I’ve watched the Unite Copenhagen talk in 2019 regarding debugging crashes that have happened in Burst - this was very helpful. But I’m curious how we could better debug a crash.dmp that had happened when a specific ISystem that used an end-frame entity command buffer to record a command.
Specifically a buffer doesn’t exist on an entity but SystemA uses an EntityCommandBuffer of the end-frame to append to buffer. The crash happens due to the buffer not existing. However when I get the accurate callstack it just points to the end-frame command buffer system, it doesn’t give me the specific system that had recorded the command, it points to the EcbWalker calls (playback processor). Unless there’s more context in the crash frame / call stack - maybe a guid that points to the specific system that had scheduled the ecb?
The solution to the crash here is to check if the buffer exists before recording the command, but I want to establish a workflow for any future crashes and better root-cause the issue - especially with a game that is released (and we have burst debug symbols)
So like extracting a journaling type of guid from the crash.dmp? Is this possible?
