The increase of plugin crashes with e10s doesn't show any glaring signatures. There is a significant increase in these two generic/undiagnosed signatures: F1398665248_____________________________ mozilla::plugins::PluginModuleChild::NPN_CreateObject We know that in the past the F139 signature has been related to nesting RPC calls from multiple plugin instances, and I have a similar hunch in bug 1252152. In particular I'm worried about us tearing down plugin instances while that instance or another instance is on the stack. In addition to the patch I'm going to attach in bug 1252152 which will collect field crashes to detect this case, I think we should do a more thorough audit of the possibilities there. There are NPP destruction guards in the plugin process which are supposed to help prevent this, but I suspect they may not be comprehensive. With e10s it's possible that this might be even worse if the chrome process decides to shut down the plugin completely while a content process is still using it. I think this possibility is less likely but still worth double-checking.
I may be able to do this next week, but I'd love help if it's available.
Assignee: benjamin → nobody
Diagnostic crashes don't show anything new or spectacular, and after a quick code review I think this is ok.
Status: NEW → RESOLVED
Last Resolved: a year ago
Resolution: --- → INCOMPLETE
You need to log in before you can comment on or make changes to this bug.