|
History
| Log In
|
|
|
|
|
|
I've had a similar problem with 1.19.1 (4) and a couple RC versions prior and Vista. Unfortunately I dont think my crash logs ever made it to Linden Lab as the crash logger would error out when restarting the viewer. My problem is this: when starting the viewer, on occasion Vista would freeze up and only a power cycle would fix it. Often I would need to remove the battery and leave it unpowered for 30 seconds or more before the computer would even attempt to boot. During the freezup, the screen appears normal but the keyboard and mouse are not responsive at all, but the drive activity led flickers occasionally. This usually happens when starting the viewer just before it displays the login screen, but has happened at least twice when restoring an already logged-in viewer that has been minimized. I do have the latest video drivers installed that are available for my system. Prior to the last couple RC viewers, my system was quite stable and never had any problem such as this.
Details from the help screen: Second Life 1.19.1 (4) Apr 2 2008 11:59:37 (Second Life Release) You are at 286506.8, 271777.7, 381.0 in Shantos located at sim4483.agni.lindenlab.com (63.210.158.133:12035) Second Life Server 1.20.0.83892 CPU: Intel Core 2 Series Processor (1994 MHz) Memory: 3070 MB OS Version: Microsoft Windows Vista Service Pack 1 (Build 6001) Graphics Card Vendor: NVIDIA Corporation Graphics Card: GeForce 8600M GS/PCI/SSE2 OpenGL Version: 2.1.2 LLMozLib Version: [LLMediaImplLLMozLib] - 2.01.14330 (Mozilla GRE version 1.8.1.13_0000000000) Packets Lost: 1442/577533 (0.2%) Pasting this from the SLDev mailing list, from davep@lindenlab.com:
I think this is being caused by the extra impostor updates that are happening in 1.20. Unstable render pipe + more passes through the pipe = more crashes. Try disabling avatar impostors and see if that helps. It looks like some part of the render pipe is leaving a vertex data stream enabled for reading, and subsequent calls occasionally read off the end of the stream. I have work in an internal branch (maint-render) that protects vertex data streams using LLVertexBuffer (removes willy nilly calls to glEnable/DisableClientState, glClientActiveTexture, and glDrawRangeElements). Time will tell if this is the culprit. Moving forward, we'll be looking at removing all GL dependencies from the newview project and creating an API agnostic interface to the graphics card. This is an early step in that direction, hot on the heels of the use of gGL.begin/gGL.end vs. glBegin/glEnd. I feel this flicker I reported may be related to the crashing. Others have reported similar flickering and momentary freezes or random crashes like I have experienced.
Could someone affected by this bug please try disabling avatar imposters and report what affect this has? It's our current leading hypothesis (see davep email, quoted above) though we haven't been able to repro the bug.
I've imported the ticket internally and attached it to the branch 1.20 QA flow. Are you sure this is a Vista only issue? I crash quite a lot with 1.20 on XP, and that just by rotating the cam around my avatar a bit. I have the feeling this is even worse when SL is still loading textures right after the login:
Login, textures still loading (lots still grey), rotate by alt-click and mousemove -> crash. Tillie -- we're seeing this for XP as well in the crash reports. The common element is that the crashes all occur in the nVidia driver dll itself -- frustrating for crash research because we don't get a call stack.
Please try turning off avatar imposters and note any change in crash frequency. Thanks, Periapse I turned off avatar impostures as suggested and have not crashed in over an hour since logging on. Seems your suspicions might be correct.
I'm not sure if this is similar to my issue, but the moment I tried running 1.20, it immediately crashes when I login. I see my avatar for about 2 seconds and then it just crashes and sends an error report.
I may reinstall later and try again, but it seemed pretty consistant to me. This is XP Pro SP2 with a Nvidia Geforce 8800 GTS running 169.21 (latest at time of posting). Ok this is more than just a software or driver issue: this burned out my LCD screens power supply entirely the other night. LL now owes me 8 bucks plus shipping and handling.
turning off avatar imposters did not prevent the camming crash from occuring.
I too am having this issue, I tried turning off the Avatar Impostures, and had good results for over an hour, then it locked up and crashed again. When I hard booted my machine, I brought SL back up and it locked again inside of 5 minutes. I am also having the same issue in the normal viewer since the update. If anyone knows of a good 3rd party client to use, please let me know, as I am now flushing my concierge money down the perverbial toilet.
Second Life 1.20.0 (84432) Apr 9 2008 03:59:09 (Second Life Release Candidate) CPU: Intel Core 2 Series Processor (2399 MHz) Memory: 6142 MB OS Version: Microsoft Windows Vista Service Pack 1 (Build 6001) Graphics Card Vendor: NVIDIA Corporation Graphics Card: GeForce 8600 GT/PCI/SSE2 OpenGL Version: 2.1.2 LLMozLib Version: [LLMediaImplLLMozLib] - 2.01.14354 (Mozilla GRE version 1.8.1.13_0000000000) Also having this issue here. Turning Avatar Impostors off has not helped. The issue appears to be much worse in areas with a large number of avis, as I can go an hour or two without crashing if I avoid busy places. Worth noting though, that aside from the crashes - framerate and overall performance seems to be improved immensely from the last version - but the crashes are so frequent I've been forced into using the release version again. The crashes seem to happen only when I'm rotating/zooming the camera - and as Tillie mentioned, especially if the area is still rezzing and filled with grey objects.
Second Life 1.20.0 (84432) Apr 9 2008 03:59:09 (Second Life Release Candidate) CPU: Intel Core 2 Series Processor (2200 MHz) Memory: 2047 MB OS Version: Microsoft Windows XP Service Pack 2 (Build 2600) Graphics Card Vendor: NVIDIA Corporation Graphics Card: GeForce 8800 GT/PCI/SSE2 OpenGL Version: 2.1.2 LLMozLib Version: [LLMediaImplLLMozLib] - 2.01.14356 (Mozilla GRE version 1.8.1.13_0000000000) I experienced the same sort of crash on my first and only time using this Dazzle/RC version: a black screen for several seconds, followed bya notification from OS (Windows Vista) that the graphics device driver had crashed. My video system is also nVidia.
Second Life 1.20.0 (84432) Apr 9 2008 03:59:09 (Second Life Release Candidate) CPU: Intel Core 2 Series Processor (1994 MHz) Memory: 2046 MB OS Version: Microsoft Windows Vista (Build 6000) (note: this is Vista Home Premium) Graphics Card Vendor: NVIDIA Corporation Graphics Card: GeForce 8600M GS/PCI/SSE2 OpenGL Version: 2.1.1 LLMozLib Version: [LLMediaImplLLMozLib] - 2.01.14329 (Mozilla GRE version 1.8.1.13_0000000000) The crash occurred about 5 minutes after logging in, when I was doing nothing in particular -- standing still and engaging in an IM conversation in an otherwise empty sim, or nearly so. Certainly no one else was even remotely near or in view, so the _rendering_ of avatar impostors should not have been a factor. I wasn't even moving, or cam'ing around. The secondlife.log file doesn't show anything extraordinary at its end -- it just cuts off abruptly in the midst of logging a series of INFO: process_improved_im events. Nope, was too premature in saying that turning off Avatar Imposters solves problem. This definitely does NOT solve it.
Now also getting C++ runtime error message reported when the client crashes after the video driver restart is reported by Vista. Also, there seems to be no crash logger activity when I restart the client after that.
I decided to trip down all my settings and see how long I would make it inworld, after about 40 minutes of stepping from Low to Ultra and back again, I decided to do a little building, as this is when the lock happens for me most of the time (That may be because I spend a considerable amount of time doing that, I don't know.) I upped my object mesh from the Low settings and these are the graphic settings I had right before I locked up today. NOTE: Atmospheric Shaders and Avatar Impostors were unchecked.
Foock, I posted here because a number of people were reporting the total locking of the system. I need to hard boot my machine when this issue comes up. I do not know what the root cause is. My machine just stops working and does not return. One moment, I am able to interact with the client, and the next I cannot. No movement, no mouse, nothing...just a static picture of where it stopped working with no access to windows or the task manager. If this is an unrelated issue, I apologize. I just do not with to record a duplicate problem, as they sounded related. I also am having the warning flickering before this event occurs.
No apologies necessary, Foock. I'm pretty new to the jira, so I'm just getting my feet wet. I don't know all the protocol yet. Graphically, the screen appears to flicker when moving before the event occurs. I have also noticed, while building, it seems when I pan the camera around what I am working on, the focused prim/object "lags" in it's rotation slightly as compared to the rest of the world. I don't know if this is a related issue or not. I am not getting any lines or bars on the screen though. I am working with an SLI rig as well. I don't think my specs listed before reflect that.
Foock, I see some errors relating to Second Life, but logged as many times as this has happened. (ex. Faulting application SecondLifeReleaseCandidate.exe, version 1.20.0.18896, time stamp 0x47fca255, faulting module SecondLifeReleaseCandidate.exe, version 1.20.0.18896, time stamp 0x47fca255, exception code 0xc0000005, fault offset 0x0064a4e5, process id 0x658, application start time 0x01c89c2d0016c921.). Is this what you are talking about? I don't usually play around in the event viewer, not sure what I'm looking for.
I am so happy this issue has been reported. Prior to being forced to go to RC1.20, I have had a very stable viewer with almost no hard crashes. But as soon as I upgraded to 1.20, the problems stated.
- hard crashes randomly at some sims but completely repeatable crashes in others. a crash of my Video driver not responding. Basically, the SL screen goes black then an error comes up "Display driver nvlddmkm stopped responding and has successfully recovered". - in these same sims (example the Holodeck Dance Club) I also get constant cycle of freezes of the viewer of about 2 second internvals. This is NOT LAGGING.... my viewer actually screen freezes. Freeze...release...freeze...release...etc.. - jerky camera motion operation. and at times, if I move the camera angles around - the graphics driver not responding occurs. I have done some investigation on the root cause as one of the sims that constantly crashes / freezes me has a mega prim sphere on it. When I removed the megaprim, the freezing stopped. This was not a problem with any of my previous versions of the Viewer. But I noticed that even with the megaprim removed, I still am crashing on this building skybox sim (yesterday I got the graphics card not responding crash about 15 times). this skybox has a lot of sculpties on it but it doesnt seem to be their simple presence - possible the increased # of them. I have an HP Pavilion Laptop (brand new) with Vista and running the latest driver on a NVidia GeForce 8400 graphics adapter. This "graphics driver not responding" error only happens when in secondlife and although since I started SL in January I have had a couple of these errors, in RC1.20 it has become a horrid part of life. tonight I will try turning off the imposter and the shading to see if the problem goes away. PLEASE FIX THIS!!! Thank you all for the detailed and useful input. As this crash spans several nVidia cards, and internally we still don't have a repro, your observations here are our best clues for diagnosing and fixing the problem.
I'll keep this group up to date on the progress of the investigation. | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Bug
Reopened
Showstopper
1.20 causes fatal driver instability with some nVidia drivers
You might want to pass the following to the internal developers. It looks like LLRenderPass::pushBatch() may be sending corrupt/bad data to the GL library. On Solaris, we're using Nvidia's GL library and that basically a recomiple from their Linux drivers so I don't think the crash I'm encountering is local to Solaris.
In any case, once I get logged in, the viewer immediately tips over. Here's the tail end of the log:
2008-04-10T12:42:00Z WARNING: dispatch: Ignoring unknown message EstablishAgentCommunication
2008-04-10T12:42:03Z INFO: setStartupState: Startup state changing from STATE_PRECACHE to STATE_WEARABLES_WAIT
2008-04-10T12:42:03Z INFO: setStartupState: Startup state changing from STATE_WEARABLES_WAIT to STATE_CLEANUP
2008-04-10T12:42:04Z INFO: setStartupState: Startup state changing from STATE_CLEANUP to STATE_STARTED
2008-04-10T12:42:04Z INFO: idle_startup: Doing first audio_update_volume...
2008-04-10T12:42:04Z INFO: idle_startup: Done first audio_update_volume.
2008-04-10T12:42:04Z INFO: updateGeometry: WL Skydome strips in 26 batches.
Program received signal SIGSEGV, Segmentation fault.
Here's the stack trace:
0xfd0c1760 in _nv001457gl () from /usr/lib/libGLcore.so.1
(gdb) where
#0 0xfd0c1760 in _nv001457gl () from /usr/lib/libGLcore.so.1
#1 0xfd13cfbe in _nv001457gl () from /usr/lib/libGLcore.so.1
#2 0x00000008 in ?? ()
#3 0x00000441 in ?? ()
#4 0x00030101 in ?? ()
#5 0x0c7eaea0 in ?? ()
#6 0x41e756f5 in ?? ()
#7 0x3e6ac6f6 in ?? ()
#8 0xfd0c1740 in _nv001457gl () from /usr/lib/libGLcore.so.1
. . .
#167 0x085291e4 in LLRenderPass::pushBatch (this=Cannot access memory at address 0x449
)
at ... /newview/lldrawpool.cpp:440
Previous frame inner to this frame (corrupt stack?)
Since I don't have the source for libGL I'm stuck here.