Orbiter-Forum  

Go Back   Orbiter-Forum > Orbiter Space Flight Simulator > Orbiter Beta
Register Blogs Orbinauts List Social Groups FAQ Projects Mark Forums Read

Orbiter Beta Topics related to Beta releases of Orbiter and Orbiter development.

Reply
 
Thread Tools
Old 03-19-2016, 04:07 PM   #136
Eduard
Orbinaut
Default

Quote:
Originally Posted by jedidia View Post
 Can you be more specific? does the crash happen in a place where you could reasonably expect some terrain loading and unloading going on? To me, the whole thing still smells of a memory leak in the terrain engine, but I could easily be wrong.
Good question. No I do not believe it's terrain loading related problem, because I also experience the problem when doing mid-course corrections half away, when being at the boundary of Earth's major SOI, still very far away from the Moon. I warp time to that point to get more stable data in TransX.
So the first thing I asked myself was if it could be a TransX-problem. But completely disabling TransX does not solve the problem.

And the crash does not seem to happen when I haven't don any engine- or RCS-actions with the selected vessel after starting a scenario.
When I orbit the Moon I and warp time, I really hear a lot of harddisk reading and see indeed some slowing down because of it (I have downloaded all hi-res textures), but this does not crash my Orbiter process. And I believe this part has been programmed very well now and tested very well by everyone.

I think it may be related to someting with celectial bodies or gravitation, but not the terrain thing or surface textures.

So It has to be an Orbiter core-problem indeed. Not occurring with my Orbiter 2010 instance, but it is something new in the Beta.
I am running Windows 10 and I have an GeForce GTX 750 Ti videocard.

Last edited by Eduard; 03-19-2016 at 04:25 PM.
Eduard is offline   Reply With Quote
Old 03-19-2016, 05:00 PM   #137
jedidia
shoemaker without legs
 
jedidia's Avatar
Default

Quote:
And the crash does not seem to happen when I haven't don any engine- or RCS-actions with the selected vessel after starting a scenario.
Interesting... I have no thruster action at all going on in my tests. I wouldn't suspect a connection between the issues if it wasn't for the exact same symptoms described.

Might help if you post the scenario and exact steps for reproduction, so I and others can try to confirm.

Quote:
And I believe this part has been programmed very well now and tested very well by everyone.
Oh, it's not about the quality of the coding. It wouldn't be C++ if it wouldn't be leaking at some point during development
jedidia is offline   Reply With Quote
Old 03-19-2016, 07:17 PM   #138
jarmonik
Beta Tester

Default

Quote:
Originally Posted by jedidia View Post
 ...with terrain resolution set to 128 in D3D9 client.
That will cause the system to run out of memory, 128 is a way off being anything practical at a moment. With that setup surface vertex data alone can take 1GB of memory and that's too much for 32-bit apps. With tile resolution setting 32 the total vertex data consumption is something like 25-50MB.

It's recommended to keep the tile resolution at 32 until we have a better LOD control.
jarmonik is offline   Reply With Quote
Thanked by:
Old 03-21-2016, 11:44 PM   #139
Eduard
Orbinaut
Default

Update:

I think I have found out something new.
I experience this at least when using the D3D9Client (which uses better graphics and thus uses more memory = creating bigger risk for crashes when there are memory leaks, I think). But I cannot say for sure that the client is the (only) cause.

What I found out is the following thing:
When disabling surface elevation (nice to know there is a setting for it so I can turn it of if I like), the number of crashes appeared to decrease.
After re-enabling it, I noticed the number of crashes increased dramatically.
And now I probably now why: disabling the checkbox "Surface elevation" also defaults the setting to "linear interpolation" (after restarting the Orbiter Launchpad) automatically.
This is strange, because this is not the same default as the default when starting Orbiter the first time (or after removing Orbiter_NG.cfg), because that default is "cubic interpolation".

And that setting "linear interpolation" appears to be much more unstable than "cubic interpolation", with much more crashes when being close to Earth or Moon.

I don't have that scenario of the transfer between Earth and Moon anymore. I have to recreate it. I don't know if the CTD there has the same cause as this more elevation related CTD. It has to be tested again.

Last edited by Eduard; 03-21-2016 at 11:47 PM.
Eduard is offline   Reply With Quote
Old 03-22-2016, 11:23 AM   #140
martins
Orbiter Founder
Default

Quote:
Originally Posted by Eduard View Post
 Update:

I think I have found out something new.
I experience this at least when using the D3D9Client (which uses better graphics and thus uses more memory = creating bigger risk for crashes when there are memory leaks, I think). But I cannot say for sure that the client is the (only) cause.
That would be quite an essential piece of information. Can you try without the D3D9 client to see if you can reproduce the crashes? If it is reproducible, can you post a scenario?

Also, what makes you think that the problem is down to a memory leak? Do you crash due to running out of 32-bit address space? (which has been observed before, though this happens without memory leak, but because of over-enthusiastic application of memory-greedy render settings).
martins is offline   Reply With Quote
Old 03-22-2016, 10:50 PM   #141
Hlynkacg
Aspiring rocket scientist
 
Hlynkacg's Avatar


Default

Just a heads up, found a possible new bug

Code:
double man_pitch = GetManualControlLevel (THGROUP_ATT_PITCHUP, MANCTRL_ROTMODE, MANCTRL_JOYSTICK);
Moves with the joystick x-axis even when linear control mode has been selected
Hlynkacg is offline   Reply With Quote
Old 03-22-2016, 11:23 PM   #142
Eduard
Orbinaut
Default

Quote:
Originally Posted by martins View Post
 That would be quite an essential piece of information. Can you try without the D3D9 client to see if you can reproduce the crashes? If it is reproducible, can you post a scenario?

Also, what makes you think that the problem is down to a memory leak? Do you crash due to running out of 32-bit address space? (which has been observed before, though this happens without memory leak, but because of over-enthusiastic application of memory-greedy render settings).
Ok, I have tried some things again now, I have created three scenariofiles and tested them. I have uploaded this files here:

> Eduard.scn: Standing on brighton beach, when using timewarp for a while at high speed, I get a CTD in the D3D9Client.

> Eduard1.scn: Transfer from Earth to Moon, using DG-S. When repeating I sometimes experience a CTD, but it's difficult to reproduce.

> Eduard1A.scn: A modified version of the previous scenario: I replaced the DG-S by Shuttle-A (and checked very well that I did it in the correct way).
Now I get much more CTD's, when using the D3D9Client with setting "linear interpolation".
To reproduce: 1. move the vessel with the RCS. 2. Killrot. 3. timewarp 1000x or 10000x.
With 1000x, I am still very far away from the moon and I already get a CTD.

I have tried to reproduce exactly the same CTD with the stock client orbiter.exe, but I do not see it there anymore.
But I still think it's very strange that the problem occurs much more frequently with the Shuttle-A version than with the DG. So I am not shure if we can say that it's only a problem of the D3D9Client.

And finally, I also noticed this: Sometimes when coming very close to the moon, the stock client also crashes sometimes, at least when using "linear interpolation". This could be a purely terrain related problem of course. I don't know. But it's important to know that this also happens with the stock client.

I hope this information helps you.
Attached Files
File Type: scn Eduard.scn (2.9 KB, 7 views)
File Type: scn Eduard1.scn (4.1 KB, 0 views)
File Type: scn Eduard1A.scn (4.1 KB, 3 views)
Eduard is offline   Reply With Quote
Old 03-23-2016, 12:22 AM   #143
jarmonik
Beta Tester

Default

I suppose, you could zip and post the D3D9ClientLog.html from /Modules/D3D9Client/ right after a CTD has occurred. It might give some ideas about what's going on.
jarmonik is offline   Reply With Quote
Old 03-23-2016, 01:12 AM   #144
Eduard
Orbinaut
Default

Quote:
Originally Posted by jarmonik View Post
 I suppose, you could zip and post the D3D9ClientLog.html from /Modules/D3D9Client/ right after a CTD has occurred. It might give some ideas about what's going on.
Ok. See the attachment, I have zipped the D3D9ClientLog.html together with the Orbiter.log here.
Attached Files
File Type: zip Logfiles.zip (10.2 KB, 3 views)
Eduard is offline   Reply With Quote
Old 03-23-2016, 01:43 AM   #145
jarmonik
Beta Tester

Default

Quote:
Originally Posted by Eduard View Post
 Ok. See the attachment, I have zipped the D3D9ClientLog.html together with the Orbiter.log here.
I get a log like that when Orbiter hits the 2GB memory limit. We would need to reduce a distant terrain resolution to bring down the memory consumption.
jarmonik is offline   Reply With Quote
Old 03-23-2016, 02:52 AM   #146
martins
Orbiter Founder
Default

I just realised that I had set the LARGEADDRESSAWARE flag for orbiter.exe, but not for orbiter_ng. This may explain why you get crashes with the D3D9 client, but not with the inline client. You could verify that by running the external D3D7 client, which should then be afflicted by the same problem.

I'll compile both orbiter.exe and orbiter_ng with the LARGEADDRESSAWARE flag for the next beta to see if that cures the problem (at least until somebody also hits the 4GB ceiling).
martins is offline   Reply With Quote
Thanked by:
Old 03-23-2016, 03:26 AM   #147
martins
Orbiter Founder
Default New Orbiter Beta Released (r.53, Mar 23 2016)

Change log:
  1. Both orbiter.exe and orbiter_ng now have been compiled with the LARGEADDRESSAWARE flag to provide a larger address space.
  2. DeltaGlider: re-instated the atmospheric autopilot interface on the 2D panel (currently not yet accessible in VC)
  3. Camera: ground mode should now again refer to local elevation instead of mean radius
  4. Scenarios: Dissolved the 2010 Edition folder, added more 2016 Edition scenarios, including a re-recorded "Shuttle to ISS" tutorial (but need to finish annotations)
  5. Bug fix: vessel-specific MFD modes did not load from scenario file
  6. Bug fix: Playback scenarios: docked assemblies are now rigid relative to each other
Point 1 might have an effect on the problem in the preceding discussion.
Point 3 should address this issue.

This is the last update before the holidays. After Easter, I hope to upload the first release candidate fairly soon.
martins is offline   Reply With Quote
Old 03-23-2016, 03:33 AM   #148
Loru
Moderator
 
Loru's Avatar


Default

Any plans of updating particle emmiters per this thread?
Loru is offline   Reply With Quote
Old 03-23-2016, 04:16 AM   #149
jarmonik
Beta Tester

Default

There appears to be very rare CTD originating from elevation interpolation after all. Crash occur due to reading memory from offset -258 to pelev. I checked the interpolation mode after the crash and it was set to 'linear'. This occured with Orbiter rev. 52
Attached Thumbnails
ctd-2.png  
jarmonik is offline   Reply With Quote
Old 03-23-2016, 10:49 AM   #150
martins
Orbiter Founder
Default

Ok, thanks, I'll check it out. Shouldn't be too difficult to find. The 258 offset sounds like I subtracted a padding edge twice or subtracted padding from an unpadded pointer.

Quote:
Any plans of updating particle emmiters per this thread?
Yes, but I think I'll leave it for a later update. If I keep adding features this release isn't going to happen, and i wanted to take advantage of the (relative) lull in bug reports. I hope this is down to the beta getting stable, but it's probably just because everybody is getting sick of bug reporting.
martins is offline   Reply With Quote
Thanked by:
Reply

  Orbiter-Forum > Orbiter Space Flight Simulator > Orbiter Beta


Thread Tools

Posting Rules
BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
Forum Jump


All times are GMT. The time now is 09:26 AM.

Quick Links Need Help?


About Us | Rules & Guidelines | TOS Policy | Privacy Policy

Orbiter-Forum is hosted at Orbithangar.com
Powered by vBulletin® Version 3.8.6
Copyright ©2000 - 2018, Jelsoft Enterprises Ltd.
Copyright 2007 - 2017, Orbiter-Forum.com. All rights reserved.