-
debci (0.10.1) unstable; urgency=medium
* Fix documentation build
-- Antonio Terceiro <email address hidden> Tue, 30 Sep 2014 17:29:48 -0300
-
debci (0.10) unstable; urgency=medium
[ Martin Pitt ]
* Fix Makefile to be idempotent, previously it failed on already existing
symlinks.
* Add missing dctrl-tools build dependency, Makefile is calling grep-dctrl.
* Makefile: Don't call "checkdeps" on "make check", the test suite doesn't
actually need autopkgtest or the libjs packages (which are for the UI
only).
* test/test_helper.sh: Don't hardcode "amd64", to fix test failure when
building on other architectures.
[ Antonio Terceiro ]
* debci-generate-index: update packages.json atomically
* web UI improvements:
- remove limit of number of items to display in the search
(Closes: #762213)
- trigger search immediately when there is text in the search box
(this will happen if you reload the page)
- store search query in the URL, and honor that when opening any page that
contains a search box
-- Antonio Terceiro <email address hidden> Fri, 26 Sep 2014 18:22:06 -0300
-
debci (0.9) unstable; urgency=medium
[ Antonio Terceiro ]
* schroot/create-testbed: update existing chroots
* debci-generate-feeds: add description to feeds
* Incorporate results of the Google Summer of Code 2014 project by Brandon
Fairchild, providing uncountable improvements to the web interface.
http://lists.alioth.debian.org/pipermail/soc-coordination/2014-August/002295.html
[ Brandon Fairchild ]
* web UI was improved to support multiple suites and architectures, reduce
the dependency the UI had on Javascript, and provide more information to
package maintainers. The documentation was also worked on to provide
information to people interested in working on debci (specifically the web
UI)
-- Antonio Terceiro <email address hidden> Sat, 23 Aug 2014 18:10:04 -0300
-
debci (0.8.1) unstable; urgency=medium
* Documentation updates
* Fix obtaining the finish date of tests. Parsing the adt-run log started
producing bad dates when adt-run changed to displaying only the time
without the date. debci now uses the timestamp of when adt-run finished
instead.
* debci-batch: store offline mark in persistent storage to avoid going back
online after a reboot.
* debci-test: revert to updating only package-specific data; it turns out
that updating global data after every package is too expensive.
* debci-batch: update global data every 1 hour or so
* debci-generate-index: avoid crash when regenerating data for extenal
consumption from scratch. This process is not completely idempotent yet,
so it needs more work.
-- Antonio Terceiro <email address hidden> Mon, 18 Aug 2014 08:25:23 -0300
-
debci (0.8) unstable; urgency=medium
* debci-test: update global data files, not only package-specific data file,
after each successfull test run.
* debci-generate-index: ensure that no two instances can run at the same
time.
-- Antonio Terceiro <email address hidden> Thu, 31 Jul 2014 11:55:27 -0300
-
debci (0.7.3) unstable; urgency=medium
* debci-generate-feeds:
- always update the global feed.
- Point main link of feed entries to the debci log instead of the package
page.
* debci-generate-index:
- report duration of each test run together with the status
* debci-test:
- fix regression with new packages caused by not mkdir'ing the full path
to the adt-run output directory. `fake` backend adapted to emulate the
behaviour of adt-run so that the tests would now fail without the fix.
-- Antonio Terceiro <email address hidden> Wed, 18 Jun 2014 14:29:46 +0200
-
debci (0.7.2) unstable; urgency=medium
* Fix calculation of test elapsed time. We can't store the beginning
timestamp at the adt-run output directory because that will be wiped out
by adt-run before running the test.
* public/index.html: fix source code URL to point to the collab-main
repository which is the canonical one.
-- Antonio Terceiro <email address hidden> Tue, 10 Jun 2014 17:30:41 -0300
-
debci (0.7.1) unstable; urgency=medium
* Fix handling of blame with "unknown" as value. Will avoid crashing when
generating indexes
-- Antonio Terceiro <email address hidden> Mon, 09 Jun 2014 11:07:11 -0300
-
debci (0.7) unstable; urgency=medium
[ Antonio Terceiro ]
* debci-batch: put packages with fastest testsuites first in the queue
* debci-generate-feeds: allow generating feeds for specific packages
* debci-generate-index: update package feed right away when indexing
individual packages
* public/app.js: Adjust the "Pass percentage" Y axis labels so that chart is
actually a percentage chart.
* public/index.html: add link to global Atom feed in the home page.
* debci-test, debci-generate-index: more robust way of calculating test run
duration. Parsing the autopkgtest log is just not good enough.
[ Martin Pitt ]
* Fix report_status() to work for bash scripts.
* debci-test: Add --print-output option to write output directory to stdout.
With that the caller can easily check the result.
* debci, tools/server.sh: Exec the target program instead of running in a
subprocess, for efficiency and proper SIGTERM handling.
* Drop tools/convert-data. We only needed it for version 0.6, and we're
going to need a different conversion script after the next data/ dir
reorganization.
* Reorganize data directory to move the "categories" packages/,
autopkgtest/, status/, and feeds/ to the top level; they contain
subdirs for all releases and architectures. This makes autopkgtest/ fully
rsyncable and also makes it easier to the web UI to browse/present data
for multiple releases and architectures. Also split release-arch/ into
release/arch/.
* Add tools/convert-data-0.7: Convert data directory to changed format from
above.
* Fix debci-status -l if there is just one package.
* Add tests for debci-batch, covering skipping/re-running of tests under
appropriate conditions and showing the cause for re-run in the log.
* Replace test/json_validity (in ruby) with test/test_json.sh (shell with
just calling ruby for the actual JSON validation). With that we can re-use
test_helper's test bed setup instead of having to replicate it all in
ruby.
* Make test/runall.sh executable for convenience, and make it only run
test_* scripts.
* debci-batch: remove reading configuration from /etc/default/debci
* lib/environment.sh: read configuration from $debci_basedir/config/debci.conf
* Add debci_mirror config variable to configure an archive mirror. Use it
for chdist and schroot setup.
* Drop unnecessary faketime build dependency.
* Add LXC backend.
-- Antonio Terceiro <email address hidden> Tue, 03 Jun 2014 22:36:44 -0300
-
debci (0.6.3) unstable; urgency=medium
* debian-status: avoid generating invalid JSON in the presence of packages
without status files
-- Antonio Terceiro <email address hidden> Thu, 29 May 2014 12:48:48 -0300
-
debci (0.6.2) unstable; urgency=medium
* debci-generate-index: don't try merging dependencies diff when no package
has been blamed before.
-- Antonio Terceiro <email address hidden> Wed, 28 May 2014 18:14:06 -0300
-
debci (0.6.1) unstable; urgency=medium
* Fix chdist calls on stable (See #736504)
-- Antonio Terceiro <email address hidden> Thu, 15 May 2014 12:51:25 -0300
-
debci (0.6) unstable; urgency=medium
[ Martin Pitt ]
* Move schroot creation from bin/debci-setup to
backends/schroot/create-testbed, as it is schroot backend specific and
other backends will need different setup.
* process-package: Call backend test-package with second argument that
specifies the adt-run output directory. Implement that in schroot backend
with adt-run --output-dir, and fake backend when passing $DEBCI_FAKE_DEPS.
This will be used to collect and present log files, artifacts,
package/version lists, etc.
* Eliminate list-base-system and use autopkgtest's testbed-packages output
file instead. This will also work with remote workers and other virt
servers.
* Eliminate check_version() and take this information from autopkgtest's
output.
* Move report_status() into lib/functions.sh as we'll need it in more places
than just process-package.
* Move "run needed" policy from process-package to bin/debci. With this,
debci-test can be called manually at any time to retry tests or get called
by e. g. britney which already evaluates when to run tests. Remove the
"run test at most once a day" check completely as package updates can
happen several times a day. As a by-product this also fixes debci's -f
option to force a test run.
* backends/schroot/list-dependencies: Ignore individual apt-get failures as
that gets called for binaries which don't exist on the selected
architecture. Fixes operation for e. g. glib2.0.
* Use chdist instead of the local schroot for package queries:
- Add scripts/setup-chdist to create/update a chdist for the selected
release.
- Use that in list-dependencies instead of querying the schroot. Move that
script to scripts/ as it is now not backend specific any more.
- Update lib/functions.sh to use chdist instead of the schroot.
- Add devscripts dependency for chdist, and demote schroot dependency to
recommends.
This allows us to use other backends, like QEMU or remote ones.
* Reorganize data directory:
- Split out autopkgtest output into ${debci_data_dir}/autopkgtest/:
This now contains a full directory for each test run with all of
adt-run's --output-dir files. This provides access to individual test
stdout/err logs, the package version lists, and additional artifacts
that tests produce. With this, autopkgtest/ becomes eligible for being
on a remote file system and can be synced/rebuilt independently of the
metadata.
- Use run IDs with format YYYYMMDD_HHMMSS[machine or tags] instead of the
old YYYYMMDD-n as the former works for a distributed file system and
remote workers.
- base.txt is now replaced with adt-run's testbed-packages output file.
- test-packages.txt gets dropped as there is no need to keep it around
permanently. Build it from the per-test -packages logs from adt-run.
- Move status.json into status/ directory and remove the redundant
latest.json.
- Move packages.json into status/ directory.
* Add tools/convert-data: Convert data/ directory to changed format from
above. This can be dropped again after running it on the production debci
instance.
* app.js: Filter out query strings from URL for constructing the .json URLs.
Fixes wrong URLs after using the package search field.
* backends/schroot/test-package: Drop unused temporary dir.
* backends/*/test-package: Write exit code into "exitcode" file in output
dir, so that we retain it even when we switch to asynchronous/remote
processing.
* backends/fake/test-package: Write log file into output dir, like adt-run.
* Move test result evaluation, log and meta data creation from debci-test to
debci-generate-indexes. This prepares the way for running adt-run
asynchronously or remotely, and picking up the results from the
autopkgtest/ data directory.
* Reduce the per-test run <timestamp>.log files to the debci metadata
(triggers, package versions, package diffs, overall result), as the actual
autopkgtest log is now kept in data/*/autopkgtest/. This avoids
duplicating the log contents.
* app.js: In the per-package test list, add link to latest autopkgtest log
and artifacts (other output files).
* debci-status --all: Output a list also if there is only one package.
* debci-status --all: Sort package names so that they appear sorted in the
web UI.
* Avoid some error messages noise when calling generate-index without any
existing test data.
* Drop debci-test --reason option and make debci-batch put the reason into
the status directory directly.
* Fold debci-cleanup into debci-generate-index.
[ Antonio Terceiro ]
* schroot backend: pass --session-id to adt-virt-schroot. This makes it
easier to debug test suite that leave rogue schroot sessions behind.
* rename `debci` binary to `debci-batch`. `debci` stays reserved for
providing a entry point binary like `git` that will invoke subcommands
that are kept out of $PATH to avoid cluttering the namespace.
* debci-batch: make batch locking suite- and architecture-specific
* debci-batch: add --offline and --online options. --offline will prevent
new test runs from being started, and --online will revert to normal
operation.
* Rename scripts/*:
- scripts/process-package: absorbed by the previously almost empty
bin/debci-test
- scripts/list-dep8-packages: absorbed by the previously almost empty
bin/debci-list-packages
- scripts/list-dependencies → bin/debci-list-dependencies
- scripts/setup-chdist → bin/debci-setup-chdist
- scripts/blame → bin/debci-blame
- scripts/cleanup → bin/debci-cleanup
- scripts/generate-index → bin/debci-generate-index
* add `debci` binary to not clutter the namespace. When you call `debci
COMMAND` it will set PATH to include the debci bin/ and then call
`debci-COMMAND` with the supplied arguments.
* debci-blame: replace usage of deprecated IO#lines method with IO#each_line
* debci-generate-index: accept duration as command line switch
* backends/schroot/create-testbed: drop usage of eatmydata to avoid breaking
test suites that actually depend on a real fsync(). To compensate the
performance loss when installing packages, set dpkg `--force-unsafe-io`
option in a configuration file (Closes: #747308).
* Added Ruby bindings and the corresponding documentation
* Dropped usage of faketime in the test suite (Closes: #747715)
* backends/fake: output a timestamp like the ones adt-run outputs
* Added new debci-config, a helper to output value of debci configuration
values
* debci-generate-feeds: added generation of per-package and global Atom
feeds. Called from debci-batch
* public: now using FontAwesome from fonts-font-awesome package
* public/app.js: add a link to the per-package Atom feed
* public/style.css: make the status indicators use FontAwesome icons
-- Antonio Terceiro <email address hidden> Thu, 15 May 2014 12:34:33 -0300
-
debci (0.5.2) unstable; urgency=medium
* scripts/process-package: skip blame check without a dependencies diff
* lib/*.sh: guard `set -eu` to avoid doing that on interactive shells
* lib/functions.sh: fix list_binaries to work against the sources index
- This fixes listing binaries coming from packages that were binNMUed
-- Antonio Terceiro <email address hidden> Sat, 05 Apr 2014 20:03:22 -0300
