How's stability in 2019.x?

How's stability in 2019.x?

Basically I want to know if I need to upgrade my plugin to 2019 ASAP, or can wait for a bit.

 

I haven't done a huge stability or performance test but our 2018 branch worked right away in 2019. And the automatic tests plus our manual smoke tests are green. I need todo a full performance regression test though

 

I'm less interested in performance and more in crashes and bugs. I.e. whether I should target 2019.x now or still stay on 2018.3. 2018.3 introduced few troublesome changes of its own and I still haven't taken them all into account.

how long since you switched to 2019.x?

 

Haven't done any real playtests in 2019 only our manual smoke tests, a list of things we know have broken before when moving between major versions. Those didnt catch anything.

Though to be 100 procent sure I would need a few hours of actual playtesting.

Edit: installed 2019.1.0 day one and are now on 2919.1.1

 

Time traveler detected.

Alright, thanks for the info.

 

The base platform seems stable. New features of course still have instability. What are you interested in upgrading for?

 

2019.1.1 is stable for me so far, but I don't think any preview packages are

 

Or my brain moves faster than my fingers

 

I have a scene converter plugin which is linked in my signature, and I just wanted to know if it would be a good starting to target 2019.x instead of 2018.3. I.e. whether this is a platform a lot of people would switch to and not something disastrous like one of the 5.0.x updates which broke animation state machines and such.

That's the rough idea.

Maybe I'll jsut give it a whirl later and see how bad it is. 2018.3 screwed up some code that relied on PrefabUtility, and that part is still in smoldering but barely working state. (-_-)

 

They will. Just like from 3.x to 4.x to 5.x etc.
If you really worry, If I were you, I would just wait for 2019.3 or 2019.4, where 2020 phase will enter.
Other than that. Just try, test and see what will happen.

 

Fair point.

 

Some useful informations for you :

https://forum.unity.com/threads/share-your-unity-2019-1-0f2-experience.670816/

https://forum.unity.com/threads/unity-2019-1-is-now-available.661885/page-4

Stay on 2018.xx at the moment....

 

I will translate: "stay away from the 2019, because the forty-third public variable gets serialized wrongfully after inspector scrolling".

I know this is a painful bug for you (we all have those, believe me), but this is ridiculous.

Obviously there are bugs, but nothing crucial, the base version is stable, unless you use multiple dozen member variables in your scripts exposed to the inspector. Which you shouldn't, so you're probably safe.

And people started to use 2019.1 and the number will grow steadily. I use 2019.1 since the first alpha (I created my current project back then) and I don't have any major issues lately (only with some preview packages, but that's separate concern).

 

No crashes on Windows Standalone so far. Building for Android is broken for me, bug reported and it's been reproduced by Qa. Waiting for the fix to see if Android runs crashfree too.

 

In all honesty, I dealt with scenarios where I needed HUGE number of inspector fields, so this stuff is a big deal for projects that rely on on scriptableobject-based data.

Based on responses so far it looks like the situation isn't bad, but it may be a good idea to wait for something like 2019.3 or 2019.4 before switching to it.

 

Im not sure on general stability, but I know prefabs are unfortunately still broken - they're loosing MB references.

 

Is there any additional info on that? Because one of the big features in 2018.3 was the new prefab system.

 

heres big thread of ppl reporting this:
https://forum.unity.com/threads/edi...the-object-referencing-it-is-reloaded.568120/
and heres mine
https://forum.unity.com/threads/probably-serious-prefabs-bug-causing-losing-data-references.673318/

 

Hi, I want to use the new unity 2d lighting system. Should I wait a bit or Upgrade from 2018.4 to 2019.2 right away?

 

Given that 2018 LTS isn't really stable, I don't know how stable 2019.2 would be tbh. Wish instead of new features Unity decided to make what it's already there better and without bugs. What is the point of unfinished features for years that you can't even use because you don't know when they will break again?
So I would advocate to only use LTS versions and leave the bugs to others (at least LTS should have slightly less bugs).
The only code that shouldn't work should be yours, while you are developing, and not Unity's code.

 

This post isn't really helpful without examples of what you're talking about.

 

In my experience, whatever you do, avoid 2019.2 like the plague. The compilation times in it are absurd: upwards of three times 2019.1, accurate as of 2019.2.9f1. 2019.3 is better in this regard. In fact, in a test I did, it was downright a miracle for an empty project.

 

I'm talking about me encountering bugs while I work, and about 2 crashes per day. Given that 2019 has even more features than 2018, I wouldn't want to try that.

I don't really think an example is need, but I can give you personal examples and also this, that is more or less better than my personal experience:
https://issuetracker.unity3d.com/pr...175&status=1&unity_version=2019.2&view=newest
~800 bugs being worked on for Unity 2019.2
https://issuetracker.unity3d.com/pr...175&status=1&unity_version=2018.4&view=newest
~360 bugs being worked on for Unity 2018.4

Notice that there are bugs more than half a year old, and this is only the list of bugs that appear here (my reported bugs don't appear here for some reason, probably the same with many others) and are being worked on. There are bugs that Unity can't reproduce and won't fix.

As you can see, LTS has less active bugs, and one reason for why that is that somewhat makes sense is that it has less bugs (as there are other reasons, like how many users there are using the version, or if Unity isn't validating as many bugs in that version).

In case you want to release a build in the next 6 months, given the time it takes for Unity to fix bugs, I would stay with the Unity version with less bugs, that being the latest LTS version (even thought it also has bugs and problems, but less than other versions).

Personal examples:
- Editor.log file gets filled with data until the hard drive is full. (Workaround, mark Editor.log as read-only, but then if Unity crashes you won't get any log. yay).
- Animator randomly makes Unity crash after trying to allocate unreasonable amounts of RAM (I've seen a requests of more than 12 Million TB).
- Animator's ordered interrupt doesn't work as specified in the manual, making it unusable on certain cases
- Random crashes when using git to add and remove files (Unity can't always handle files changing and reimporting them). Apparently this bug was fixed, but may not be backported (sometimes it says a bug was backported to an LTS version but it's still there)
- Lightprobes can't be selected to edit them (this one was recently fixed).
- It's not possible to rearrange parameters or layers in the Animator (this one took more than a month to get a fix).
- This one is a stupid one, but here it is too: Drag an drop an animation with any of this names to the Animator Window and it's name will be empty: "Pillow_Hit.Lethal_BD", "Pillow_Hit.Basic", "Pillow_Hit.abcde", "abcdefghij.abcde".
- Can't compile with Burst to Android.
- Random bugs I encounter and I just find a workaround quickly enough and just forget about them. One of them was that using the GL API inside the editor made Unity crash, so I replaced everything with Gizmos and it didn't crash anymore.
- Upgrading my project from Unity 2018.4 LTS to Unity 2019.2 isn't even possible, Unity freezes

 

HELL YEAH, THIS IS THE GOOD S***. This is the exact kinda post I like to see when people talk about stability issues because:

• It's in depth
• It contains steps to reproduce the issues
• It includes references to potential regression issues

This is like the ideal "here's some S*** that's busted in Unity" post.