On Mon, 30.06.14 13:27, Colin Walters (walt...@verbum.org) wrote: > > On Mon, Jun 30, 2014, at 12:49 PM, Lennart Poettering wrote: > > > This sounds confused. Why would the cloud images not use a tmpfs for > > /tmp? > > https://git.fedorahosted.org/cgit/spin-kickstarts.git/tree/fedora-cloud-base.ks#n181
Oh, man. This is just stupid. > > Anyway, to allow to avoid the discussion I have now added > > ConditionPathIsSymbolicLink=!/tmp to tmp.mount. This will skip mounting > > of the tmpfs if /tmp is already a symlink. This should fix your issue? > > Hmm, that would effectively force tmp to *not* be a tmpfs with the > current tree. > > We just pushed a different fix for continuous to just make /tmp a > regular directory, since there I'm comfortable assuming /tmp-on-tmpfs. Not following? > This makes physical /tmp be more broken for ostree systems, but maybe I > can just ship some logic to bind mount it to /sysroot/tmp if it's not > already a mount point, or so. I'll figure something out. Hmm, I already commited the fix I mentioned. Shall I revert it? It kinda makes sense to me to leave the condition in, after all with a symlink you revert redirect things to something else, and systemd cannot make sense of that. Hence: if you redirect things then drop in a .mount file for the new location, and don't do any mounts on /tmp itself. Lennart -- Lennart Poettering, Red Hat _______________________________________________ systemd-devel mailing list systemd-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/systemd-devel
