Some issues about tpacall

Question

1. In server side, I found that the return value of tpacall is different every time, that is the return value add one at a time, but their addresses are the same. Could someone explain that?
2. The last parameter of tpacall() and tpcall() is flag, whose explanation is:
Flag options. You can list a group of flags by using the logical OR operator.
But I found that the common use is:
tpacall("PRINTER", rbuf, rrlen, TPNOTRAN|TPNOREPLY);
"|" is bit operation rather logical operation ?
3. As for the tpacall, in my mind, the two services are already separated and are in two transactions, but why could it use TPNOTRAN|TPNOREPLY?
Thx a lot.

Answer 1

1. Except when the TPNOREPLY flag is used, the return value from tpacall() is a handle that can later be passed to tpgetrply(). When the TPNOREPLY flag is used, tpacall() returns 0 on success. In either case, the only error return value from tpacall() is -1.
2. You're correct that TPNOTRAN|TPNOREPLY is a bitwise OR, and this is what is intended.
3. The TPNOTRAN flag indicates that the called service is not part of any transaction that the caller may already be a part of. A service called with TPNOTRAN can fail without causing the caller's transaction (if any) to be marked as rollback-only.
If tpacall() or tpcall() is called from within a transaction without using the TPNOTRAN flag, then both the callier and the called service are part of the same global transaction and either both will succeed or both will be rolled back.
The TPNOREPLY flag indicates that no reply to the service is expected. In order to use this flag, you must specify the TPNOTRAN flag or you must make the call outside of a transaction.
A "PRINTER" service is a good candidate to use the TPNOTRAN|TPNOREPLY flags, since the printer may not support transactions or have a way to reply that a document could or could not be printed.
Regards,
Ed