I'm using rev322062, compiled (cc -o t t.c) the t.c below, which is a quite small test case just for this bug.

The lifecycle of t and DTrace are not equal - thus I try to communicate from the user application (t) to DTrace when to start instrumenting and when to stop.  I do this by t writing START and STOP onto stdout, and the D script to set a thread-local variable.

The probe sysctl::write:entry in t.d is not executed - the same one in t2.d, where a printf is added - is executed.  My expectation from DTace is that t.d's sysctl::write:entry would also be executed.

I execute "doas ./d.t" in one shell, then execute ./t in another shell.  (Running dtrace -c obviously works because DTrace then exits once t exits, but this is not a solution for my use case (a much bigger instrumentation where I only want to trace after some initialisation).

t.c
==============================
#include <stdio.h>
#include <unistd.h>

int main (int argc, char ** argv) {
  printf("START");
  printf("BOOOOOOOO\n");
  printf("STOP");
  printf("done now\n");
}
==============================

t.d
==============================
#!/usr/sbin/dtrace -Cs

#pragma D option quiet

#define prog "t"
#define act execname == prog && self->started == 1

syscall::write:entry
/arg0 == 1 && (string)copyin(arg1, 5) == "START"/
{
  self->started = 1;
}

syscall::write:entry
/act && arg0 == 1 && (string)copyin(arg1, 4) == "STOP" && self->started == 1/
{
  exit(0);
}
==============================

t2.d
==============================
#!/usr/sbin/dtrace -Cs

#pragma D option quiet

#define prog "t"
#define act execname == prog && self->started == 1

syscall::write:entry
/arg0 == 1 && (string)copyin(arg1, 5) == "START"/
{
  printf("starting\n");
  self->started = 1;
}

syscall::write:entry
/act && arg0 == 1 && (string)copyin(arg1, 4) == "STOP" && self->started == 1/
{
  printf("exiting now\n");
  exit(0);
}
==============================