So tired because regressions after each update !!!

The really problem with Unity is the regression at each update ! : Performance, Lights, Shaders, etc…
I pass my time to search if I made something wrong since 12 years, and each time it’s a Unity regression problem.
I’m really tired of being slowed down by these problems !

So, with last 2022.1.20f1 / URP / ANDROID, it’s the same problem than : Unity Issue Tracker - [Mobile][URP] All lights are emitting light from one light spot when the Platform is set to Android or iOS

After the fix, we start again… and if I sent to every regression I found, I can’t imagine how many problem reports I would have posted in 12 years.

So to get through this regression, just switch from Forward to Deferred, but that’s really a bad thing because the mobile performance in Deferred are bad.

I’m posting here because I’m tired of it.

2022.1f and 2022.2b bugs/Regressions or missings features are not listed by Unity :

- Ray tracing artifacts around objects in HDRP (under certain lighting circumstances or when not using the SSAO, GI or Lightmap)
- Raytracing no longer reflects the part of objects that has a different material (on an object with more than one material) in HDRP (regression) Only the first material is taken into account.
- The additional light bug on URP on Android (regression)
- Raytracing does not reflect transparent materials (I tried everything)
- Not in HDRP, but the SMAA is totally broken with the FSR in 2022.2b URP
- An transparent Object is not visible through another transparent object in HDRP (works fine in URP)
- When you using OnSceneGUI(), all changes in inspector appear only when the mouse is on Scene View (No more updated like before) (Regression)
- Some GUI draw ignore Colors (Regression)
- New Light bug with nearest or FSR upscale in 2022.2.12b on Android + Deferred (with bloom, all lights are broken) (Regression, not present in 2022.2.10/11)
And other things that I don’t have in mind right now.

2 Likes

I wonder, given how much pain Unity upgrades are giving you, what is the strategy you follow when you decide to upgrade a project to a newer version?

2 Likes

Before, I make a backup, after I go back to the old version…

This is not a real solution, because to take advantage of new features, you often have to wait a year or more after the final version (two years for LTS), but as in the case mentioned earlier, the same problem can come back several years later…

Moreover, on Issues Tracker the problem is normally corrected in version 14 of the URP… This is no longer the case, since with the latest beta version of 2022.2, the problem is back.

While I understand your frustrations I don’t think there’s a real solution to be had aside from slower release cycles both from Unity’s perspective and yours. Which would just lead to the problems they had before they started this practice of releasing major updates multiple times per year.

1 Like

I’d argue that saying this ignores the core of the problem. We have slower release cycles but it feels like the problem is only getting worse, even between minor version updates and even LTS updates. There needs to be a more definitive internal strategy when it comes to handling the development of the engine and performance testing.

5 Likes

We’re only having slower release cycles from the perspective of the previous system but I was thinking more along the lines of what we had back in the 5.x and earlier days where we received one massive update every few years.

I’d argue that wasn’t extremely the case either, largely because the releases back then were often extremely large as far as what changed between versions. For instance, 4.2 → 4.3 brought us a whole new 2D workflow and a pretty substantial overhaul to animations, we got a whole new UI system in 4.6… I’d argue the pace stayed largely the same and did end up slowing, it’s just it’s easier to see when releases happened thanks to the changing of the version numbering.

I said it before and I’ll say it as many times it takes: the problem is that Unity doesn’t make games.

“But Unity has close relationships with some game studios! They don’t need to waste resources making games!”

That’s not enough: it only catches bugs/regressions after they made into production releases.

External studios have negative incentive to be using Unity betas/alphas in production (it’s too risky and unstable) or to be updating their Unity version regularly (too disruptive the larger your team is).

If something breaks after an upgrade, the studio will be the one losing money rolling back, looking for workarounds, filling bug reports, and sometimes even delaying their launch while they wait for a fix from Unity. This means studios which make in-depth use of Unity are unlikely to be providing enough data points (feedback, bug reports) on betas/alphas, TECH releases, and even the first few LTS versions (as result of bad LTS stability reputation).

It’s obvious whatever internal testing processes Unity uses have been getting less reflective of “real world usage” over the years. Combined with the “hardcore devs” reluctance in using their own time and money to test pre-release Unity software, the visibility of possible bugs and issues becomes limited until it’s too late and lots of devs are affected.

Unity needs their internal testing to better simulate real world usage. This means having one or more groups of people using Unity on projects that are significantly more complex than just demo scenes, build at reasonable times and run well on their target platforms, and are maintained after completion.

We almost had that.

8 Likes

And I’ve said it before and I’ll say it again: it doesn’t fucking matter if Unity makes games or not when the fundamental problem is that management will not allocate resources where they’re needed in the first place and increasingly fragments the development of packages and engine components themselves. Making games is not some sort of panacea for internal operations and what we’ve heard from former employees shows that that’s where the problems truly lie.

2 Likes

Even if they fix their management, not using their software as it’s used in the real world before releasing will still cause major bugs to make it to production.

You’re right in that the current state of mismanagement won’t even allow Unity to get to that point. It’s like a 50 person game studio trying to make a 100 hour open world RPG while having a two-person QA department and expecting the developers testing their own code is “enough”.

Anyway, the problem is that when a final version comes out, we must have a debugged product…
But when I see that the Alpha version 2023.1 does have the bugs of the final version 2022.1 and the version 2022.2 beta, I find it mind-blowing…
Each time, I look for solutions to circumvent the problem.
I’m not a beta tester to report all the bugs every time.
In the change log, compared to the versions of 6/8 years ago, there is almost nothing… However, there are a lot of bugs or regressions to fix.
I’ve loved Unity since my debut in 2009 on Mac in its version 2, but I’m getting tired of looking for problems that slow me down and that aren’t mine.
And it’s getting worse and worse…

So, before starting a new version (2023), finish the previous final version.

Take a look at the next major update 2022 (2022.2 beta), in scene view shading mode, all views are broken the same on Mac, Raytracing (HDRP) does not reflect objects with more of a material…And all bugs of the final version (2022.1) are here too ! I don’t know if I want the next version…

I gotta agree. It’s been increasingly annoying how, even without extreme changes to the engine on my side or using any 3rd party binaries, the engine just… feels so unstable.

But the worst part is the TIME it takes for them to fix an issue.

-WEBGL build still has NO SOUND(fixed in the past, issue came back a few updates later, still here)

Yes, we can consider that the LTS is the “Final” edition (and I’m nice), the final version is a Beta and the Beta is an Alpha version…

And at the same time, the future version is in alpha (preview?)

As a result, versions still not finished, fewer major updates, much less fixing in the changelog and reporting of bugs to be corrected, always more regressions and the price of the pro version which will increase…

Find the mistake?!

1 Like

2021.3LTS has been the most frustrating release so far. Unusable inspector in .3 and .4, and then Unity refused to perform GetComponent in various, seemingly unconnected edge cases. Seems like my main scene got corrupted, rebuilding it so far fixed GetComponent with the same exact code, I was troubleshooting the issue for weeks and couldn’t find anything - no reports anywhere, can’t reproduce in other scenes or fresh projects. I thought I was going nuts for a moment there. These issues, together with the recent Unity news, have really taken a toll on me. Never again will I start LTS project before .20 release unless I’m forced to by some SDK requirements.

Not that waiting helps much, they’ve been actively porting confirmed 2021 regressions to 2020LTS as well.

2 Likes

2022.1.20f1 (and 2022.2b) + Android plateform + URP

Additionnal Lights bugs exemple…

First light + Forward : OK !

Second Light + Forward : OK !

First + Second light+ Forward : WTF ?

First + Second light + Deferred : OK !

I love it !!!

1 Like

About this one, what is the max lights per object in the urp settings you use for android?

New Project, simple scene, defaut values !

1 Like

It’s definitely gotten worse ever since 2020, especially with 2021. Something has changed. We’ve had speedy updates ever since post 5.0, so it can’t just be due to that.

If your “ultra stable” LTS version is constantly breaking, something is wrong.

3 Likes

Better exemple : 1 light per object… same Issue :

Same issues, with Mobile Cfg…

@DevDunk ,Do you have your answer ?
1 directional light, 2 additionnal max in VR (Quest 2), but you can push (when it work) up to 8 per object on good mobile, depend geometrie