Duy Nguyen <pclo...@gmail.com> writes: >> This shouldn't be needed either. My assumption was that if someone >> explicitly asked for multiple threads we're better off failing than >> silently ignoring their request. The default/automatic setting will >> detect the number of cpus and behave correctly. > > No. I can have ~/.gitconfig shared between different machines. One > with multithread support and one without. I should not have to fall > back to conditional includes in order to use the same config file on > the machine without multithread.
Our attitude has been to fail unsatisfyable requests from the command line loudly to let the users' know, and tolerate having elements from the future versions of Git in the configuration file by ignoring them or tweaking them (silently or with a warning). So this hunk is in line with the current practice, and we should keep it this way. It is a separate discussion if we want to rethink that whole attitude, that discussion may be worth having, and I think that discussion is wider than this single topic.
