Disappointed with performance

Discussion in 'Developer Preview Archive' started by angel_m, Dec 22, 2011.

  1. Joachim_Ante

    Joachim_Ante

    Unity Technologies

    Joined:
    Mar 16, 2005
    Messages:
    3,555
    I am very curious to see a project folder with the bad android PowerVR performance.
    Please file a bug with a project folder attached that has bad powerVR performance on 3.5.

    Thanks.
     
  2. jlevel

    jlevel

    New Member

    Joined:
    Jan 23, 2012
    Messages:
    20
    I tried to sent my project, but I had some problems with the bug report, was a lot of GB and was stuck for some reason. I'll try to narrow down more and send it soon.

    Anyway, I discovered another behaviour:

    I switched "Graphics level" from OpenGL 2.0 to OpenGL 1.x and I have a huge performance boost than OpenGL 2.0 on Unity 3.5.

    It's curious because on Unity 3.4 I received better performance with OpenGL 2.0 than 1.x

    Not sure if it will help
     
  3. unity_sg

    unity_sg

    New Member

    Joined:
    Sep 14, 2010
    Messages:
    90
    Same problem here, a decrease of performance about 25% on a big project (windows)
    Unfortunately I can't send the project to report a bug
     
  4. Lypheus

    Lypheus

    Member

    Joined:
    Apr 16, 2010
    Messages:
    625
    Upgraded from 3.5b -> 3.5f and I'm seeing a huge increase in performance, using stats window i'll get around 120-400fps vs 18-70fps from the previous build (in worst case, 1772 draw calls @ 120fps with 3.5f vs 18fps with 3.5b). For comparison, the wiki fps script reports about ~60fps with new build vs ~18fps with beta.

    Demo is at: http://forum.unity3d.com/threads/123849-TPS-Nightfall-Cooperative-Survival-Sandbox-Game

    No complaints here.
     
  5. a1s2d3f4

    a1s2d3f4

    New Member

    Joined:
    Mar 20, 2011
    Messages:
    413
    Im pretty sure i dropped 30 FPS from 3.5b to 3.5f
     
  6. Tiles

    Tiles

    Member

    Joined:
    Feb 5, 2010
    Messages:
    2,244
  7. rbisso

    rbisso

    Member

    Joined:
    Jul 13, 2011
    Messages:
    23
    @Joachim - Two fixes have brought us back up to approximately our 3.4 framerate: First, we implemented the culling masks on our realtime and baked lights, as you suggested (Saved about 300-400 draw calls in our worst-case view). Second, as suggested in another thread, we temporarily disabled our blob shadow projector (which saved us another 275-300 draw calls). Thanks so much for your help and we'll be looking out for the projector fix.
     
  8. TheGeoff

    TheGeoff

    New Member

    Joined:
    May 25, 2010
    Messages:
    12
    One of the issues with sending a 3.4.2 project, at least for us, is the the 3.4->3.5 project is pretty hap-hazard. We had to do it a couple of times to get all of our tags, layers, etc to import properly. And even then some of the physics settings are off.
    Did anybody with the performance problems have 3.4->3.5 Upgrade issues? Maybe the upgrade process is where the bug it?
     
  9. Paulo Henrique

    Paulo Henrique

    Member

    Joined:
    Dec 12, 2010
    Messages:
    221
    ~30% Performance boost in a huge scene without realtime lights.
     
  10. tarragon

    tarragon

    New Member

    Joined:
    Dec 28, 2010
    Messages:
    38
    Sorry, your tests are wrong. QED: 99% of forums.
     
  11. why789

    why789

    New Member

    Joined:
    Sep 23, 2011
    Messages:
    184
    than the dynamic shadows do not project onto a surface that is baked (culled out by the realtime light...)
    Isn't that the whole point of dual-lightmaps??? realtime shadows mingling with baked shadows??

    Unity 3.5 changes it so if you are near enough to use the near lightmap and you only have 1 auto light, the realtime shadow will "overwrite" (meaning you won't see the baked shadow..) the baked shadows. Which I think completely defeats the purpose... So with that you have high quality baked shadows far away with crappy realtime shadows close.

    If you use 2 lights as suggested by Joachim Ante, one auto and another baked only, than the auto light does not project shadows onto the baked only surfaces. Defeating the purpose of dual lightmaps.

    WTF? Is anyone else noticing how backwards this is? or is it just me?
     
  12. Tiles

    Tiles

    Member

    Joined:
    Feb 5, 2010
    Messages:
    2,244
    Yes, this is backwards. But who says that improvements means to have something better afterwards :D
     
  13. why789

    why789

    New Member

    Joined:
    Sep 23, 2011
    Messages:
    184
    so now what do I do?! Could I bake the maps and than after they are baked mark all static objects as "Do not cast shadows"? Than just switch this back on before I rebake?

    EDIT: just realized, even if it doesn't cast shadows it will still light it up and that takes CPU power as real as making it too bright.
    Sigh... 3.5 officially messed up Dual-Lightmaps. Now that I think of it, Single lightmaps are also out of the picture even if I manually set up the light intensity, because the dynamic lights will still light up and cast shadows on static objects -_-
     
    Last edited: Feb 27, 2012
  14. HazeTI

    HazeTI

    New Member

    Joined:
    Jan 12, 2012
    Messages:
    83
    Joachim - We've also submitted a bug report with a test project we've created. Case 445654. Note that our problem is not with an increase in draw calls as expressed by others. The test project we submitted has exactly the same number of draw calls between 3.4 and 3.5. However, the frame time is between 2 and 3 times slower across the board (physics, animation, skinning, batching rendering) on 3.5.
     
  15. Joachim_Ante

    Joachim_Ante

    Unity Technologies

    Joined:
    Mar 16, 2005
    Messages:
    3,555
    What i said was that if you want to use different lighting rigs for dynamic vs static lighting, then you can use culling mask. If you want to use the same lighting, then you can still use a light marked as "Auto". This still works perfectly fine with forward and dual lightmapping.

    If you wish to have different lighting on dynamic and static objects, then it makes sense that there will be no shadowing information cast from the dynamic to the static objects as they are being rendered exclusive of one another.


    I believe the issue is in the setup you have. Have you filed a bug report with a complete project folder so we can take a look at it?
     
  16. Joachim_Ante

    Joachim_Ante

    Unity Technologies

    Joined:
    Mar 16, 2005
    Messages:
    3,555
    Do you have a project folder uploaded somewhere? Please report a bug and post the case number here.
     
  17. Joachim_Ante

    Joachim_Ante

    Unity Technologies

    Joined:
    Mar 16, 2005
    Messages:
    3,555
    We are investigating this specific issue. Very nice bug repro project. Thank you.
     
  18. why789

    why789

    New Member

    Joined:
    Sep 23, 2011
    Messages:
    184
    I thought dual lightmaps do not work with forward rendering, also I would like more than 1 directional light that is casting shadows, I would like multiple point lights to cast shadows as well (realtime)

    I am using deferred lighting at the moment.
     
  19. Tiles

    Tiles

    Member

    Joined:
    Feb 5, 2010
    Messages:
    2,244
    In 3.42 it does. But you need to write a shader version with enabled dualforward surface shader directive. Like i did in this thread here. Unfortunately this shader doesn`t longer work in 3.5. When i upgrade my little testfile i get a pink mesh and a bunch of warnings. Seems that the new lighting model has broken this feature.
     
    Last edited by a moderator: Feb 28, 2012
  20. n0mad

    n0mad

    Member

    Joined:
    Jan 27, 2009
    Messages:
    3,731
    Speaking about light culling, I don't know if it's intended or not, but baking lights completely ignore layer separation.
    Try this : Create 2 layers, put one GameObject in layer 1, and a light only culling layer 2. Then bake. GameObject receives the light.

    I was planning to use fake local lights to create some nice color variations on certain objects, but it is just impossible with this behaviour, as it colors the whole area. Dunno if it's intended, but I don't feel it's right.

    edit : yet another stuff I don't understand : before 3.5, creating an empty prefab and asking for its name property returned the prefab's name. Now it returns a Null. Is it on purpose ?
     
    Last edited: Feb 28, 2012
  21. n0mad

    n0mad

    Member

    Joined:
    Jan 27, 2009
    Messages:
    3,731
    I'm also encountering huge hickups in the Profiler (deep mode) whenever I try to move the timeline mark right or left.

    It's eating up to 1 Go in the windows process manager, and even crashed Unity after a dozens of freezes.



    Additionally, I'm facing far more "GetThreadContext failed" messages at Unity startup (+ Unity not starting at all, inevitably) than with previous versions (it seems like this issue is going more and more frequent as versions go up). The only positive point is that with 3.5 windows now detects the program failure, and therefore doesn't force me to shut it down via Process Manager.

    edit : the Profiler crash is a Mem leak, obviously. RAM is going above the roof in no time after I just try to navigate in the profiler timeline, even if I completely stop the player. Even if I keep doing nothing in Unity after that, the whole Editor + Profiler gets severly unresponsive.

    edit 2 : Ok I nailed the moment where it goes up : it is when I move the time mark to a place where there was a huge CPU load (900 ms, case below, when Adding AnimClips at runtime). Whenever I move the mark outside of that point, RAM and responsiveness goes back to normal.


    $Unity_profiler_memleak.png

    then :

    $Unity_profiler_memleak2.png

    then :

    $Unity_profiler_memleak3.png


    Is there any stable 3.5 somewhere ? Not being able to work with the deep profiling will get me stuck in work for some time ...
     
    Last edited: Mar 31, 2012
  22. n0mad

    n0mad

    Member

    Joined:
    Jan 27, 2009
    Messages:
    3,731
    And a third observation about performance difference with 3.42 :

    Animation.AddClip now is more than 10 times slower than with 3.42. I was adding several clips during Awake to an Animation component, it took approximately 0.3 sec on my PC, but now with 3.5 it takes more than 3 seconds (exact same clips).

    update : It's odd .... I reimported all the FBX, rebuilt all the animations (nothing changed for them since 3.4.2 of course), and now it's "only" 3 times slower than 3.4.2 (900 ms instead of 300).


    Of course during all the tests I tried to disable firewall + antivirus, nothing changes.
     
    Last edited: Feb 28, 2012
  23. Tim C

    Tim C

    Unity Technologies

    Joined:
    Feb 6, 2010
    Messages:
    1,558
    What is the bug number for this?
     
  24. Tiles

    Tiles

    Member

    Joined:
    Feb 5, 2010
    Messages:
    2,244
  25. MABManZ

    MABManZ

    Member

    Joined:
    Sep 29, 2011
    Messages:
    139
    I just went back to 3.42, my performance on Android dropped from 30-50 fps on a Galaxy S device to 15-35 fps...checked for duplicate lights and re-imported all assets, checked quality settings but nothing helps
     
  26. n0mad

    n0mad

    Member

    Joined:
    Jan 27, 2009
    Messages:
    3,731
    Last edited: Feb 29, 2012
  27. akwok

    akwok

    New Member

    Joined:
    Apr 21, 2009
    Messages:
    31
    I am also having this memory issue while using deep profiling and select time frame with high CPU usage. It crashes Unity. iMac, Lion, 4GB ram.
     
    Last edited: Mar 2, 2012
  28. n0mad

    n0mad

    Member

    Joined:
    Jan 27, 2009
    Messages:
    3,731
    Good idea to put down machine specs, mine is :
    intel core i7 920, 10go RAM, win 7 x64
     
  29. MaDDoX

    MaDDoX

    Member

    Joined:
    Nov 10, 2009
    Messages:
    697
    Exactly the same with our RageTools Pro "vilas walk" animation, as shown in our youtube video. It got 40 FPS in my machine, it now drags at 12 FPS. It's also a chore to use the profiler now, as has been noted here, but after some struggling I've noticed (through deep-profiling) that many apparently simple methods like Vector3_op.add and mult are taking waaay too long. No idea if this is related or not to the bug, just thought I'd mention it 'coz I found it really strange.

    The only good side of all this is that we're going through a completely different optimization route which should help our users in the end. It's just sad that it has delayed our release so much, I have a couple users really desperate to get their hands on the product :(

    PS.: There are absolutely no lights in our scene. It's all vertex-shaded polygons created by RageSpline.
     
  30. n0mad

    n0mad

    Member

    Joined:
    Jan 27, 2009
    Messages:
    3,731
    I'm a bit surprised to be honest with Unity's stance on these problems :

    - All threads speaking about a draw call rising are now marked as [SOLVED] by an admin, as if everything has returned to normal. A lot of people are saying they have much more DCs with 3.5, and I'm really not sure they all used this "bug workaround" Joachim talked about in this thread. There are even some users reporting that they weren't under the scenario Joachim described at all.
    - The stance of "we cannot solve your problem if you don't give us your project to test" is unacceptable in this case, professionally, because there are countless people saying their FPS dropped with 3.5. With so many people reporting performance drops, for whatever reasons Unity can find about it, there is still a major problem with 3.5. With such an amount of reports, it should mean that Unity could reproduce the problem internally. It's always better to provide a concrete package to test, but on the other hand I don't find it very friendly to make believe that nothing will be improved until everybody sends their project.

    I don't mean to put any pressure on the team, as it seems they're experiencing a lot actually (I guess big studios using Unity is part of it), but right now it seems like we just have to wait for a potential improvement. We need more communication with the team on these forums about 3.5, at least to relieve us from any possible doubt about the future.

    TL;DR : we need more feedback.
     
    Last edited: Mar 3, 2012
  31. Tiles

    Tiles

    Member

    Joined:
    Feb 5, 2010
    Messages:
    2,244
    Best would be to mark Unity 3.5 as Beta again until the patch arrives that fixes the drawcall issues.
     
  32. ragethecat

    ragethecat

    New Member

    Joined:
    Jan 19, 2012
    Messages:
    14

    We are also seeing very similar slowdowns in the Webplayer after pushing to 3.5. Increased DC as well as a loss in framerate in excess of 20 fps. We are using 2 cameras for the player, one for the environment and one for weapons. Perhaps this is an edge case Unity devs have not tested much?

    It is frustrating because we gained nothing from the switch to 3.5 unless you count the ability to burn lightmaps and occlusion without running out of memory and crashing. This seems like a rushed release with limited QA time. If and when Unreal starts running in a browser, users will begin to bail if updates continue on this path. If you are truly supporting AAA studios and games, step it up.
     
    Last edited: Mar 4, 2012
  33. goat

    goat

    Member

    Joined:
    Aug 24, 2009
    Messages:
    3,035
    I'm just generating a mostly flat ocean floor landscape and even without any assets if I leave Unity 3.5 sitting a while it won't respond when I make it the active window again as is extremely sluggish to respond. This is Windows 7 x64 with a ATI 2400 All-In-One. Similar to the way osX 10.7 freezes when I would import even simple assets in Unity 3.4.2.

    On Windows, since it didn't completely freeze I checked memory usage: 300,000K first time but only 120,000K the second.

    Mono is running and it's slugglish always. I like leaving Mono running because it is so slow. Could this cause the problem?

    I'll either have to remember to save and close (I lost my work) everytime I leave my chair or start profiling Unity.exe
     
  34. Quietus2

    Quietus2

    Member

    Joined:
    Mar 28, 2008
    Messages:
    2,061
    That sort of sounds like the garbage collector has fragmented memory to hell. If it is a mono issue, there's not a thing you can do about it unless Unity updates mono. That tends to happen historically only during major point releases.

    Besides the profiler, a good thing to do is check Mono's bug tracker for relevant issues.

    Awhile back, the community found bugs in mono which caused asynch loading and networking to be unusable in production as they randomly would fail. With a bug that mission critical, it still took 3 months for Unity to admit there was a problem and another 6 months for them to actually fix it with the arrival of 3.0 and a mono update.
     
  35. dstgermain

    dstgermain

    New Member

    Joined:
    Feb 3, 2012
    Messages:
    1
    This happens here too, I get a 100% reproducible hang, the editor becomes completely unusable and I'm forced to kill the process.

    Xeon E5620 2.4 GHz (2 processors)
    6 GB ram
    Windows 7 64 bit

    Is this problem registered somewhere with the devs on the Unity side?
     
  36. n0mad

    n0mad

    Member

    Joined:
    Jan 27, 2009
    Messages:
    3,731
    Bump ...
    So few feedback from Unity, it's scary ...
     
  37. angel_m

    angel_m

    Member

    Joined:
    Nov 4, 2005
    Messages:
    880
    I am experiencing a new and curious behaviour (maybe it has been reported already) and it is the unstable and variable (without apparently any reason,) draw calls count.
    Some times the statistics window shows about 1500 drawcalls and other day without any change in the scene (of course always with the camera in the exact same position and angle) the stats shows about 1000 drawcalls. :eek:

    P.D. The fps also drops when the draw calls are higher. At least this is logical.
     
    Last edited: Mar 13, 2012
  38. n0mad

    n0mad

    Member

    Joined:
    Jan 27, 2009
    Messages:
    3,731
    I'm adding yet another drawcall unexplainable rising case :

    This time with a simple HUD ... Using 2DToolkit, so this should always be 1 drawcall as I'm using only 1 atlas.

    Details here.
     
    Last edited: Mar 20, 2012
  39. goat

    goat

    Member

    Joined:
    Aug 24, 2009
    Messages:
    3,035
    Well after developing thew initial work on Windows and Android we have now problem after going to iOS and osX.

    This is an (really) easy lightweight app.

    Try to put the project and osX and test it on an 1st Gen iPod iOS 3.1.3 and iPhone 3G iOS 4.2.1 (both arm6/openGL ES 1.1)...

    Unity 3.5 (f5)
    osX Lion 10.7.3
    xCode 4.3.1
    iOS SDK 5.1

    Target: tried different combinations but make sure subset of net 2.0, arm6, iOS 3.1.3, slow and safe

    1. Well the game just hangs after being copied and set to running on both platforms
    2. So I try it on the Android on the MacMini. Works great.

    So I try demo projects that I know worked in December:

    1. ChickenLord's Shader Bundle Asset 'Cube' Demo: hangs on both after working fine in December
    2. Unity's own StarTrooper Demo: hangs on both after working fine in December

    So I'm like OK, that was Unity 3.4.2 I'll downgrade but in the mean time osX has had a couple of minor upgrades to 10.7.3, iOS SDK is now at iOS 5.1, and xCode is not 4.3.1!!!

    And I did publish a simple preliminary test (non Unity) app to the app store today so I know the toolchain from Apple works.
    .
    .
    .
    one last thing: on a couple of the projects (ChickenShader's Cube StarTrooper), xCode reported Unity aborted app startup because of something fundamentally simple to check: it says:

    You are using Unity iPhone Basic. You are not allowed to remove the Unity splash screen from your game

    I'm like I haven't replaced the splash screen.

    This happened once or twice in 3.4.2. All the time in 3.5

    0x8614ec: bne 0x008614d8 ; checkBytes(std::string, std::string, bool) + 100 at VerifyiPhoneSplashScreen.cpp:58
    0x8614f0: cmp r1, #0
    0x8614f4: bne 0x00861508 ; checkBytes(std::string, std::string, bool) + 148 at VerifyiPhoneSplashScreen.cpp:68
    0x8614f8: ldr r0, [pc, #20]
    0x8614fc: add r0, pc, r0
    0x861500: bl #1967568 ; 0x00a41ad8 printf_console at LogAssert.cpp:659
    0x861504: bl #5707836 ; 0x00dd2d48 abort
    0x861508: sub sp, r7, #8
    0x86150c: pop {r4, r5, r7, pc}
    0x861510: subseq r6, lr, r0, ror #11
    0x861514: ldrsheq r8, [r7], #-112

    Just to be sure. I deleted the Unity generated iPhone-Unity project and re-saved it (earlier I just 'replaced' it) with the same results.

    So BECAUSE all of these apps are 'hanging' as if on an assert and not 'coring' I think this is the case for every single one of the apps I couldn't get tested on iOS today.

    I always knew Assert was lame. Please fix this.

    If not this, any other ideas?
     
  40. goat

    goat

    Member

    Joined:
    Aug 24, 2009
    Messages:
    3,035

    As far as this goes: I am replacing the Windows machine in June. There is something (I believe it is Office 10, it said installed 'caching' SW????) that is killing this machine and just a simple check of memory and CPU usage doesn't show anything obvious.
     
  41. goat

    goat

    Member

    Joined:
    Aug 24, 2009
    Messages:
    3,035
    OK, so I found out this has caused a lot of trouble for others as well. I'll just co-install xCode 4.3 as xCodeOLD.app as recommended until a Unity update is made.
     
  42. supernat

    supernat

    Member

    Joined:
    May 3, 2011
    Messages:
    39
    I had a drop from about 40 fps on Droid X to 11 fps, and the reaction time to input is highly sluggish now. The game I'm working on is in the infancy stage and only consists of a single opaque background texture and a single texture (transparent) for all menu items using EZ GUI. I tried reducing the atlas texture sizes to 64 pixels with no noticeable increase in frame rate, so it could be a fill rate problem or just something in the unity engine eating up frame time. I just submitted the bug report with my original 3.4 project.

    I don't think it's an issue with my PC, but just for clarification, I have:
    Intel 980X OC'd to 4.2GHz, 12 GB ram, dual GTX 460, Windows 7.

    I haven't had any issues with the editor that others are experiencing. I know many of you are frustrated, but the massive new features and improvements in 3.5 will long outlive the few bugs that exist right now. If you have a project that needs to get out the door right now, why not go back to 3.4 and wait for a stable 3.5. I think it will definitely be worth it in the end.
     
  43. rendermat

    rendermat

    New Member

    Joined:
    Mar 30, 2012
    Messages:
    1
    We discovered the same issue: extreme performance issues on Galaxy Nexus with PowerVR.

    Our project uses no lights - all lightning is lightmapped. Vertices are at all times below 20k. Drawcalls below 40.
    Profiler shows nearly no impact by scripts, no physics burden etc. 90% of our performance is eaten up by graphics.
    We tested this by deep profiling and completly deleting all other influences (scripts etc.).

    We still get low framerates.
    Still our game runs smoothly on the Samsung Galaxy S2 (MaliGPU) and the Asus Transformer (Tegra2).
    On the Galaxy Nexus however, the game is laggy beyond playability.

    Android performance with 3.5 seems to be unacceptable. Even on the iPhone3S it should be possible to run 20k vertices with 40 Drawcalls.
    Textures are few (as expected by this low drawcall count), compression is PowerVR (or TXT for the Transformer) optimized, and resolution is low (1064). We only use one lightmap to further reduce drawcalls. 90% of all shaders are mobile unlit. No non-mobile shaders in use at all. No alphas (they seem to have clipping and z-sorting issues) and no transparancies. No garbage collection allocation. So where does this lag come from????

    Moreover we observe regular spikes in the profiler - even in empty projects (category CPU others: setRenderTexture.Aktive().
    The Main Thread runs in 1.9ms on an MacBook Air but at spikes takes up to 4ms. We also abserved an connection between those spikes and Keyboard input - even if there is no script answering to those (project empty): Spike at key down and key up. What's going on there? I mean 4ms for and unused keypress in an empty project??? While following the forums we also encountered lot's of posts dealing with these peaks.

    We will file a bug report and attach our project sonn. Thanks for your help.
     
  44. Joachim_Ante

    Joachim_Ante

    Unity Technologies

    Joined:
    Mar 16, 2005
    Messages:
    3,555
    The framerate issues have been tracked down to 3 sources:

    1) A bug in the Projector, causing a big increase in draw calls. This is fixed in 3.5.1

    2) An intended change in behaviour of how lightmapped static objects or affected by realtime lights. This fixes a bug but some "incorrect" lighting rigs will get increased draw calls. This can be fixed easily in the project. Details are here:
    http://forum.unity3d.com/threads/116479-Disappointed-with-performance?p=833498&viewfull=1#post833498

    3) A performance regression in the skinned mesh renderer on some iOS and Android devices. This is due to some driver issues with multithreaded skinning. This is fixed in 3.5.1.

    4) On Android specifically Adreno chipsets have very slow dynamic VBO performance. We added a specific driver workaround.

    If you have a performance regression and it doesn't apply to those 3 cases, please file a bug and post the case number here.

    We are in the final stages of testing 3.5.1. It will ship soon.
     
  45. hippocoder

    hippocoder

    Digital Ape Moderator

    Joined:
    Apr 11, 2010
    Messages:
    11,532
    Can't disagree hard enough. Have you debugged something as big as unity? No, you haven't. Short of an MMO you won't find anything this complicated on so many different platforms. You also have to debug it for deferred and forward rendering. The more reports unity gets, the better big picture they get.

    Submitting a bug report is ESSENTIAL. Unlike other software you might have used with bug report tools (microsoft, anyone?) - unity really does need these reports to work with. Yes I submitted physynth as well as a bug report.

    10 forum users complaining about slowdown does not equate to unity figuring it's a problem since you can slow it down by changed behaviour, or any other number of things. It could be shaders. It could be skinned meshes. It could be any number of things on any number of platforms.

    Submit a bug report, it helps everyone who uses unity.
     
  46. n0mad

    n0mad

    Member

    Joined:
    Jan 27, 2009
    Messages:
    3,731
    I submitted 2 bug reports, and even wrote TWO full step-to-step repro case, Hippo. One here, and one other here. You can't be more precise in how to reproduce. Yet zero feedback. How come you say that Unity needs tons of different projects to "understand" the bugs when it's so easily reproducable ... ? And for those 2 cases, it would have been far less an evidence for their QA if I just sent my project and said "there's a bug, it doesn't work. Go figure it. Thanks" (caricature).

    What about the Profiler systematic slowdown/crash ? Does it also need tons of projects to be sent to see that whenever you roll over a high peak, it slows down to the point of being unusable ? A Pro feature ? (reminder, all the people who reported that bug in these forums were having the same issue, at the same time, under the same conditions).

    It took me a lot of time to explain in every detail what setup I was using in all those cases, so that it could be 100% reproducable.
    I don't even freakin know if the 2 x 2 hours of my time I took to clearly write those repro cases were worth it, as I don't even know if anyone at Unity read it. It's fresh, and the thread was at the top list for an entire week. Joachim even answered another person in this very thread, and didn't give any feedback about the ultra detail repro case I created. Not even a "it's supposed to work like this", or "we're looking at it"... It was simply ignored.
    This looks like some sort of punition, I don't know.
    But right now you can understand my reaction. Anyway I'm patient by nature, so that's not such a big deal to wait and see for me, and I won't hold any unnnecessary grudge. But clearly there's a communication problem right now.

    edit : Also, Unity lately telling people that chatting is useless is in complete contradiction with this sticky.

     
    Last edited: Mar 31, 2012
  47. Joachim_Ante

    Joachim_Ante

    Unity Technologies

    Joined:
    Mar 16, 2005
    Messages:
    3,555
    Hi n0mad,

    When i said. File a bug report. What i mean is using Report Bug.app and creating a bug report that includes a project folder so that we can reproduce the issue. I am happy to help you out, but I need actionable data to work on.

    The reason others got replies is because they filed a full bug report with sample project and it was possible for us to reproduce the issue.

    Once you have filed a bug report with your issue, please post the case number here.

    Some more info on how to file a bug:
    http://blogs.unity3d.com/2009/04/15/bug-reporting-and-you/

    (If you have already filed bug reports, please just post the case number here)
     
  48. n0mad

    n0mad

    Member

    Joined:
    Jan 27, 2009
    Messages:
    3,731
    Hi Joachim, thanks for your answer.
    However, I'm not sure to understand why I should file a project folder when the two bug reports I'm talking about are full repro cases (case 452620 case 454280, they can be found in these forums too), starting from an empty scene ?

    They were bugs that I spotted using my project, yes, but for convenience reasons I stripped all the noise (= my 1.7 Go project) down to just a fully independant, simple repro case.
    Isn't it better ? My project got a ton of interdependancies, and I just don't want to send the whole 1.7Go package over the web just for a bug that can be reproduced with a simple, independant guideline.

    This is kind of why I'm not very satisfied with the way things are handled actually, as it is clearly said that no bug will be looked at if it doesn't contain a concrete project file. But you didn't even look at the bug itself, and if it was needing a project file to be inserted at all ? Kinda odd ... :/
    Anyway I'm not expecting as much as the CTO to have a look at those cases, I understand you got more important bugs to look at, but at least someone from Unity to give a bit of feedback on actual bugs...
    There's not even a beta bug correction list, we don't even know what's being worked on or not ...
    Put it bluntly, it's either "give your project, or shut up and wait".

    edit : I got a perfect example for what I'm trying to say :
    I just found an unexpected behaviour with OnDestroy() method, but nothing in the docs gives me a hint about it being to be expected or not. So thinking the behaviour doesn't reflect the definition given in the docs, I conclude it's a bug, and reports it's a bug :

    case 455705 :

    What am I supposed to think ? Is it intended ? Did I really spot a bug ? Before bombing your QA with reports that could possibly not be bugs, isn't it understandable to search for a confirmation by communication first ?
     
    Last edited: Apr 3, 2012
  49. Wozik

    Wozik

    New Member

    Joined:
    Apr 10, 2009
    Messages:
    662
    n0mad, you have a valid point.

    But there are two things I'd explain. First, since as of recent report we have a million unity users, you can only guess how much reports we get. So we implemented a rating system with which the bugs that gets reported with repro projects, screencasts and attachements get much higher priority and don't get lost in the noise. Hence the multiple requests for a repro project. Not because we're not able to create script with a OnDestroy() function on DontDestroyOnLoad() objects.

    I've verified your case is not reproducible with the 3.5.1 and replied with the following message:

     
  50. angel_m

    angel_m

    Member

    Joined:
    Nov 4, 2005
    Messages:
    880
    I have tested my project in 3.5.1 and the performance is exactly the same as with 3.5.
    At least it is not worse...