Also, I noticed (gcc didn't ;-) that the mknod parameters have been inadvertently transposed. To be safe we should also specify the permissions for the device nodes we're creating (and possibly explicitly set umask).
Finally, I think we need a slightly more holistic view on Upstart-in-initramfs-less-environments since we cannot assume any devices exist, including /dev/null, /dev/console, /dev/kmsg, /dev/tty. On Ubuntu, I don't think we'd see an issue since when a system is installed from the live media, some devices already exist. However, we cannot assume that for all systems.
Hi Steve,
Good catch! The patch seems to have a stray 'needs_devtmpfs' at line 304:
301 if (stat ("/dev/ptmx", &statbuf) < 0 || !S_ISCHR( statbuf. st_mode) st_dev) != 5 || minor(statbuf. st_dev) != 2) statbuf. st_mode) )
302 || major(statbuf.
303 needs_devtmpfs = 1;
304 if (needs_devtmpfs
305 || stat("/dev/pts", &statbuf) < 0 || !S_ISDIR(
306 needs_devtmpfs = 1;
Also, I noticed (gcc didn't ;-) that the mknod parameters have been inadvertently transposed. To be safe we should also specify the permissions for the device nodes we're creating (and possibly explicitly set umask).
Finally, I think we need a slightly more holistic view on Upstart- in-initramfs- less-environmen ts since we cannot assume any devices exist, including /dev/null, /dev/console, /dev/kmsg, /dev/tty. On Ubuntu, I don't think we'd see an issue since when a system is installed from the live media, some devices already exist. However, we cannot assume that for all systems.