender belfry]

noooooooo

[bitter sparrow]

oh the upload lol

[slender belfry]

@agile dirge the synchronized steps with the animation are supremely cute as well hehe

[bitter sparrow]

packages look good under manual inspection of their internal structure

besides macos cause i cant open .dmg

i'm doing a zip named bevy-template-windows.zip which contains the structure bevy-template/{bevy-template.exe,assets/}

the version number isn't needed in the package name because itch.io and github release both have custom ways to specify version number

[slender belfry]

Can you upload it somewhere?

I can check

[bitter sparrow]

will make a PR as soon as i get this green + add cargo cache

[bitter sparrow]

you should be able to see it here? https://github.com/benfrankel/bevy_jam_template/actions/runs/9910265871

in the artifacts

[slender belfry]

yeesss

sec

I'm sorry

[bitter sparrow]

rip

[slender belfry]

wait, I'll try the build artifact

[bitter sparrow]

seems 7z can extract .dmg on linux, i'll try inspecting the internals

[slender belfry]

I think that's good for the build artifact?

Contents of the dmg look like this

Info.plist:

  <?xml version="1.0" encoding="UTF-8"?>
  <!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
  <plist version="1.0">
      <dict>
          <key>CFBundleDevelopmentRegion</key>
          <string>en</string>
          <key>CFBundleDisplayName</key>
          <string>bevy-jam-template</string>
          <key>CFBundleExecutable</key>
          <string>bevy-jam-template</string>
          <key>CFBundleIdentifier</key>
          <string>Pyrious.bevy-jam-template</string>
          <key>CFBundleName</key>
          <string>bevy-jam-template</string>
          <key>CFBundleShortVersionString</key>
          <string>v0.1.0</string>
          <key>CFBundleVersion</key>
          <string>v0.1.0</string>
          <key>CFBundleInfoDictionaryVersion</key>
          <string>6.0</string>
          <key>CFBundlePackageType</key>
          <string>APPL</string>
          <key>CFBundleSupportedPlatforms</key>
          <array>
              <string>MacOSX</string>
          </array>
      </dict>
  </plist>

[bitter sparrow]

hm

internal structure looks the same as bevy-template

[slender belfry]

The executable flag is not set

It tries to open the game as a document

[bitter sparrow]

is that the case for bevy-template as well or just mine?

interesting

[slender belfry]

just yours

pretty sure that's the issue

can you throw in a chmod +x before packaging?

[bitter sparrow]

ok. i wonder if that's due to the binary being uploaded as an artifact then downloaded to be packaged

maybe the flag gets lost during that sequence

[bitter sparrow]

yeah i can try that

[slender belfry]

And that was not a problem on Linux?

[bitter sparrow]

i'll check

i only looked at the file structure didn't try running it

[slender belfry]

I mean, I guess Linux users are used to setting chmod +x manually

[bitter sparrow]

that is also true 😄

yeah the flag is missing on linux as well, runs after adding it

and yeah confirmed the flag is set for bevy-template

cool it's consistent then

only weird thing is the wasm file shouldn't be marked executable (currently neither of them are)

wow

https://github.com/actions/upload-artifact/issues/38

it's literally a known issue since 2019

[slender belfry]

And yet again the template runs into edge cases 🎉

[bitter sparrow]

why was bevy-template not hitting this i wonder

ah the download is when the zipping occurs, apparently

wild bug to not fix ngl

is it a big deal if the .wasm is marked executable

if not i have to edge case for that as well

which is do-able EDIT: did it

[bitter sparrow]

linux and web builds confirmed working now

will add cargo cache then make pr to bevy-template

[bitter sparrow]

cargo cache worked, i updated my CI as well and will try to bring the improvements to bevy-template

https://github.com/TheBevyFlock/bevy-template/pull/130

starting with ci.yaml because it's easier

the main difficulty here will be the web build, because it uses trunk

which creates the final package directly instead of creating a binary first like all the other platforms

web: 0 binaries, macos: 2 binaries 😄

i think web build will be better off not being in the matrix for bevy-template

so i'm still a bit confused about cargo test @slender belfry

cargo test --all-targets: does that test docs?

cargo test does i believe

[slender belfry]

Wrote something new

You're right

(I think)

[bitter sparrow]

😭

   Documentation tests
       Documentation tests are also run by default, which is handled by rustdoc. It extracts code samples from documentation comments of the library target, and then executes them.

this is why i'm confused about cargo test vs cargo test --all-targets

[slender belfry]

You know what

I'll just run it

let's see

[bitter sparrow]

like it seems to me that --all-targets actually reduces how many targets are being tested (specifically removing doc tests)

from the man page

[slender belfry]

I don't think that is the case. When I use rust-analyzer without --all-targets, it will not show me all errors, but also take up less space on disk. There is something going on with things like binaries and examples.

[bitter sparrow]

rust-analyzer may have different logic than cargo-test

for example cargo-check man page says that by default it checks binary and library targets

[bitter sparrow]

and then that --all-targets checks all targets

but --doc isn't even an option

[slender belfry]

Yeah there's an open issue for that since years

[slender belfry]

This suggests that any weird behavior is entirely cargo's fault

[bitter sparrow]

all targets ||except docs||

[slender belfry]

Some Targets Are More Equal Than Others

[bitter sparrow]

rea

https://github.com/rust-lang/cargo/issues/6669

[slender belfry]

--all-targets-for-real-for-real

very.

cool.

[bitter sparrow]

😩

hm well i can just do two commands to be extra sure

in the same job

first all-targets, then docs

[slender belfry]

Sounds great acceptable

[bitter sparrow]

updated pr

[slender belfry]

Left some final comments

[bitter sparrow]

resolved

[slender belfry]

Approved

(But need to change branch protections since the jobs have different names now 😄)

there we go

Oh heck

Missed something small

[bitter sparrow]

ope

[slender belfry]

That's why the docs step is taking so long, hehe

[bitter sparrow]

ah that is intentional

documenting deps so that links to deps will be valid

whether that's actually needed idk, but that is my rationale

[slender belfry]

Huh, does that actually not fail otherwise?

[bitter sparrow]

not sure

i'll add the no-deps for now

[slender belfry]

I'll test locally real quick

https://github.com/TheBevyFlock/bevy-template/pull/131

small PR

[bitter sparrow]

it wasnt this slow in my template :p

[slender belfry]

hehe it can't actually merge before yours because I already changed the branch protection rules

[bitter sparrow]

so i think the reason the ci is slow rn

is because there is no cache for the doc test command yet

and then all-features balloons the time

[slender belfry]

we also need RUSTDOCFLAGS='-D warnings'

otherwise the docs thingy doesn't do much

[bitter sparrow]

Failed to save: Unable to reserve cache with key Linux-(ffa9cf99a 2024-06-03)-doc-9d2c396dd6346a52c23fcf9d3d30f3533e6b2c25ccd0477a4d22ba88992e775e-60db438780d04b1e73edaa22633702f4b51bd392256300103bbf383b4a87f279, another job may be creating this cache. More details: Cache already exists. Scope: refs/pull/130/merge, Key: Linux-(ffa9cf99a 2024-06-03)-doc-9d2c396dd6346a52c23fcf9d3d30f3533e6b2c25ccd0477a4d22ba88992e775e-60db438780d04b1e73edaa22633702f4b51bd392256300103bbf383b4a87f279, Version: 4464b218375e2702da53c92c2e51ffdd3cdd5aeba23dec304d9c1239b4ef594a

output of CI trying to cache after doc test

[slender belfry]

works fine with --no-deps 🙂

[bitter sparrow]

ok now try something that exists?

[slender belfry]

all good 👍

[bitter sparrow]

ok great

ci looks good

fast and no cache warnings

[slender belfry]

Nice!

Left you a final final comment

[bitter sparrow]

love that for cargo

can i do --deny warnings instead of -D at least?

[slender belfry]

No

[slender belfry]

idk

[bitter sparrow]

i mean in the RUSTFLAGS

[slender belfry]

no clue honestly

[bitter sparrow]

ok

[slender belfry]

The deeper you look into it, the more cargo gets

[bitter sparrow]

$ rustdoc -h | grep deny
    -D, --deny LINT     Set lint denied
$ rustc -h | grep deny  
    -D, --deny LINT     Set lint denied

should be good

[slender belfry]

yay!

[bitter sparrow]

also i assume i can just add this to the global env right?

[slender belfry]

sure

At that point, you might as well set RUSTCFLAGS to deny warnings as well to be safe? idk

[bitter sparrow]

oh i did add RUSTFLAGS heh

[bitter sparrow]

redundant?

[slender belfry]

Not sure

since it's clippy and not rustc

eeeh let's leave it there

Approved

[bitter sparrow]

sounds good

should automerge blazingly fast

[slender belfry]

why is it recompiling the tests 👀

[bitter sparrow]

idk. it seemed to hit the cache

[slender belfry]

Maybe because of the RUSTFLAG?

[bitter sparrow]

ah could be

[slender belfry]

kinda makes sense, since we changed the build config

[bitter sparrow]

release.yaml will be even more fun

[slender belfry]

❤️

cargo test is still recompiling everything

is it possible that cargo test --doc is yeeting the target?

OH OF COURSE

--all-features is missing

sneakily added it to my already approved PR

[slender belfry]

#1261408668159709315 message

[bitter sparrow]

@slender belfry https://github.com/TheBevyFlock/bevy-template/pull/132

the LoC savings aren't as big because web build is still a separate job

i think i could combine the two matrices into one by allowing multiple targets per job so that macos can do its thing

[bitter sparrow]

well that saved another 49 LoC. will have to test that it works

[bitter sparrow]

may be possible to throw the web platform back into the matrix now that there's no intermediate job where it has to produce only binaries

[bitter sparrow]

actually i just realized this unparallelizes the macos builds 😅

it may be worth the tradeoff to simplify the workflow, especially since windows takes forever anyways. but probably not

https://github.com/orgs/community/discussions/14484

this would solve the issue

unless there's not enough cpu to run two cargo build in parallel on a single runner :p

are built dependencies shared between x86_64-apple-darwin and aarch64-apple-darwin targets? the second target never compiles dependencies

i wouldn't expect them to share though

[bitter sparrow]

my current take: keep the simpler approach for now. macos still finishes 3min faster than windows

and only like 30sec slower than linux

in the few release runs that i've tried

if macos becomes the bottleneck in the future, it can be re-evaluated

[dawn crag]

I got curious about runtime performance with various size optimizations: https://github.com/rparrett/bevy_wasm_bench

[bitter sparrow]

looking at the data, there's a lot of it. did you come to any conclusions?

[dawn crag]

No idea how that generalizes to other HW, and I didn't make a huge effort to make things not noisy beyond closing all running programs, so not very scientific.

Yeah, I have a few takeaways.

opt-level = "s" is great (faster than 3), opt-level = "z" is slow. At least for this benchmark on my hardware.

[bitter sparrow]

its interesting that lto="fat" makes wasm-opt significantly faster than with "thin"

[dawn crag]

lto = "thin" seems better and faster than "fat".

[bitter sparrow]

the combined time is still lower with thin but not by that much, more like 10-30 seconds diff afaict

but the runtime is pretty much equal

[dawn crag]

I am generally slightly suspicious of the data but it took like half a day to gather so I don't know if I'll revisit it too soon, lol.

[bitter sparrow]

codegen-units=1 making the combined time lower (in some cases)... interesting

[dawn crag]

z + z is significant for filesize, but not worth the runtime perf cost IMO.

[bitter sparrow]

opt level S having lower frame time than opt level 3 is interesting

[dawn crag]

That possibility was alluded to be someone or some document somewhere, so it's cool to see that potentially confirmed for this bench, anyway.

[bitter sparrow]

seems like the winner is S + S + lto=thin + codegen=1, and strip=debuginfo doesn't really matter

we're currently doing Z + S + lto=thin + codgen=1

s: optimize for binary size.
z: optimize for binary size, but also turn off loop vectorization.

... why would we turn off loop vectorization

ig z is meant to be extra aggressive size minimization. we totally don't need to default to that

it does make a significant difference to size though, so mentioning it as an option for sure

[bitter sparrow]

@dawn crag it seems codegen-units=1 had a similar effect on compile time as lto="thin"

im making a pr to the bevy website to update the compile options advice

so i'll probably reword them to both say "slightly slows compile time"

or just "slows compile time" because it's not exactly slight, it's just not huge

interestingly codegen-units=1 doesn't seem to do much besides increase compile times if you already enabled lto

maybe it shouldn't be recommended at all

or that might be overfitting to the benchmark

https://github.com/bevyengine/bevy-website/pull/1553

[bitter sparrow]

it's possible codegen-units=1 doesn't do much because the crate being benchmarked is small

idk if codegen-units applies to the dependencies being compiled, maybe it does (like 1 codegen unit per dependency)

[hidden pond]

BTW just submitted this PR for the cargo-generate stuff:
https://github.com/TheBevyFlock/bevy-template/pull/133

[bitter sparrow]

did this, now +130 −272 😄

[slender belfry]

That's pretty good 💪

[bitter sparrow]

three different types of template variables with cargo-generate in release.yaml is crazy

necessary obviously, but crazy

(cargo-generate variables, github workflow variables, and shell variables)

[slender belfry]

reviewed 👍

[bitter sparrow]

we left the same comments 😭

[slender belfry]

😄

[slender belfry]

Left a tiny nitpick

Look "good" otherwise, in the sense that for me, a mere mortal, it sure looks like a workflow.

[bitter sparrow]

hehe. the diff is kinda hard to read because the file is mostly fully changed

[slender belfry]

I like how all the duplication is gone 🙂

[bitter sparrow]

yep. adding new features to all builds (like caching) is really easy now

[slender belfry]

Yeah I just looked at the final file

[bitter sparrow]

the real test is whether the release will succeed. ive been testing on my own template, but i don't use trunk so there are some differences

[slender belfry]

Just checked, yes, exactly the same comments hahaha

[slender belfry]

Move fast, break things!

Approved

[bitter sparrow]

made the issue: https://github.com/TheBevyFlock/bevy-template/issues/134

[slender belfry]

thx!

[bitter sparrow]

also it seems the bevy github CI template is trying to be the bare minimum "this is what setting up CI looks like" type of thing

so making the CI fast with caching etc. is probably out of scope there

adding more comments would probably be in scope

but i don't personally feel like maintaining another release.yaml file rn

so bevy-template's README should probably have its own custom explanation of how to set up CI / CD. it should still try to keep it simple though

maybe link to a separate document like README-CI?

[hidden pond]

@slender belfry gone ahead and updated PR based on the feedback

[slender belfry]

https://hackmd.io/SuC0b_aWSZaloSmy_GVZXg

I already wrote a bit about how to interact with the release.yaml

[slender belfry]

It all makes sense when you consider the CI template an extension of bevy/examples and not an actual template

[slender belfry]

Could you merge main into cargo-generate in that PR as well to get @bitter sparrow' work in?

[slender belfry]

Left some more comments

BTW, I really wanted to have the README done by today, but my health again took a bit of a downturn :/
I'll try to get it done tomorrow. As soon as that's finished, I'll record a little video of the template for the Bevy Assets page, submit a PR there, and then go and spam #crates and #jam

[bitter sparrow]

yeah the README is the last thing that needs to get done, everything else is just extra

so we can focus that

[slender belfry]

I'm really really happy with how the template turned out

I think this is legitimately the best showcase of a Bevy project in existence today

[bitter sparrow]

i agree for the most part. i only think we're overusing observers

[hidden pond]

Alright I've updated the PR.

[slender belfry]

Very possible. I'm curious how a version using Commands would look.

[bitter sparrow]

check out https://github.com/benfrankel/bevy_jam_template/blob/97971d49045fcac22ae60386b603cac08e9bfff3/src/screen/title.rs#L44

[slender belfry]

@hidden pond did you find the EditorConfig plugin?

[hidden pond]

Yes

[bitter sparrow]

(this is using a few extension traits patching things that should be provided by upstream bevy, in util/patch.rs)

note that the "entity spawning functions" have full world access and don't have to return anything, and the code calling the entity spawning function receives an EntityCommands for further entity building just like the usual Commands::spawn

because "entity spawning function" is actually just a generalization of the insert entity command

[slender belfry]

Looks great now, thanks!

[slender belfry]

Looks pretty sweet 👀 But yeah, it also looks too custom to hand it to a beginner

[bitter sparrow]

maybe for 0.16. i'm not gonna push for it at this point. but also i think you could definitely say the same about the observer approach

it's mostly an upstream bevy issue

[slender belfry]

Those look pretty nice as well. Have you opened a PR on Bevy?

[bitter sparrow]

PR no, but they all have issues. i'll probably try and implement them

[hidden pond]

I'll add in the Licenses stuff sometime later as that'll require...a bit more involved stuff.

[slender belfry]

Oh btw, I'd like to collect all the issues and PR we've created while working on the template.

[bitter sparrow]

as a post-mortem?

[slender belfry]

Yep

[bitter sparrow]

sounds good, should be easy to search based on author

[hidden pond]

(Also we can do other things. Like provide a template README if we wanted, and also exclude certain files from the 'template'. So that if you use the template those files don't get copied over. If we wanted).

We also probably should get CI setup for the template to ensure the template actually works now that I think about it...but that's not super important right this moment. That'll be useful with the merging from main into cargo-generate branch, to make sure something didn't accidentally break.

[lunar sail]

It always surprises me how hard it is to get Rust CI right lol

--all-features --all-targets --workspace --please-im-begging-you-just-do-it

[bitter sparrow]

starting a release workflow now for testing. somehow the linux build completed in under 3 minutes

do we know why github forgot that the release workflow is named Release?

it's showing as .github/workflows/release.yaml in the UI

while ci still shows as CI

doesn't matter it's just weird lol

[bitter sparrow]

successful :)

need to test that the packages are good though

web build is good

ok the executable is not being renamed properly, x86_64-unknown-linux-gnu sounds spooky and is not intentional lmao

linux build does in fact work though

[bitter sparrow]

small PR to fix this: https://github.com/TheBevyFlock/bevy-template/pull/135

[slender belfry]

approved

[bitter sparrow]

do we know why the CI tests recompile every time?

maybe it would be better split into a few jobs to parallelize

not a big deal

[bitter sparrow]

someone/something fixed this

[slender belfry]

It started when we added the --doc to the second tests

Maybe it will work better if we split the regular test and the doc tests into two jobs?

[bitter sparrow]

i'll make a pr to try that rq

https://github.com/TheBevyFlock/bevy-template/pull/136

[slender belfry]

Let's wait until the CI through and then push something to trigger it again

[bitter sparrow]

ok makes sense

current CI is done

pushed an empty commit to the PR

doc tests hit cache, other tests did not

https://github.com/TheBevyFlock/bevy-template/actions/runs/9922966385?pr=136

[slender belfry]

the heck o.O

[bitter sparrow]

technically it hit the cache but it rebuilds anyways

the old command:

run: cargo test --all-targets

the new command includes --workspace --all-features

but cargo test --workspace --all-features --doc is fine

[slender belfry]

can you change it to cargo test --all-targets to see if that really hits the cache in our current setup?

[bitter sparrow]

will do

actually tbh i'll try cargo test --workspace --all-targets first. we only have one workspace so there should be no difference

will try a few configurations tho (change commit -> empty commit)

realized that clippy is recompiling too, it's just way faster than test

because it doesn't need to produce a binary

        run: cargo clippy --workspace --all-targets --all-features -- --deny warnings

[bitter sparrow]

this still recompiles...

all the CI runs for the PR seem to be pulling from the same cache

Cache restored from key: Linux-(ffa9cf99a 2024-06-03)-test-9d2c396dd6346a52c23fcf9d3d30f3533e6b2c25ccd0477a4d22ba88992e775e-60db438780d04b1e73edaa22633702f4b51bd392256300103bbf383b4a87f279

Cache restored from key: Linux-(ffa9cf99a 2024-06-03)-test-9d2c396dd6346a52c23fcf9d3d30f3533e6b2c25ccd0477a4d22ba88992e775e-60db438780d04b1e73edaa22633702f4b51bd392256300103bbf383b4a87f279

the key format seems to be: <OS>-(<CARGO_VERSION>)-<JOB_NAME>-<IDK>

ok from the source:

        key: ${{ runner.os }}-${{ steps.cargo-version.outputs.cargo-version }}-${{ inputs.cache-group }}-${{ hashFiles('**/Cargo.toml') }}-${{ hashFiles('**/Cargo.lock') }}

[slender belfry]

Oh, so all jobs share a cache?

[bitter sparrow]

so it's a hash of (runner, cargo --version, job-id, Cargo.toml, Cargo.lock)

job-id can be manually overwritten when you use the action

it doesn't hash the actual command we're running (cargo test --etc)

so every time i change the command... it re-uses the outdated cache for that job

[bitter sparrow]

all runs of the same job

that's interesting. i wonder if replacing the job-id with the actual command being run would be helpful

oh i bet what's happening

cargo test --workspace --all-targets --all-features is fine

but all of our CI runs have been pulling from an outdated cache, since at one point we changed the command but we never changed Cargo.toml or anything since then to trigger a re-cache

[slender belfry]

OOOOOOOOO

smart!

so we can close the PR and do a cargo update, which would be good anyways

[bitter sparrow]

we could save the exact cargo command being run into an environment variable

then pass that through to the cache action

so that if we change the command in the future, it would trigger a re-cache

[slender belfry]

Don't think that's necessary since we don't touch the commands again

[bitter sparrow]

true but it's a footgun for editing the CI.yaml

[slender belfry]

Hmm

Fair enough

Actually, is there a wait to recache on changes to the workflows?

That would be enough

[bitter sparrow]

ah

hm, i think that would require an upstream change to Leafwing-Studios/cargo-cache, if it's possible

the only input we have for when a re-cache occurs is the cache-group: input, which is what i was calling job-id

this should still be different for each job that needs its own cache

[prisma delta]

Yeah feel free to PR changes

[bitter sparrow]

oh actually it isn't strictly required. we can pass in cache-group: <workflow file hash>-<job name>

if upstream could do that automatically it would be better ofc

[slender belfry]

Please do, that would save others the trouble 🙂

(if it's not too much work)

[bitter sparrow]

yeah i'll investigate whether it's possible

i could make it hash .github/workflows/**/*

as for knowing exactly which file in there called the action.. idk

that would make any workflow fully re-cache when any other workflow changes

tbf workflows changing doesn't happen often, but that still sounds undesirable for an upstream feature

[bitter sparrow]

confirmed that this worked

Test finished in 1:37

${{ github.workflow_sha }} seems promising, will test how it works

[bitter sparrow]

nope! i found something that seems to work though

[slender belfry]

Really? Sounds like expected behavior to me tbh

[bitter sparrow]

it means if i change release.yaml it'll reset the ci.yaml cache for example

[slender belfry]

Yeah that’s slightly crappy, but better than the current behavior imo

[bitter sparrow]

i found a way to hash only the specific workflow file but it's a bit of a hack

[slender belfry]

I'd upstream it

[bitter sparrow]

i take ${{ github.workflow_ref }} which is for example TheBevyFlock/bevy-template/.github/workflows/ci.yaml@refs/branch/main or w/e

then i strip away the repository prefix TheBevyFlock/bevy-template/ and the git ref suffix @refs/branch/main

leaving just .github/workflows/ci.yaml, and then we can hash the file at that path

[slender belfry]

That's clever 👀

[bitter sparrow]

https://github.com/Leafwing-Studios/cargo-cache/pull/17

[bitter sparrow]

since this will probably be fixed upstream, what should i do for bevy-template in the meantime?

[bitter sparrow]

i suppose i can remove workarounds from ci.yaml and then make a cargo update PR like you suggested here

https://github.com/TheBevyFlock/bevy-template/pull/137

this also pulls in the new release of blake3 with the recompile fix

[bitter sparrow]

hm on the other hand, this would prevent sharing a cache between CI and release workflows (assuming that was possible before?)

i mean ig they're not doing the same type of builds in our case anyways, but in other cases that may be a downside

[slender belfry]

I think that's fine, GitHub has a generous 10 GiB of cache anyways

Wait, per that thread someone made recently, release does not benefit from any cache at all anyways