I've noticed that the session list scroll lags. An...

# confetti

[Arkadii Ivanov]

I've noticed that the session list scroll lags. And also the transition animation from the session list screen to the session details screen also lags. The cause of the latter issue might be the former issue.

[Arkadii Ivanov]

The speaker list scroll also lags.

[Stylianos Gakis]

Are you 100% sure you don't have a debug build you're looking at?
Probably not the best example, but on my pixel 7 pro these do not lag at all. I didn't try with an older phone though

[Arkadii Ivanov]

Oh, you are right! I totally forgot that I had a local build. The version from GP works fine!

[Stylianos Gakis]

Hehe been there before 😅

[yschimke]

@John O'Reilly do you have GDE minutes for FTL devices? I'd love to setup 1 single scroll benchmark run daily or weekly. I know issue above is resolved, but would be good to measure and improve.

[John O'Reilly]

What are FTL devices? I have general google cloud credit...not sure if it covers that

[yschimke]

https://firebase.google.com/docs/test-lab

[yschimke]

Using this to run macrobenchmark https://developer.android.com/studio/test/gradle-managed-devices#gmd-ftl

[yschimke]

Could automate generating the baseline profiles, and benchmark scroll performance.

[John O'Reilly]

I think the firebase stuff should be covered by same cloud project account we have. btw @mbonnin was just looking at billing since last July (when I think the $500 credit was applied)......is this what you'd expect?

[John O'Reilly]

these are the accounts btw

[John O'Reilly]

this is the breakdown of costs

[John O'Reilly]

@yschimke but to answer your question, we should be able to use that Confetti-Firebase project.....can add you to that if you're not already there

[yschimke]

I think I am.

[yschimke]

OK, I might try adding this when I have some free time. Start with a single mobile benchmark test, run weekly.

[yschimke]

Then automate baseline profile generation.

[mbonnin]

@John O'Reilly this is in line with what I expected. App Engine is mostly gone. We can probably clean up a bit of storage cost (I think old images might rot there) but there's not much we can do about networking. That's the minimum charge for the global forwarding rule.

[mbonnin]

We should start a Patreon for when the GDE credits are over 😄

[John O'Reilly]

Hopefully they'll renew it

[John O'Reilly]

Both the GDE status and the cloud credits 😃

[John O'Reilly]

We're not using App Engine at all any more?

[mbonnin]

Nope

[Stylianos Gakis]

Actually out of pure coincidence I was looking yesterday a bit into how the confetti backend is setup. Because it's an area I know nothing about, and I thought I might be able to learn a thing or two from there. Until I looked into the CI step to build the backend, and eventually reached this file https://github.com/joreilly/Confetti/blob/main/backend/terraform/main.tf and I was like holy shit and just stopped there 😅
Is there some description of how everything is setup for it, or is the setup too complicated to be able to be fit into some Readme or something like that?

I swear every time I look into what backend needs to do to setup their deploying setup, so everything besides writing the code itself of the service, I feel discouraged to pursue it further and properly learn what's going on there. Do you think perhaps Confetti isn't the right place to start to get into this?

[mbonnin]

I think Confetti is quite good as a backend tutorial actually. It’s built by a front end engineer who knew nothing about backend a couple of years ago

[mbonnin]

Terraform is the main alien thing indeed but the awesome thing with terraform is that it brings all the configuration in source control, meaning we don’t have to click in the gcloud web console.

[mbonnin]

There are a bunch of terraform tutorials out there but it’s not that hard. The main weird thing for me was realizing that terraform was storing its own GCP state outside of GCP so that now there are 2 states (that can become out of sync). But AFAIK, this is still not a solved problem. AWS has cloud formation but GCP has nothing sad panda

[mbonnin]

If you want to dig more, we can give you access to the Google cloud console. We don’t need to click there but seeing what the different resources are in the UI helps

[Stylianos Gakis]

You know what, that would actually be super useful. If I can get access in the least privileged way possible just so I can peruse around it would probably be very useful to me. And yes you're right, making terraform not look like alien script is an important step to understand what is going on.
When you say there are 2 states, how does that happen here? That Terraform file describes how things should be configured, and it tells GCP, but where is the second instance stored?

[mbonnin]

@Stylianos Gakis you’re added! (account [e-mail])
> That Terraform file describes how things should be configured, and it tells GCP, but where is the second instance stored?
main.tf is just the definition, the “class” file. The state is the “realisation” of them. Contains unique ids for resources, etc... It’s stored in 2 places
• in GCP itself
• in the confetti-tfstate GCP bucket: https://console.cloud.google.com/storage/browser/confetti-tfstate/terraform/state?pageSta[…]e%22:(%22f%22:%22%255B%255D%22))&project=confetti-349319
IIRC this is all because GCP doesn’t give API access to all its internal state so terraforms has to track it in parallel. This is a very sad state of things IMO

[mbonnin]

(configured here)