On Tue, Oct 13, 2015 at 12:16:17PM +0000, Mattia Rizzolo wrote: > On Tue, Oct 13, 2015 at 02:21:38PM +0300, Niko Tyni wrote: > > Apparently the test fails when it can't find any "regular" (as opposed > > to "special") filesystems. In this age of building on virtual hosts > > and chroots, I'm not sure this is a valid assumption, though it does > > look strange that the build on reproducible.d.n doesn't report anything > > mounted as the root file system. > > I've no clue which file systems are considered "regular" and which ones > are "special" for Sys::Filesystem,
Reading lib/Sys/Filesystem/Linux.pm, the full list of "special"file system types is binfmt_misc, debugfs, devpts, fusectl, fuse.gvfs-fuse-daemon, mini_fo, nfsd, proc, procbususb, securityfs, swap, sysfs, tmpfs, udev > but what I'm pretty confident about > is that a pbuilder's /proc/mounts usually contains no entry for the root > file system. > So I assume the absence of an entry for / is not the cause of the ftbfs, > since that's also what happen on "build 1" (and on the test build I just > ran here locally). There's lots of diagnostic output in the build1 log including # filesystem - mounted - special - device - options - format - volume - label - type # / - 1 - 0 - /dev/vda1 - rw,relatime,errors=remount-ro,data=ordered - ext4 - 1 - 1 - ext4 and a long list of other file systems like things under /srv/workspace/pbuilder/61616. It looks like these come from either /etc/fstab or /proc/mounts. So perhaps the difference between the chroots is in /etc/fstab ? Still not sure how much of this can be expected from a 'sane' build environment... -- Niko Tyni nt...@debian.org