We use django-reversion to track history of admin/maintainer changes to
the site, and part of our extension on top of django-reversion involves
annotating each "revision" with a description of the changes that were
made. In django-reversion 2.0.0+ the pre_revision_commit signal that we
were using to do this annotation is gone in favour of just using
Django's standard pre-save signal. This was a little challenging to
adapt to for our purposes but not impossible.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
resolve_variable isn't available with Django 1.10+, and it's not even
used here, so just drop it from the import line.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
The patterns() function is deprecated in Django 1.8 and gone in 1.10, so
we should switch over to the new list format.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Rename publish() to publish_view() and call it directly from the view
rather than using a string name (which is not possible in Django 1.10+).
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
The utils.setup_django() costs a lot of time, but both update.py and
update_layer.py call it, so move layer validation from update_layer.py
to update.py to avoid calling update_layer.py when possible can save a
lot of time.
Now we don't have to call update_layer.py in the following cases:
* The branch doesn't exist
* The layer is already update to date on specified branch (when no
reload)
* The layer dir or conf/layer.layer doesn't exist
We can save up to 98% time in my testing:
$ update.py -b master --nofetch [--fullreload]
Before Now Reduced
No update: 276s 3.6s 98%
Partial update: 312s 87s 72%
Full reload: 1016s 980s 3%
Note:
* All of the testing are based on --nofetch
* "No update" means all layers on the branch is up-to-date, for
example, when we run it twice, there is no update in the second run,
so we only need about 3s now, which is the most common case when we
use cron to run it per half an hour.
* "Partial update" means part of the layers have been updated.
* "Full reload" means all of the layers have been updated.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
This makes it easy to see which layers failed. For example:
ERROR: Failed layers on branch master: openembedded-core meta-python
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
We have an update.py running periodically in background, but we also
need to run it manually, for example, run it to update actual_branch,
the manually run usually failed because can't get lockfile, we have to
run it again and again. A timeout option helps a lot in such a case. Now
the following command can make sure we can run the command successfully:
$ update.py -b master -a actual_branch -t 2000
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
The previous commit broke the layer order, e.g.:
A -> B -> C -> D
The algorithm is checking the dependencies one by one, and until we find D, add
D to layerquery_sorted, and add it "collections", the one in "collections"
means it's dependencies are OK, then C, B and A will check against collections,
so that update_layer.py will update them one by one. The previous commit added
A/B/C/D to collections directly, so that when check against it, all the
dependencies are met, thus broke the layer sorting, and then there would be
failures if we pass layer A to update_layer.py before B, C and D (suppose they
are newly to database). This commit fix the problem.
BTW., why I use collections to record the one whose dependencies are matched,
but not directly use layerquery_sorted, it is because collections contains both
the ones to be added/updated and the ones in database, but layerquery_sorted
only contains the ones to be updated/added.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Add a page with basic statistics for the index - number of layers,
recipes, classes, machines and distros on an overall basis (distinct
names) and per branch, since I've been asked a few times for this kind
of information. It's currently only linked from the Tools menu for
logged-in users, but the URL will work for anyone.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
If the user hit Ctrl+C during the initial info gathering then it didn't
break out of the loop in update.py, so you had to hit Ctrl+C for as many
layers as were involved in the update. Look for exit code 254 from
update_layer.py and stop if it is returned since that indicates Ctrl+C
has been used.
Additionally, ensure we return exit code 254 and print a message from
the main update script when it is interrupted in this way.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
For reasons that are not immediately clear to me, if you don't specify a
name when adding the ViewSet it takes one from the model used in the
queryset in the ViewSet. The new layers view uses LayerBranch and so it
silently replaced the layerBranches URL in the router, even though the
URL itself was still present. Unfortunately that's broken Toaster.
Specify a name to restore the old URL.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
It was a bit awkward to query layers externally - you had to do multiple
queries and you couldn't get the YP Compatible version info at all. Add
an additional LayerBranch-based view that exposes the branch name,
layer fields, YP Compatible Version and active maintainer information
with just one call.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Django REST Framework now requires a field specification for every
ModelSerializer, so specify '__all__' to retain the current behaviour.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Layer maintainer and layer note information wasn't available through the
REST API since it wasn't needed for Toaster, but for other uses it is
useful.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
If you query on a boolean field you can use the string "False" to match
False in the database; however if you try the same with __isnull then
the query will match every record which is obviously undesirable. If
__isnull is being used, then convert the value to a boolean so that the
query works properly.
An example of this type of query:
http://127.0.0.1:8000/layerindex/api/layerBranches/?filter=yp_compatible_version__isnull:false
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
tinfoil is not needed in cases like the layer is already up-to-date or the
layer is invalid, so only init it when needed.
This can save about 1min when run "update.py -b <branch>" (124 layers).
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
This can save a lot of time, here is my testing data when PARALLEL_JOBS is 10,
this is the fetch time only, I hacked it to stop when the fetch is done to get
the data (124 layers):
$ update.py -b <branch>
Before: 2m30
Now: 16s
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Fixed:
$ ./update.py -l foo -b new_branch
INFO: Fetching remote repository foo
DEBUG: run cmd 'git fetch -p' in 'foo'
[snip]
DEBUG: run cmd 'git checkout origin/new_branch' in oe-core
ERROR: error: pathspec 'origin/new_branch' did not match any file(s) known to git.
The "new_branch" is newly created, it doesn't exist in local repo since it
isn't fetched, it only fetches the layer specified by -l, so only the foo layer
is fetched. This patch fixes problem.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
The obsolete dependency is the one which is in database but not in
conf/layer.conf anymore. The old code had a problem for newly created
layerbranch, the new layerbranch has no dependencies, so no need remove. And it
had a side effect was that when need_remove was cleaned up, it would be set
again in the next for loop, thus might wrongly remove dependencies. This patch
can fix the problem.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
If your classic recipe search returned a single result, then since
commit c8c25fb641 you got a standard
recipe page instead of the proper page with fields you can edit. To fix
it, just drop the redirection functionality from the classic recipe
search, since we don't really want it here.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
We were using the layerbranch id to search for the specified layer,
which is most likely to return either no results or results for the
wrong layer. We can also avoid specifying the id field at all here as
the filter() function can handle real objects.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Unsatisfied layer dependencies shouldn't error out of the script -
broken metadata isn't supposed to terminate the index update process.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
We need to set a default for this variable or you get an
UnboundLocalError if no non-root superuser/staff accounts are set up in
the database.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
If you use a traditional database engine (such as MySQL/MariaDB) then
maximum character field lengths are enforced, however depending on the
configuration this may result in an exception rather than a warning and
truncation and will also break the per-layer transaction. To avoid that
ugliness, add a signal handler to do it internally, which as a bonus
lets us know if field lenghts are too short for data when using database
engines that don't enforce lengths (e.g. SQLite).
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
If a database error occurs when we save a recipe (e.g. because a
database-level constraint is voilated) it will mess up the transaction.
Unfortunately that means we need to break out of updating the entire
layer rather than catching the error, because if we do catch it we just
get errors on every update after the initial error; failing early and
giving up on the transaction is a little better in terms of not filling
up the update logs with further useless errors.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
* Allow the model's __str__() function to control what is shown for each
dependency item as per other model admins
* Enable searching for PACKAGECONFIGs by recipe name
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Use the more typical handling of the return of get_or_create() and check
if the item was created before saving.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Fix duplicate PackageConfig records being created each time a recipe
is being updated - we need to delete the old ones first.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Since we added the "packageconfig" to Recipe we need to exclude it from
the admin or we see an error when we try to open ClassicRecipe in the
admin site.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Added a model for the PACKAGECONFIG variable, which has a one to
many relationship with the Recipe model.
Added models for static build dependencies and dynamic build
dependencies, both of which have a many to many relationship with
the Recipe model.
These objects are created in update_layer.py and are displayed on the
Recipe detail page.
Added a depends search option for recipes, allowing users to search for
recipes that depend on any particular recipe. Use "depends:recipename"
in the recipe search to activate this.
Fixes [YOCTO #12129]
Fixes [YOCTO #11415]
Signed-off-by: Amanda Brindle <amanda.r.brindle@intel.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Use JavaScript to check if the search box for recipe search is
empty before querying the database. This will prevent the "502
Bad Gateway" error that occurs when the query takes too long due
to the large list of recipes. Since there are so many recipes
spread across the layers in the OE index, there's no point in
allowing a user to search without a keyword in order to browse
the list; it simply isn't digestible as a whole.
For the Machines, Classes, and Distros pages, the search behaviour is
unaffected, however to make it more obvious that you can browse the list
add an explicit "browse" button.
Fixes [YOCTO #11930]
Signed-off-by: Amanda Brindle <amanda.r.brindle@intel.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Add another tab to search for classes.
Fixes [YOCTO #11207]
Signed-off by: Amanda Brindle <amanda.r.brindle@intel.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Don't assume that every model will have a "search_allowed_fields"
attribute - if it doesn't, default to all CharFields in the model. This
fixes searching via the REST API.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
When publishing a layer, send an email notification to all of that
layer's maintainers. Include information on how to edit the layer, plus
contact details for the first active staff user if there are any
problems (we could make this configurable in future, but for now this is
sufficient).
Fixes [YOCTO #11208]
Signed-off-by: Amanda Brindle <amanda.r.brindle@intel.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
It's really irritating to be forced to specify a value for this field
especially as it'll get auto-populated by the update script. Set
blank=True to allow that. While we're at it, touch up the description a
bit to make more sense.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
If you're diagnosing problems with the bitbake server when running the
update script, then you need to be able to look at
bitbake-cookerdaemon.log, but you couldn't do that after the fact
because the temporary directory it gets written out to was being
unconditionally deleted. Add a --keep-temp option which preserves it and
some debug messages to tell you where it is.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Allow admin to create Yocto Project versions with the version name,
description, an icon link that represents that version, and a link that
contains more information about the version.
Admins who have the "set_yp_compatibility" permission can then
decide if a layer has Yocto Project compatible certification,
and if it does, admin can choose which version of Yocto Project
the layer is compatible with. If a layer is deemed compatible,
the version's icon will appear next to the layer's name, and
the icon be a clickable link to a page with more information.
Fixes [YOCTO #11452]
Signed-off-by: Amanda Brindle <amanda.r.brindle@intel.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
If a log level is set on the command line with -q/-d,
set tinfoil's log level to the appropriate log level.
Fixes [YOCTO #11931]
Signed-off-by: Amanda Brindle <amanda.r.brindle@intel.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
This patch adds asynchronous task execution using a Celery backend
and RabbitMQ task queue, so that the layer submission process to
proceed even in the event that sending the notification email fails,
and establishing an asynchronous execution mechanism that we can use
in the future e.g. for triggering parse operations from the web UI.
This pertains to bug 11197:
https://bugzilla.yoctoproject.org/show_bug.cgi?id=11197
It updates the README to reflect the installation and configuration
of a basic RabbitMQ setup, adds a 'tasks.py' file to contain task
definitions, updates the 'edit_layer_view' function to send
emails to administrators about new and updated layers asynchronously,
modifies the 'settings.py' to include a default configuration
for a RabbitMQ connection, and updates the Dockerfile to start a Celery
worker alongside the Gunicorn daemon.
Fixes [YOCTO #11197].
Signed-off-by: Diana Thayer <garbados@gmail.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
Before, if a user edited a layer's name, they would be redirected
to a url utilizing the old name and then receive a 404 Page not
found error. Now, the url utilizes the new name.
Fixes [YOCTO #11932]
Signed-off-by: Amanda Brindle <amanda.r.brindle@intel.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
At the moment the URL that is presented in the review email will have
http:// or https:// prefix depending on what the user who submitted the
layer was using, but that's irrelevant - we actually want https:// if
the server is capable of it since the reviewer may be redirected to log
in (and Django's login_required decorator will always redirect to the
login page if http is being used as far as I observed, which is a bit
annoying if you are already logged in.) Add a setting which if enabled
will substitute https:// as the prefix.
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
The branch is not needed any more when it has been removed from the repo, so we
also need remove its layerbranch, otherwise it still can be got from the web,
which causes confusions.
Note, we have to move the location of tinfoil's code to avoid creating tinfoil
when not needed, otherewise, if tinfoil is created (but not needed), and the
program exists earlier before "try ... finally" block, then tinfoil.shutdown()
doesn't run so that it is not shutdown. Move the code back, right before where
it is needed can fix the problem.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
-p, --prune
Before fetching, remove any remote-tracking references that no longer exist on the remote.
Fixed:
$ git push origin :test_branch
$ ./update.py
The test_branch was still in fetched local repo which was incorrect, it should
be gone since it has been removed by upstream.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
It doesn't need to be so ahead since we only need it when writing database, and
a following patch will remove layerbranch from database when the branch had
been removed from the repo, it's not easy to do the work in
transaction.atomic() block.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
It never works since it is in the middle of transaction.atomic() block, and
update.py doesn't need it any more, so remove it.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
* Problems
The update.py couldn't handle new (not only new branch, in fact, but also
existing branches, see below for more info) branch well, for example, there are
3 layers: layer_A, layer_B and layer_C, and create new branch "branch_1" for
them, and they have depends:
layer_A -> layer_B -> layer_C
The "->" means depends on.
Then run "update.py -b branch_1", there would be errors like:
ERROR: Dependency layer_B of layer_A does not have branch record for branch branch_1
Though update.py runs "update_layer.py" twice, but it didn't help since
layerbranch was None when it was failed to create in the first run.
The reason is if update.py updates layer_A firstly, it would fail since it
can't find layer_B:branch_1 in database (not added to database yet), similarly,
if add layer_B before layer_C, it would also fail. Only layer_C can be added
(assume it has no dependencies). So we have to re-run update.py again and again
to make it work, here we may have to run update.py 3 times, and more runs are
needed if the dependency chain is longer.
* Solutions:
Make update.py pass layers orderly to update_layer.py according to dependencies
can fix the problem, we can get intial dependencies from tinfoil, add an option
"-i, --initial" to update_layer.py to get them.
Not only new branch, but also existing branches may have the problem, because
collections and dependencies maybe changed in the coming update, so we can't
trust database when the layer is going to be updated, for example, if there are
10 layers in database, and 3 of them will be updated (-l layer1,layer2,layer3),
then we can not use the 3 layers' collections data from database, we need get
them from tinfoil.
* Performance improvement:
It should be the same as before in theory, I have tested it with 97 layers:
- Before: 11m6.472s, but only 75 layers were added, 22 ones were failed, I
have to re-run update.py again and again (maybe 4 times to make all of
them added). So:
(11 * 60 + 6)/75*97/60 = 14m35s
- Now 12m10.350s, all the layers are added in the first run.
So about 2 minutes are saved.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Paul Eggleton <paul.eggleton@linux.intel.com>
