It would be useful if there were a way to distinguish between a postinst failing because of errors or a postinst deciding to delay its execution until first boot in the install log. At the moment all non-successful postinst are assumed to be delayed on purpose. I suggest exit 75 for "I wish to be executed on the host please" (#define EX_TEMPFAIL 75 is a semi-standard in sysexits.h) and anything else being reported differently. The goal being that OE's rootfs task can warn when postinsts are delayed without using exit 75 and we can notice if eg a postinst that should run on the build host is failing.
Shouldn't that be a convention on how postinst are authored and not part of opkg? pkg_postinst_PACKAGENAME() { if [ x"$D" = "x" ]; then # Actions to carry out on the device go here else exit 75 fi }
Alex's patches make this moot, closing.