gnetlist ignores "pure" nets between io blocks on a sheet

[wzab]

In the attached simplified design I use two sheets to define connectors, and the another one to describe interconnections.
Interconnections are defined using the passive-1.sym blocks.
When the sheet describing the interconnections contains a "pure" connection between two passive blocks (see interconnections.sch and test1.net) the connection is fully ignored.
If I add any "real" element to that connection (see interconnections2.sch, where I've added the testpoint, and the resulting test2.net) - gnetlist generates the correct netlist.

PS. This bug is an extension to the 2148310 bug (I was not able to add the new attachment to the previous bug report, and in fact this bug seems to be gnetlist related, not gschem related).

[wzab (wzab-users)]

I've just stated, that the problem appears, when there is no element with "refdes" attribute connected to the net on the particular sheet.

[nobody (nobody-users)]

I've found a "quick & dirty" workaround. I attach to such "pure" nets the testpoints with refdes set to RMV? (? is then replaced with a number).
It assures proper generation of the netlist.
Then I filter the network with the command:

$cat generated.net | sed -e s/RMV[0-9]*-1// > filtered.net

It removes the testpoints from the netlist.
It is not very elegant solution (as the schematics contains testpoints not existing on PCB) but at least it allows to reasonably handle pure net-to-net interconnections.