I'm trying to profile node 0.10.33, the case requires profiler to be activated in code. so --prof is not suits.
v8tools based profilers returns profile logs in memory using json tree, which is not suits for any visualization tool.
Main problem using node-profiler is logs received by it is not readable by any tool: nprof, chrome://tracing/ etc. For example nprof gives lots of "Code move event for unknown code" errors and
[Unknown]:
ticks total nonlib name
146518 98.8%
as result.
The same time using just --prof gives valid logs.
As i see it node-profiler just calls to v8 built-in profiler resuming it, for thus we calls node with "--prof", "--prof_lazy", "--log" parameters. So logs must be identical.
Where im wrong?