On Thu, Jun 29, 2017 at 03:24:46AM +0000, Tomasz Chmielewski wrote: > With lxd 2.14: > > # mkdir /tmp/testdir > # touch /tmp/testdir/file1 /tmp/testdir/file2 > # lxc file push -r /tmp/testdir/ testvm1/tmp # <---- note the trailing > slash after /tmp/testdir/ > # echo $? > 0 > # lxc exec testvm1 ls /tmp > file1 file2 > > > With lxd 2.15: > > # mkdir /tmp/testdir > # touch /tmp/testdir/file1 /tmp/testdir/file2 > # lxc file push -r /tmp/testdir/ testvm2/tmp # <---- note the trailing > slash after /tmp/testdir/ > # lxc exec testvm2 ls /tmp > testdir > # lxc exec testvm2 ls /tmp/testdir > file1 file2 > > > This breaks many scripts!
Hmm, so that just proves that we really more systematic testing of the way we handle all that mess in "lxc file push". For this particular case, my feeling is that LXD 2.5's behavior is correct. It's the same behavior as you get from "cp -R" and from "scp -r" which are the templates we've been using for that command. But the fact that this behavior changed and didn't cause any test regression is concerning and something that should be addressed. Christian mentioned looking at the testsuite for "cp" and having our own do pretty much the same tests for things like path handling, symlink dereferencing, ... -- Stéphane Graber Ubuntu developer http://www.ubuntu.com
signature.asc
Description: PGP signature
_______________________________________________ lxc-users mailing list [email protected] http://lists.linuxcontainers.org/listinfo/lxc-users
