Thanks for your help!
At that point I have been calling
make for a number of years without problem (until the support for Pkg3 was added). I asked the question in the hope someone would recognize why this might have happened. Debugging Travis is unfortunately beyond my skill set.
run(pipeline(...)) statements I can see make executes the first part (generating a C++ program) correctly but fails during the compilation step (no binary is ever added). I’m sure occasionally I’ll take a stab at this problem again, but for now I’ll just disable this part of the Travis testing. My original assumption (write permissions had somehow changed) appears incorrect.
Again, thanks for your help.