On Wed, Apr 22, 2020 at 03:36:09PM +0000, Karstens, Nate wrote: > There was some skepticism about whether our practice of > closing/reopening sockets was advisable. Regardless, it does expose what > I believe to be something that was overlooked in the forking process > model. We posted two solutions to the Austin Group defect tracker: I don't think it was "overlooked" at all. It's not safe to call system() from a threaded app. That's all. It's right there in the DESCRIPTION: The system() function need not be thread-safe. https://pubs.opengroup.org/onlinepubs/9699919799/functions/system.html > Ultimately the Austin Group felt that close-on-fork > was the preferred approach. I think it's also worth > pointing that out Solaris reportedly has this feature > (https://www.mail-archive.com/austin-group-l@xxxxxxxxxxxxx/msg05359.html). I am perplexed that the Austin Group thought this was a good idea.