Multiple crash regressions when upgrading from 1.08.15 to 1.09.05

Hey, we recently upgraded our FMOD library from “10815” to “10905” and we’re getting several in-dll FMOD crashes.

A few bits of relevant information:

  • These crashes are all in Windows.
  • All our sounds are using FMOD_ACCURATETIME and FMOD_NONBLOCKING modes.
  • These crashes are fairly uncommon - About several thousand crashes for several hundred thousand players over the span of a day.
  • These crashes did not occur in FMOD 10815.
  • I can provide PDB files over email.

If necessary, I can provide more information over email:
syang@roblox.com

Thanks!

Crash 1:
.dmp file: http://spotcos.com/fmod_crash/crash1/crash1.dmp
Stacktrace:
fmod!FMOD_File_SetDiskBusy+0x484c
fmod!FMOD_File_SetDiskBusy+0x1fca
fmod!FMOD::DSPConnection::setUserData+0x271b
fmod!FMOD::AsyncThread::wakeupThread+0x7c1f
fmod!FMOD::ChannelControl::stop+0x3892
fmod!FMOD::AsyncThread::wakeupThread+0x8520
fmod!FMOD::AsyncThread::wakeupThread+0x7e25
fmod!FMOD::ChannelControl::setPaused+0x4d
–Roblox Code–

Crash 2:
.dmp file: http://spotcos.com/fmod_crash/crash2/crash2.dmp
Stacktrace:
fmod!FMOD::DSPConnection::setUserData+0x2153
fmod!FMOD::AsyncThread::wakeupThread+0x1c5d
fmod!FMOD::AsyncThread::wakeupThread+0x7fd6
fmod!FMOD::Channel::setPosition+0x54
–Roblox Code–

Crash 3:
.dmp file: http://spotcos.com/fmod_crash/crash3/crash3.dmp
fmod!FMOD::AsyncThread::wakeupThread+0x1e06
fmod!FMOD::AsyncThread::wakeupThread+0x30c9
fmod!FMOD::AsyncThread::wakeupThread+0x83ae
fmod!FMOD::AsyncThread::wakeupThread+0x8ddb
fmod!FMOD::ChannelControl::stop+0x3944
fmod!FMOD::AsyncThread::wakeupThread+0x93b0
fmod!FMOD::SystemI::createMemoryFile+0x40f5
fmod!FMOD::SystemI::createMemoryFile+0x481f
fmod!FMOD::SoundI::setUserDataInternal+0x5bd7
–Roblox Code–

Crash 4:
.dmp file: http://spotcos.com/fmod_crash/crash4/crash4.dmp
fmod!FMOD::ChannelControl::stop+0x207d
fmod!FMOD::AsyncThread::wakeupThread+0x93b0
fmod!FMOD::AsyncThread::wakeupThread+0x8d23
fmod!FMOD::ChannelControl::setPaused+0x4d
–Roblox Code–

Could you try running with the logging version of FMOD? Do you see any ERR or ASSERT messages in the output?

Thanks to your provided crash dumps we were able to isolate the cause of this bug and it has been fixed for our next release.