On Wed, Feb 21, 2024 at 10:01:30PM +0100, e9hack wrote: > Am 21.02.2024 um 01:21 schrieb Daniel Golle: > > > > Yep, I didn't think about empty variables when I built this... > > > > Can you test this please: > > https://github.com/openwrt/procd/pull/3 > > > > This does fix the issue. > > If I test a modified executable, I replaced it by a link to a link on a > automatically mounted usb stick. The link on the usb stick points to the > modified executable. e.g.: > > /sbin/dnsmasq is renamed to /sbin/dnsmasq.old > /sbin/dnsmasq -> /mnt/test/dnsmasq/dnsmasq > /mnt/test/dnsmasq/dnsmasq -> /mnt/test/dnsmasq/dnsmasq-test-something > > If the modification bricks the router, I replace the last link on the usb > stick by > > /mnt/test/dnsmasq/dnsmasq -> /sbin/dnsmasq.old > > or > > /mnt/test/dnsmasq/dnsmasq -> /rom/sbin/dnsmasq > > and the router is operational again. > > This doesn't work for procd since procd is needed before the usb stick is > mounted. > > How can I do similar things for procd?
There isn't any as convenient option, for the reason you named. When hacking on procd the qemu targets (malta, armvirt, x86) are very handy for testing and debugging. Once things work well there I move on to a "friendly" board (ie. with dualboot/recovery feature, serial console, pstore which is also good for PID 1 crashes, ...). > > Regards, > Hartmut > > > _______________________________________________ > openwrt-devel mailing list > openwrt-devel@lists.openwrt.org > https://lists.openwrt.org/mailman/listinfo/openwrt-devel _______________________________________________ openwrt-devel mailing list openwrt-devel@lists.openwrt.org https://lists.openwrt.org/mailman/listinfo/openwrt-devel