On 2019/04/22 14:56, Marc Espie wrote: > Here's a log traceback of a failure to build py-unicorn. > > As far as I can tell, the python tools do run whatever build setup is > around at that point, and thus fail abysmally if, say, you pkg_delete > stuff between the scan and the run. > > e.g., it found pbr_pbr.json, and then it fails running it, because it's > gone, as it's not an actual dependency of py-unicorn. > > what would be required would be some kind of exception handling to fail > gracefully in that case... e.g., near the end of the log around > ep.resolve() > > > (and of course it's not even a case of asking junk to wait until AFTER > configure, as in this specific case, running setup actually happens very > late in the name of the port... as usual in python, everything is thoroughly > entangled together).
Interesting as I don't recall ever running into this (2 builds/week approx), perhaps I have less churn though (2+2+4 cores).. Have you seen this before or is it new? I wonder if something changed in setuptools.