tests/kill4: Fix tests with wrong assumption

Previously, some tests in kill4-p.tst depended on traps reacting to a
signal sent by the shell. Since the traps were set in subshells, the
processes executing the trap built-in might not be the direct child
processes of the shell. In that case, the children were terminated by
the signal, possibly before the subshells responded to the signal,
causing the subsequent wait built-in to return too early in the main
shell process.

To resolve this issue, the test cases are rewritten without using traps.

tests/kill4-p.tst

@@ -42,32 +42,34 @@ setup 'halt() while kill -s CONT $$; do sleep 1; done'
 mkfifo fifo1 fifo2 fifo3
 
 test_oE 'sending signal to background job' -m
-(trap 'echo 1;      exit' USR1; >fifo1; halt) |
-(trap 'echo 2; cat; exit' USR1; >fifo2; halt) |
-(trap 'echo 3; cat; exit' USR1; >fifo3; halt) &
+# The subshells stop at the redirections, waiting for the unopened FIFOs.
+(>fifo1; echo not reached 1 >&2) |
+(>fifo2; echo not reached 2 >&2) |
+(>fifo3; echo not reached 3 >&2) &
 halt &
-<fifo1 <fifo2 <fifo3
-kill -s USR1 %'(trap'
-wait %'(trap'
+kill -s USR1 '%?echo'
+wait '%?echo'
+kill -l $?
 kill -s USR2 %halt
 wait %halt
 kill -l $?
 __IN__
-3
-2
-1
+USR1
 USR2
 __OUT__
 
 test_oE 'sending to multiple processes' -m
-(trap 'echo; exit' TERM; >fifo1; halt X) &
-(trap 'echo; exit' TERM; >fifo2; halt Y) &
-<fifo1 <fifo2
-kill '%?X' $!
-wait '%?X' $!
+# The subshells stop at the redirections, waiting for the unopened FIFOs.
+(>fifo1; echo not reached 1 >&2) &
+(>fifo2; echo not reached 2 >&2) &
+kill '%?fifo1' '%?fifo2'
+wait '%?fifo1'
+kill -l $?
+wait '%?fifo2'
+kill -l $?
 __IN__
-
-
+TERM
+TERM
 __OUT__
 
 )