ADDRESS instruction without operands doesn't swap environments.

[RossPatterson]

TRL2 says that an ADDRESS instruction with no operands switches to the previously active command environment:

If ADDRESS is specified without either an environment name or an expression, then commands will be routed back to the environment that was selected before the previous lasting change of environment was made (and the current environment name is saved). Repeated execution of just "ADDRESS" will therefore switch the command destination between two alternative environments.

(Sec. 7 pp. 41)

The ANSI standard agrees:

At any given time, there will be two environments, the active environment and the alternate environment.

AddrInstr:
/* If ADDRESS keyword alone, environments are swapped. */
if \#Contains(address,taken_constant),
 & \#Contains(address,valueexp),
 & \#Contains(address,'WITH') then do
   call EnvAssign TRANSIENT, #Level, ACTIVE, #Level
   call EnvAssign ACTIVE, #Level, ALTERNATE, #Level
   call EnvAssign ALTERNATE, #Level, TRANSIENT, #Level
   return

(Sec 8.3.1, pp. 74)

bREXX doesn't do this. Instead, a bare ADDRESS always sets the environment to the "system string", which in CMS370 is "CMS":

static void
C_address( void )
{
	if (symbol == semicolon_sy) {
		_CodeAddByte(OP_PUSH);
			_CodeAddPtr(systemStr);
			TraceByte( other_middle );
		_CodeAddByte(OP_STOREOPT);
			_CodeAddByte(environment_opt);
		return;
	}

(compile.c)

static void
I_StoreOption( const PLstr value, const int opt )
{
	long	l;

	switch (opt) {
		case environment_opt:
			if (LLEN(*value) > 250)
				Lerror(ERR_ENVIRON_TOO_LONG,1,value);
			if (_proc[_rx_proc].env == _proc[_rx_proc-1].env)
				LPMALLOC(_proc[_rx_proc].env);
			Lstrcpy(_proc[_rx_proc].env,value);
			break;

(interpre.c)

We should change bREXX to do what TRL2 and the standard say.

[RossPatterson]

I have fixed this bug in the CMS370 version of bREXX. I'm willing to tweak the fix as necessary to make it fit this version, if one of the contributors wants to merge it in.

@vlachoudis @rvjansen @0branch @mgrossmann @Juergen-Git