Changelog
autopkgtest (4.2.2~ubuntu16.10.1) yakkety-backports; urgency=medium
* No-change backport to yakkety
autopkgtest (4.2.2) unstable; urgency=medium
* tests/adt-run: Adjust expected cpu_model in
NullRunner.test_tree_output_dir()
* autopkgtest-buildvm-ubuntu-cloud: Run setup script with -x in verbose
mode. This provides more useful information when updating/adjusting
setup-testbed.
* setup-commands/setup-testbed: Run all apt operations with
DEBIAN_FRONTEND=noninteractive.
autopkgtest (4.2.1) unstable; urgency=medium
* Fix trailing '\' in test command with ssh runner
* ssh: Filter out trailing spaces from commands in auxverb
* Apply patches for tests with unbuilt-tree (Closes: #844454)
* setup-commands/setup-testbed: Install dpkg-dev
* Fix stray ')' in --apt-pocket handling
* Adjust expected cpu_model in NullRunner.test_tree_output_dir()
* PEP-8 fixes
* Robustify tests against kernel versions with trailing space
autopkgtest (4.2) unstable; urgency=medium
[ Martin Pitt ]
* ssh: Fix result of tests that break the testbed. (LP: #1630578)
* qemu: Fix user/password login mode without a ttyS1 root shell.
- Concatenate bytes with each other, not a str to a byte.
- Sync/flush console after sending password and ttyS1 shell command, wait
until login is complete before continuing.
- Send password to sudo for ttyS1 shell command, to also work with
password-requiring sudo. (LP: #1630963)
* qemu: Fix reboot in user/password login mode.
* ssh: Add --capability option.
This is useful to run tests which require an isolation level or have
breaks-testbed without a setup script.
* autopkgtest-build-lxd: Ask "lxc profile" for default bridge instead of
/etc/default/lxd-bridge. The latter went away in recent LXD versions, and
"lxc profile show default" also works in earlier versions.
* Make --apt-upgrade consider a "404" error as test failure.
Previously all errors from "apt-get update" were considered temporary
failures, i. e. the setup command always exited with 1. But if we specify
e. g. a nonexisting release from a distro or a PPA, this won't just go
away by itself -- we want the test to actually fail instead of getting
stuck in an eternal "retry on temporary failures" loop.
* schroot: Don't fail on stderr of schroot as long as it succeeds.
(LP: #1637898)
* qemu: Hide detected udev file system properties on /dev/baseimage.
(Closes: #842299)
[ Simon McVittie ]
* VirtSubproc: open arbitrary files in binary mode.
* VirtSubproc: if check_exec status is nonzero, include stderr in message.
* qemu: put the shared directory in /run. If the virtual machine's root
filesystem is read-only, we won't be able to create /autopkgtest.
* qemu: Move eofcat into /tmp. This avoids having to write it to /bin, which
might be read-only. Re-create eofcat on every boot for this.
* source_rules_command: log the result we got if it is not as expected
(Closes: #842302)
* Add 'needs-reboot' restriction. This allows tests to be explicit about
needing to reboot the machine, rather than assuming that
'isolation-machine' is enough. It is plumbed into the existing 'reboot'
capability, which is distinct from 'isolation-machine'.
(Closes: #842300)
* Add --setup-commands-boot for commands that must be run at every boot.
This can be used for doing "mount -o remount,rw /" before any dpkg
operation, or transient setup like writing files into /run/.
(Closes: #842091)
-- Martin Pitt <email address hidden> Mon, 05 Dec 2016 17:48:08 +0100