On 11/29/2017 08:50 PM, enh wrote: > let me know if you'd rather have the #ifdef in portability.h --- i'm > not sure whether your aim is to have all #ifdefery in there, or > whether something should be needed in two places first. > > (the other alternative would be to always try both locations, but that > seemed even worse.)
Trying both sounds right to me. I could see Red Hat Enterprise or similar doing this too. Alas there isn't a real /dev spec since the Linux Foundation absorbed and destroyed the Linux Standards Base. There are entries in devtmpfs's block subdir but they're all symlinks ala 7:0 -> ../loop0 Still, if you're using devtmpfs you get the top level entry, and if it's not there looking in the subdir as a fallback works for android and maybe elsewhere. Sigh, looking at my losetup I have pending changes to it, of course. Lemme check those in too... Ah right, I was adding losetup -A to set autoclear, but that can't work because you have to open the loopback device to call ioctls on it,and when you close the device's filehandle the association is cleared. So mount has to do it internally and then call mount() before closing the filehandle so it's pinned by the mount reference. And moving loopback_setup() into lib/ requires stripping the GLOBALS references out and turning them into function arguments... Right, break that part _out_ and check in the rest. (git --patch is lovely, athough it means you have to pull it to the "clean" directory and retest after committing.) Rob _______________________________________________ Toybox mailing list [email protected] http://lists.landley.net/listinfo.cgi/toybox-landley.net
