lp wrote:I didn't allocate enough space for the two shorts in that structure
That's no problem, as long as the corresponding bits in the mode parameter are not set, those members aren't accessed. They aren't even documented yet in tos.hyp, so i guess that was an extension added later.
The interpreter does set the path. I will have a look at it.
I got the traces by adding printfs in the XaAES kernel (just before create_process is called there in shel_write), and in sys_d_setpath0 in the mint kernel. The first message comes from XaAES, the 2nd from create_process calling sys_d_setpath0 because the defdir element was set, and the last one from the interpreter.
You can also try with litchi.gfa, it complains that the resource file cannot be found.