Philippe Troin writes: > The checking "if tcsetpgrp() actually works" configure test stops the > configure script if it is ran in the background. The enclose patch > prevents this from happening by ignoring SIGTTOU for the duration of > this test. This is a second spin on the same idea, following up on comments from Jens Petersen in . Extra stuff taken care of: - works if stdin is redirected (opens /dev/tty instead of relying on stdin) - detects if ran without controlling tty, and aborts Stuff taken care of with the first patch, included with this patch: - does not stop if ran on the background. I think this patch will address all of Jens's concerns. There are still some quirks, but I believe they will only show up very infrequently: - configure completely fails if ran without a controlling tty. Should not be a major problem, processes running the background still have a controlling tty. Running without a ctty only happens when running from cron or some other batch facilities. Anybody doing that? - there still is a tiny race window if the shell decides to switch the foreground process between the tcgetpgrp() and tcsetpgrp() calls. It is probably better to go this route rather than running this particular configure test on its own pty: this would add a lot of very platform specific cruft to the test itself, and configure is supposed to be robust. Phil.