These are chat archives for highfidelity/hifi

30th
Mar 2015
Judas
@judasshuffle
Mar 30 2015 14:27
sandbox is down o.o
Leonardo Murillo
@murillodigital
Mar 30 2015 14:27
@judasshuffle thnx for reporting, checking
@judasshuffle pls try again
Judas
@judasshuffle
Mar 30 2015 14:29
thanks its back :)
David Rowe
@ctrlaltdavid
Mar 30 2015 16:45
Is Interface crashing on start-up about 80% of the time a known problem that's being worked on?
Windows build 2179 often crashes at start-up for me (in sandbox and hq)
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 16:50
^^ same - what I see is it crashes first couple of times no matter where I start then is fine from then on. Have yet to capture it doing so while running from debugger. Once it starts normally it seems to be fine from then on for hours at a time - go do something else then it crashes on start once or twice and is ok again. It - seems a little better since rebuilding with this morning's GIT commit. (Win 7)
nevermind on seems better - was luck. Trying to get it to do so while running in debugger - but so far I've never been able to, only wants to crash on a release build compile.
Andrew Meadows
@AndrewMeadows
Mar 30 2015 18:17
Anybody got a callstack on that Windows crash during login?
I wonder if it is related to something I just fixed in PR#4529
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 18:23
If there was a delay leading to a retry on getting AV could that cause what #4529 aims to address? It's so intermittent and difficult to catch. I got crash once today but had opened in release build instead of debugger.
Andrew Meadows
@AndrewMeadows
Mar 30 2015 18:29
Yes, I think it is possible that #4529 could crash on a race condition where the avatar's skeleton wasn't properly loaded before the first step of the physics simulation.
So even if the body's model was valid, it might still crash (in theory).
David Rowe
@ctrlaltdavid
Mar 30 2015 18:41
It doesn't crash with my debug build; no call stack, sorry.
Andrew Meadows
@AndrewMeadows
Mar 30 2015 18:46
well, #4529 merged already so maybe that particular intermittent Windows crash on login will go away
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 18:49
Makes sense - had a ton of rain over weekend and cable got unhappy - was having crappy net and long ping times. Problem was worse then and I could see a null ptr ref but wasn't setup then to get stack trace. Today net is betterish and problem lesser.
also the debug builds with all extra console output are slower starting up maybe giving some extra time for "things" to happen vs release builds.
David Rowe
@ctrlaltdavid
Mar 30 2015 18:59
Windows build 2181 which includes "fix crash for bad avatar body on login" still crashes at start up at about the same rate. : (
Both with and without scripts running.
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 18:59
yes
David Rowe
@ctrlaltdavid
Mar 30 2015 19:01
Clearing cache and deleting interface.ini didn't help. (Crashed even before logged in after deleting interface.ini.)
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 19:08
Getting a shutdown crash in btCollisionWorld.cpp as well
Andrew Meadows
@AndrewMeadows
Mar 30 2015 19:19
ok @ctrlaltdavid, that means the crash you're seeing is unrelated. Hence the callstack would be helpful since it would provide some clues.
@OmegaHeron where are you located when you get that shutdown crash? I'll try to reproduce.
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 19:20
LAN connected domain you can't get to. Let me try from my public space to see if it does same.
@AndrewMeadows - Enter domain Heron - close Interface get crash. I'm on Windows 7. Unhandled exception at 0x01978723 in interface.exe: 0xC0000005: Access violation reading location 0xDDDDDE01. Breaks at line 93 of btCollisionWorld(). I wish I had a clue how to save call stack in VC - never thought I'd say I miss gdb.
Andrew Meadows
@AndrewMeadows
Mar 30 2015 19:25
ok, I'll try it
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 19:25
I'm at GIT pull from last commit (close of 4499)
Andrew Meadows
@AndrewMeadows
Mar 30 2015 19:26
Hrm... I need to update and rebuild.
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 19:26
It was doing so from 4529 pull as well
AndrewMeadows @AndrewMeadows starts build and goes to lunch...
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 19:40
@AndrewMeadows when back - I have the startup crash caught in MS-VS. I'm leaving it at that point and maybe I can get some info for you. Call stack at url below
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 19:47
Enrico Speranza
@Vytek
Mar 30 2015 20:09
2183 on Windows 8.1 64 now stable....??? No :-(
David Rowe
@ctrlaltdavid
Mar 30 2015 20:29
@AndrewMeadows Start-up crash unfortunately doesn't happen with debug builds for me, so no call stack for that.
Shutdown crash on the other hand ... crashes in PhysicsEngine::~PhysicsEngine() at the call to delete _dynamicsWorld.
shutdown-crash.png
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 20:31
I managed to capture a call stack for it by using release build with debug - never could get it to happen on a "pure" debug build.
David Rowe
@ctrlaltdavid
Mar 30 2015 20:32
Ah, that's the call stack you pasted above? Great!
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 20:32
have it frozen in the debugger now at the startup crash still
yeah that's it
David Rowe
@ctrlaltdavid
Mar 30 2015 20:33
I forgot about RelWithDebInfo ... might make that my default Release build
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 20:34
yeah - I was going nuts trying to figure out how to capture - it's never happened in debug build - I stumbled across RelWithDebug and got lucky
David Rowe
@ctrlaltdavid
Mar 30 2015 20:43
@OmegaHeron I just reproduced the crash in RelWithDebInfo, with the same call stack as you.
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 20:43
cool!
David Rowe
@ctrlaltdavid
Mar 30 2015 20:44
For the record, line 615 of the crash is: _currentRotation = transform.getRotation();
Andrew Meadows
@AndrewMeadows
Mar 30 2015 20:50
No, sometimes the debugger claims it is line XYZ when in fact the real cause of the crash is at XYX-1. Such is the case here, I believe.
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 20:50
Glad you got exact repro on it - I was afraid it was a one off
Andrew Meadows
@AndrewMeadows
Mar 30 2015 20:51
What is the value of _ghostObject in that context? 0x0 or something else?
I'm rebuilding the RelWithDebug config
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 20:53
hopefully it'll crash for you - I'm a nub at VS debugging - I get unable to read memory when trying to inspect _ghostObject
Andrew Meadows
@AndrewMeadows
Mar 30 2015 20:54
You're up to date with upstream master?
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 20:55
yes
Andrew Meadows
@AndrewMeadows
Mar 30 2015 20:55
Hrm...
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 20:55
It's crazy intermittent - I can launch dozens of times with no problem - then it'll hit
Andrew Meadows
@AndrewMeadows
Mar 30 2015 22:30
the crash on shutdown is 100% reproducible. I'll fix it first.
Andrew Meadows
@AndrewMeadows
Mar 30 2015 22:56
Woot! I crashed on login!
Joe Large [Omega Heron]
@OmegaHeron
Mar 30 2015 22:58
\o/
It's weird how it's usually just the first couple times I run interface after a system startup or re-compile. Then it's fine.
Andrew Meadows
@AndrewMeadows
Mar 30 2015 23:16
crash on shutdown is fixed here: #4532
still working on crash on login
David Rowe
@ctrlaltdavid
Mar 30 2015 23:18
Note that Philip's just merged a PR from Seth, "fix a couple of crashes".
Andrew Meadows
@AndrewMeadows
Mar 30 2015 23:31
Ah yes, he fixed that one, but not quite correctly. Mine will have a merge conflict so I guess I should resolve that.