BUST: 1.1.X Crashes (Your Help Requested!)

[smjjames]

Just now, J.Random said:

@smjjames, do you want me to get banned or something? I won't start another "release" rant here.

lol.

Not trying to, no, just saying that he wouldn't be able to do a 'debug build'. Though some sort of mod that taps into the logs and provides more logs or something MIGHT be feaseable. After all, modders do their own logging for mods, so.....

Probably going to have to wait until everybody gets back on Monday and they have a chance to discuss it.

[sackfalte]

I have the SPH symmetry issue with my heavily modded RO/RSS/RP-0 install. It will also absolutely not let me into the VAB at all, and crash. The output_log file looks identical to the ones already posted - there is a 'Crash!!!' reference, but then the file just cuts off.

I just made a second install, clean, stock, and tried to replicate the issue for 10 straight minutes, but KSP wouldn't crash. 

 

This is on a Windows 10 computer, playing KSP_64, GTX 980Ti, 16 GB RAM.

This would of course point to some mod causing the issue. I won't have the time to single the culprit out until next week though :/

 

Here is the link to the log file: https://www.dropbox.com/s/rcezop58k9t9tup/output_log.txt?dl=1

Edited May 19, 2016 by sackfalte

[smjjames]

7 minutes ago, sackfalte said:

I have the SPH symmetry issue with my heavily modded RO/RSS/RP-0 install. It will also absolutely not let me into the VAB at all, and crash. The output_log file looks identical to the ones already posted - there is a 'Crash!!!' reference, but then the file just cuts off.

I just made a second install, clean, stock, and tried to replicate the issue for 10 straight minutes, but KSP wouldn't crash. 

 

This is on a Windows 10 computer, playing KSP_64, GTX 980Ti, 16 GB RAM.

This would of course point to some mod causing the issue. I won't have the time to single the culprit out until next week though :/

Are there any exceptions happening before the crash? Maybe you could post up the output log anyway. But narrowing down the cause of that specific crash is a good idea.

[sackfalte]

2 minutes ago, smjjames said:

Are there any exceptions happening before the crash? Maybe you could post up the output log anyway. But narrowing down the cause of that specific crash is a good idea.

Here is the output_log file:  https://www.dropbox.com/s/rcezop58k9t9tup/output_log.txt?dl=1

Exceptions that I found (and which were displayed in the ExceptionDetector):

 

1
(Filename: C:/buildslave/unity/build/artifacts/generated/common/runtime/UnityEngineDebugBindings.gen.cpp Line: 64)

MissingMethodException: Method not found: 'PopupDialog.SpawnPopupDialog'.
 

2
KSP-AVC -> System.InvalidCastException: Cannot cast from source type to destination type.
  at KSP_AVC.AddonInfo.ParseVersion (System.Collections.Generic.Dictionary`2 data) [0x00000] in <filename unknown>:0
  at KSP_AVC.AddonInfo.GetVersion (System.Object obj) [0x00000] in <filename unknown>:0
  at KSP_AVC.AddonInfo.ParseAvc (System.String json) [0x00000] in <filename unknown>:0
  at KSP_AVC.AddonInfo..ctor (System.String path, System.String json, RemoteType remoteType) [0x00000] in <filename unknown>:0

[smjjames]

Looks like you actually entered the editor, but then something happened. Though theres a whole bunch of spam of stuff from TestFlightInterop and ullage thing, mostly over something called 'TinyTim'. But I have no idea if that's normal or not.

[MarvinKosh]

DxDiag

KSP Log

What I tried:

Stock KSP 1.1.2

Career mode

• Built a couple of test rockets using the HECS AI control core and launched them.
• Both rockets had fairings so that I could make the probe contained inside without worrying about aerodynamics. Between the fairing and the probe was a separator/decoupler so that the probe's engine could detach from the base of the fairing.
• The probe on rocket 2 was fairly unremarkable: four solar panels attached at the top of the HECS core, an Oscar fuel tank under the core, a small reaction wheel, two more Oscar fuel tanks, then an Ant liquid engine, then a small separator, followed by an adaptor to 1.5m, a fuel tank, a medium reaction wheel, fuel tank, and a Terrier engine. Pretty standard stuff.
• When I got bored with rocket 2's mission I reverted back to the VAB.

Events leading up to the crash:

• I clicked on the fairing, intending at first to remove it. The fairing was of course connected to all the stages underneath, so they moved with it.
• Then I changed my mind and moved the fairing back to where it was, but I did not actually click to place it.
• Apparently I changed my mind again and moved the fairing away from the point where it would have attached.
• The crash:
• I clicked to the side of the rocket, intending to leave it there for the time being while I modified the probe, but the game crashed when I did this.

My hardware: Fairly old gaming rig with newer bits including AMD Radeon R7 200 Series, no overclocking or crazy cooling systems.

Operating system: Windows 10.

Graphics driver: AMD 16.15.2211-160321a-301028C (16.3.2) which is the latest stable driver.

Other notes: I tried to access AMD Radeon Settings after the crash to get the driver version, and that crashed.

Crash logs, in case they prove useful.

[Claw]

12 hours ago, Gaarst said:

I managed to build a whole rocket without crashing after updating my drivers

My drivers are rather quite old at the moment, and I'm somewhat fearful of updating them...since everything seems to be working. 

 

And no, I can't just "release a debug build." Though if/when things get narrowed down some, I will look into building an add-on to track any relevant information (for those who wish to participate).

[AlamoVampire]

@Claw the going theory in the thread i pointed at to you is its a memory call issue which may be native to unity. there is a rumor unity is going to upgrade with a patch or something <not sure how it works for game engines to be honest> to that extent in June if I remember correctly. Will you guys be using the updated engine as supposedly it will address things like the memory thing and possibly wheels? I know when <before I had to frankly give up on KSP at the moment thanks to this crashing> most of my crashes in 1.1.0, 1.1.1 and 1.1.2 all showed ntdll.dll as a contributing factor or something like that.

[Claw]

3 hours ago, AlamoVampire said:

the going theory in the thread i pointed at to you is its a memory call issue which may be native to unity

That seems like a valid hypothesis. The thing that gives me pause is the fact that some folks are having no problems at all. So I'd still like to track down what's different or why some people are running into this with high frequency while others don't seem to experience it. The latter implies there might be a way to work around it in the mean time.

We're also aware that there are some unity updates that we'd like to investigate for upgrading. But since the team has been scattered on vacation, there's no significant news to report.

[AlamoVampire]

@Claw since i am no programmer, i can only mention what I um "hear" on here lol. BUT, again, for me, there is no sense of rhyme or reason to it, its entirely random and entirely debilitating to my ability to do anything of significance, which is why ive not touched KSP since april. I have had crashes in the VAB, SPH, pad, ascent, orbit, trans soi <read: going anywhere away from kerbin, mun, minmus, anything really> and once in the space complex screen, the one with all the buildings that practically do nothing in sandbox lol. I hope my hypothesis is of use to the team, I really do. Any chance the hypothesis can be brought up to all of the user base via a devnote or somesuch to try to focus on that area? if not, thats how it is, just a thought  

[saggerjagger]

hi been lurking these forums since 1.1 released to try find fix for kerbals constant crashes but mine is now fixed and running perfect. not one crash in over 2hrs!  

which is amazing considering it would only run for a max 5 min before a crash in vab or some times as soon as i clicked it. tried formating hardive and reinstall windows twice running bare minimun and was still the same. but on my last try before smashing the keyboard i rolled back the drivers on steam and then re updated again before realising ive never updated my bios before so did that. 

tried kerbals and its flawless now, so happy! no slowing down, freezes or crashing. im letting gforce experience run setting on optimised and looks amazing now.

specs are fx4100, asus M5A78L-M/USB3 motherboard, 8 gig ram and gtx750ti

just thought would write incase helps anyone, dont know if was just updating bios or rolling driver back and forth but i can finally enjoy this game again.

need anymore info on anything let me know

[mikegarrison]

I finally just started playing KSP again now that 1.1.2 is out, and I'm about to stop because of the CTD problem I am having in the VAB.

I'm running Win 7, 64-bit, 16 GB RAM. In the output logs the crashes seem to be happening right after the log says the game is unloading unused objects to save memory.

[AlamoVampire]

@mikegarrison This adds support to a strong and growing even stronger hypothesis that its a memory call or something like that issue in unity. See my thread  @Clawhas linked at the start of this query thread, it goes into detail. 

Edited May 21, 2016 by AlamoVampire

[mikegarrison]

4 hours ago, AlamoVampire said:

@mikegarrison This adds support to a strong and growing even stronger hypothesis that its a memory call or something like that issue in unity. See my thread  @Clawhas linked at the start of this query thread, it goes into detail. 

I had the Windows performance monitor running and it logged a "hard fault" in the memory monitor each time I had the crash.

[AlamoVampire]

@mikegarrison which lends MORE credibility to the memory issue, which I THINK is a Unity issue. Would you mind tossing that observation into my thread on the subject? 

[severedsolo]

9 hours ago, mikegarrison said:

I had the Windows performance monitor running and it logged a "hard fault" in the memory monitor each time I had the crash.

A hard fault shouldn't cause any issues though. It just means Windows has temporarily moved something stored in memory to the hard drive, and KSP wants it back. This happens all the time.

I wonder though, do the affected people have reasonably small (less than 4GB) amounts of Ram? Is Unity somehow making Windows handle hard faults badly? Or is it asking for them the wrong way? Seems unlikely though because the memory leaks have had me pushing my physical memory limits, and it must have been hard faulting - and I've never crashed.

Anyway, as requested 1.1.x has literally never crashed for me, so here are my stats:

CPU: i5-2500k (3.3Ghz - 3.8 with turbo mode)
8GB DDR3 1666Mhz Ram (Crucial)
OS SSD: Samsung 840 Evo
KSP Drive: WD Caviar Blue 500GB
GPU: Factory overclocked Gigabyte GTX 460 (latest drivers through Gefore Experience)
OS: Windows 10 Insider Build 14342
 

 

Edited May 21, 2016 by severedsolo

[steve_v]

1 hour ago, severedsolo said:

A hard fault shouldn't cause any issues though.

Indeed, I'd be very surprised if it did, though it does imply suspending execution momentarily while the OS does the required I/O. Again this shouldn't cause problems.

1 hour ago, severedsolo said:

I wonder though, do the affected people have reasonably small (less than 4GB) amounts of Ram?

Not here (16GB), and I've never seen KSP cause any paging/swapping at all. This occurs even with no swap space / page file available, and generally with >4GB true free memory (not used by cache/buffers). If it were running out of RAM in this scenario it would cause the OS (GNU/Linux here) to invoke the OOM killer to free memory, and this is pretty obvious when it happens. I'm not sure how Windows would handle such a situation, but I suspect it is similar.

I have a bunch of fairly detailed stack traces (though most are from a modded install), however @sal_vager tells me he already has over 100 of these so I haven't really bothered keeping them. I can probably generate some more from the stock game (given time) if that is any use.

What I don't have is any reliable reproduction steps: As far as I can tell, these crashes are completely random.

Aside from that, here's some system info from a "does crash" position, in case someone can spot a common thread:

Spoiler

System:     Distro: Debian GNU/Linux 8.4
                  Kernel: 4.5.0-0.bpo.1-amd64 x86_64 (64 bit gcc: 4.9.2)
                  Desktop: KDE 4.14.2 (Qt 4.8.6) dm: kdm
                  Processes: 417 Uptime: 39 days Memory: 8992.2/15998.0MB
                  Init: systemd v: 215 runlevel: 5 Gcc: 4.9.2                                                                                                                                     
Mobo:       ASUSTeK model: P9X79 v: Rev 1.xx Bios: American Megatrends v: 4005 date: 04/18/2013
CPU:         Quad core Intel Core i7-3820 (-HT-MCP-) cache: 10240 KB
                 Clock Speed: 3799 MHz
                 Microcode Revision: 0x710, 2013-06-17
Graphics:  Card: NVIDIA GK104 [GeForce GTX 680]
                 Display Server: X.Org 1.16.4 driver: nvidia Resolution: [email protected]
                 GLX Renderer: GeForce GTX 680/PCIe/SSE2 GLX Version: 4.5.0 NVIDIA 361.42
Memory:   Installed Size: 16384 MB
                 Type: DDR3                 
                 Part Number(s): KHX1600C9D3/4GX      
                 Configured Clock Speed: 1600 MHz
Audio:       Intel C600/X79 series High Definition Audio Controller driver: snd_hda_intel
Drives:      HDD Total Size: 676.2GB (31.3% used)
                 ID-1: WDC_WD3000GLFS size: 300.1GB filesystem: NTFS
                 ID-2: OCZ-AGILITY3 size: 120.0GB filesystem: ext4
                 ID-3: Crucial_CT256MX1 size: 256.1GB filesystem: ext4
Network:   Card-1: Realtek RTL8111/8168/8411 PCI Express Gigabit Ethernet Controller driver: r8169 v: 2.3LK-NAPI
                 Card-2: Intel 82579V Gigabit Network Connection driver: e1000e v: 3.2.6-k

More extensive details available on request OFC.

Edited May 21, 2016 by steve_v
Stupid, stupid editor. Bungles text alignment every time.

[Kiro]

@Claw

here is another Output log from the CTD I just had. It happened inside the VAB, while testing out some parts from the Pathfinder mod. It crashed after I cleared all the current parts and upon selecting a new root part.

Output_log

Hope this helps.

___

And here is another crash report. While building out the launch system for my base, I clicked on an engine in the menu and then immediately deleted it, and then CTD.

Output_ log 2

Edited May 21, 2016 by Kiro
additional info

[Moesly_Armlis]

On 5/20/2016 at 3:57 AM, Claw said:

So I'd still like to track down what's different or why some people are running into this with high frequency while others don't seem to experience it

Are there users that do not experience this crash?

Is there any data on the percentage of users experiencing this issue? 

Should a poll be started?

Crashes for me seem to happen quite often in the VAB.  The crashes after decouple have been few.  Crash when approaching another craft only happened once.

 

[dooots]

I also have crashes in the VAB. As far as I can tell the crashes are completely random.  Time, part count, and mods do not seem to be a factor.  The only thing in common with all the crashes as far as I can remember is they happen on a mouse click action (pickup, drop, etc.).  No crash logs from any of the crashes and all output logs have no useful information (no stack trace, not detailed enough to see when the crash happened).

Only other thing of note is doing a clean reinstall of ksp may have reduced the number of crashes, but with how random they are it is hard to tell.

CPU: i3-4130 3.40GHz
Ram: 8GBs
GPU: AMD Radeon HD 7700 Series (latest release drivers)
OS: Windows 10 64-bit Build 10586
KSP: Steam 64-bit

Edit:

Scratch the all crashes happen on mouse click, had one tonight after I detached the lower stage of a rocket and went to grab a new fuel tank.  If I had to guess the crash happened as I moused over one of the parts in the part selection area.

Edited May 22, 2016 by dooots

[fourfa]

I can add a data point.  I'm on a macbook laptop (i7 2.6GHz, 650M graphics, works OK).  Recently since 1.1.2 or so, I started having behavior where it ran great at first, then over time slower and slower, then there would be some video corruption and CTD.  I wondered if it was thermal / heat soak.  Took the back panel off, cleaned out a fair amount of dust and dog hair.  Then played about 6 straight hours with zero slowdown, no loss of responsiveness, no crashes.

Surely not relevant to the VAB/symmetry crashes or many other discussions here.  Just a reminder to check airflow, particularly on laptops with their narrow flow paths.

[John FX]

My install is very stable, running a large bucketload of mods, hitting 8GB RAM usage although that is out of the 24GB installed on my system (W7 ult, 24GB I7 GTX570)

I did have two instant CTD though, both times I was uninstalling a program on my second monitor and the crash happened at the exact moment the requester popped up and the screen went dark for me to confirm I wanted to uninstall.

Just before the crash (both times) there was a lot of

HandleD3DDeviceLost
  HandleD3DDeviceLost: still lost
Skipped rendering frame because GfxDevice is in invalid state (device lost)

and then

HandleD3DDeviceLost
  HandleD3DDeviceLost: needs reset, doing it
FullResetD3DDevice
ResetD3DDevice
dev->Reset
D3Dwindow device not lost anymore
Updating vessel voxel for Probe 5 LR
 
(Filename: C:/buildslave/unity/build/artifacts/generated/common/runtime/UnityEngineDebugBindings.gen.cpp Line: 64)

Std dev for smoothing: 3 voxel total vol: 16.0754711726719 filled vol: 2.620281506364

 
(Filename: C:/buildslave/unity/build/artifacts/generated/common/runtime/UnityEngineDebugBindings.gen.cpp Line: 64)

RemoteTech: [Transmitter]: Uploading Data... (Atmospheric Pressure Scan while in space near Phobos) - 1.25 Mits/sec. Packets to go: 6 - Files to Go: 7
 
(Filename: C:/buildslave/unity/build/artifacts/generated/common/runtime/UnityEngineDebugBindings.gen.cpp Line: 64)

NullReferenceException: Object reference not set to an instance of an object
  at CityLights.LocalCityComponent.OnPreCull () [0x00000] in <filename unknown>:0
 
(Filename:  Line: -1)

Crash!!!

probably not connected with other crashes but I thought I would let you know.

Edited May 22, 2016 by John FX

[Claw]

17 minutes ago, John FX said:

HandleD3DDeviceLost
  HandleD3DDeviceLost: still lost
Skipped rendering frame because GfxDevice is in invalid state (device lost)

This particular warn/error has been around a while. These show up sometimes when the focus is away from KSP. Personally this warning hasn't caused me any troubles, but some people have reported crashes from it. Though I think it's coincidental to something else.

 

11 hours ago, fourfa said:

Recently since 1.1.2 or so, I started having behavior where it ran great at first, then over time slower and slower, then there would be some video corruption and CTD.  I wondered if it was thermal / heat soak.  Took the back panel off, cleaned out a fair amount of dust and dog hair.  Then played about 6 straight hours with zero slowdown, no loss of responsiveness, no crashes.

That's good to hear. It used to be pretty common for us to suggest that people clean out their machines. 1.1 (with Unity 5) seems to be even more stressing to some hardware. Though there are also some graphics corruption issues with OSX, I'm glad to hear you were able to solve your issues.

 

21 hours ago, MoeslyArmlis said:

Crashes for me seem to happen quite often in the VAB.

This seems to be the most common place, so certainly the first that I'm looking at. But I personally have no issues in the VAB or SPH, so I'm hoping to narrow it down even more. Even if it means using an add-on or two to do it.

 

18 hours ago, dooots said:

I also have crashes in the VAB. As far as I can tell the crashes are completely random.  Time, part count, and mods do not seem to be a factor.

Yeah, this seems to be the extent of the information so far. Totally random, and not tied to anything in particular.

 

Thanks for posting kerbonauts. I am reading, and trying to gather as much info as possible.

 

Cheers,
~Claw

[MarvinKosh]

So I noticed a commonality in error.log files under Context:

SegCs: 0x00000033

I dunno what exactly this means, but throwing it into a Google search seems to turn up a lot of results. Of course, you don't get a crash log if everything is hunky dory so that may be skewing things somewhat, and as I said, I don't know what SegCs is. But Unity does seem to be a common factor in the results I looked at.

[fourfa]

4 hours ago, Claw said:

 It used to be pretty common for us to suggest that people clean out their machines. 1.1 (with Unity 5) seems to be even more stressing to some hardware. Though there are also some graphics corruption issues with OSX, I'm glad to hear you were able to solve your issues.

The new multithreading has to be dumping more heat into more transistors on more cores.  Perhaps prior to 1.1.2 I wasn't really fully stressing it, as I started a new career with the pre-release builds and never got back into complex builds until I finished that.  Now I've acquired the tools to watch clock speed, power consumption, and temps in real time with logging.  Should help pinpoint what's going on while the game's running.