On Wed, 2 Jul 2025 15:43:08 -0400
Mathieu Desnoyers [off-list ref] wrote:
From my perspective, making trace analysis results reliable is the most
basic guarantee tooling should provide in order to make it trusted by
users. So I am tempted to err towards robustness rather than take
shortcuts because "it does not happen often".
Another solution which I'm thinking of doing for perf is to simply
state: a deferred stack trace does not go to any event before events
were dropped.
That is, even if the stack trace is associated, if events are dropped
before getting out of the kernel, just say, "Sorry, the events before
the dropped events lost its user stack trace".
That may better, as having no data is better than incorrect data.
-- Steve