Problems with set_type_override_by_type for transactions

[cliffc]

Hi, All -

I have a small example (test1, env, agent, driver, sequencer, sequence and sequence_item) coded with trans extended from uvm_sequence_item. All items are registered with the factory. The driver and sequencer are parameterized to trans, as is the sequence. The trans has a convert2string virtual function so I can display all of the transactions executed by the sequence.

I coded a trans2 transaction extended from trans, with convert2string updates to show "trans2", and a separate test with a build_phase() that does:

set_type_override_by_type(trans::get_type(), trans2::get_type());

But when I run the second test, I still see the trans1 messages.

I also printed out the factory information in the test's end_of_elaboration_phase() and it shows that trans2 should replace trans.

I then created a second driver, extended from the original driver, added an additional `uvm_info message to the run_phase() and a third test that does:

set_type_override_by_type(tb_driver::get_type(), tb_driver2::get_type());

And when I run, the second driver messages appear, but I still do not see the trans2 messages.

Again, printing the factory.print() shows that both the tb_driver and trans types should be replaced by tb_driver2 and trans2.

Any hints as to why the transaction and transaction printing of the trans2 transaction do not seem to be working?

Thanks - Cliff Cummings

[cliffc]

Well, I'm embarrassed!

I was putting together an example that showed the difference between ::type_id::create() factory construction and new() class construction, and I forgot to remove the new() constructor from the sequence. Once all of the new() constructors were properly replaced with factory create()-constructors, the example worked fine and the trans2 print message is now executing as expected.

Sorry for the false alarm.

Regards - Cliff Cummings