On 12/04/2017 05:18 PM, enh wrote: > not sure what's meant to be going on in this file, but it looks wrong > in several ways.
Sigh, that shouldn't have been checked in. (I need tests/pending.) The theory here is that I use a known qemu testing environment (ala mkroot) to run tests in an environment that supplies specific things (like the "dummy" ethernet driver), which doesn't are about damage from failed tests that don't clean up after themselves, and isn't perturbed by whatever else the system might be doing. (On an emulated system I can reach out and grap /dev/loop1 by name. On a real system, it might be in use.) This is actually a surprisingly pervasive problem testing-wise, because if I want to test that rm -rf will go 2xPATH_MAX subdirectories deep, I don't want to worry about failures due to inode exhaustion or leaving something that the host's rm -rf might not be able to clear out either. Do it in an initmpfs, the qemu instance exits and it goes byebye. But yeah, the current one is pointless. Lemme see what I can do on my flight back to Austin... Rob _______________________________________________ Toybox mailing list [email protected] http://lists.landley.net/listinfo.cgi/toybox-landley.net
