No information about "no wait" state

[egor-rogov]

The case of not waiting (wait_event_info == 0) is no less important than the case of waiting. Absence of "no wait" counts in profile make it impossible to compare waiting time to active time.

[egor-rogov]

Looks like this requires another algorithm of iterating over backends, probably the same as in pg_stat_activity. Just commenting out if (proc->wait_event_info == 0) continue; in probe_waits() leaves a lot of empty rows in profile from some processes which are not listed in pg_stat_activity.