I submitted bug bug 8254. I hope the stack trace provides some help...
I will continue load tests with Boehm GC from now on. Best case would obviously
be that you can spot some user error on my end =)
-- roope
On Nov 5, 2012, at 8:44 PM, Miguel de Icaza mig...@xamarin.com wrote:
So in we
Hi!
I am working on a MMO game thats written in C#. We are currently running load
tests on mono 3.0.0 in linux environment.
I have noticed a few times that when we cause a spike the amount of threads we
run the server sometimes crashes with the following output:
Native stacktrace:
Is there a stress testing mode for SGEN? e.g. to force a GC at every
allocation? That can really help to nail GC bugs fast (at huge cost of
course)
On Mon, Nov 5, 2012 at 2:56 PM, Rodrigo Kumpera kump...@gmail.com wrote:
First, make sure you have symbols and gdb installed. Then, please, file a
On Nov 5, 2012, at 7:16 AM, Roope Kangas ro...@grandcrugames.com wrote:
I have noticed a few times that when we cause a spike the amount of threads
we run the server sometimes crashes with the following output:
...
Any ideas on what to do? This kind of behaviour does not occur with Boehm GC.
Yes, look in the mono man page for MONO_GC_DEBUG. We have multiple options
that can help triaging bugs. Collecting on every allocation
sometimes does help, but it depends very much on the nature of the crash.
And from the looks of the report, it's not possible to say if it's a GC
or a runtime bug
Hi!
We noticed a bug in our codebase that might contribute to this. If I understand
correctly the program was going through a lot of data (iterating through 3D
voxel geometry comparing data) and it had a bug in equals that caused boxing.
So in we were generating _a lot_ of small classes,
So in we were generating _a lot_ of small classes, loosing references to
those new instances and doing that on very many threads at once.
Sounds like a GC stress tester =)
Anyhow, I will try to repro the bug with more debug and file it to
bugzilla.
This would be fantastic, thank you!