Python Runner starts tests directly (not waiting for GUI)

[AmutioAlex]

I have been using cocotb for a while and decided to switch to the Python runner approach instead of using a Makefile.

The problem that I found is that, when I use Xcelium as a simulator and open the GUI, the test starts right away and does not wait to be started from the GUI. This is the current code that I run directly, not using pytest:

runner = get_runner("xcelium")
runner.build(verilog_sources    = [...],
             hdl_toplevel       = ...,
             always             = True)
runner.test(hdl_toplevel_lang   = "verilog",
            hdl_toplevel        = ...,
            test_module         = ...,
            gui                 = True)

Note that the GUI is opened, thus the gui flag works correctly. The problem is that before the GUI loads, the test has already started. It is not being triggered by the user.

I took a look at the source code of runner.py for the version that I am using (cocotb 1.8.1) and I am guessing that the main culprit here are lines 1002 and 1003, which always include the following options to the xrun command: -input @run and -input @exit.

I first thought that this was being fixed already for the next release, but the problem persists on the runner.py file of the master branch. Lines 1352 to 1360 state that in every case the test is run right away.

I think it would be wise for the runner to behave like the Makefile in this sense: if the GUI is started then the user has to trigger it.

Further information of my setup in case it is needed:

• cocotb 1.8.1
• Red Hat Enterprise 7.9 64-bit
• Xcelium 21.09.s003
• Python 3.10.1

[ktbarrett]

Sounds appropriate to me, feel free to provide a PR.