Introduce new AssistData class that contains all data
the framework automatically generates for assist. Currently
populated with a very simple tree structure representing
the app's view hierarchy.
Reworked how we populate the class name for accessibility
info, so this is provided through a new method call on View
that subclasses can override. This method is also used
to populate the class name in AssistData.
Change-Id: Ibd0acdc8354727d4291473283b5e4b70894905dc
This change adds support for duplex printing. The print UI now has a duplex
option which allows the user to choose one of the supported duplex options
by the currently selected printer. The chosen duplex mode is propaged to the
print service that manages this printer.
Change-Id: I807ba9da2723531535c0e1e33f2f4e1b503a54b0
We are updating the printed content asynchronously. If the user cancels
printing before the first update is complete the spooler components are
in a destroyed state but we attempt to handle the completed update resulting
in a crash. Now if printing is cancelled we ignore the update result.
bug:18525491
Change-Id: I1ef47a17de19896cba2c7a1dd2bfc205065e0a5a
A recent change modified the way we destroy the remote renderer from
asynchronous to synchronous. This caused problems since it was possible
that the remote rendering service is unbound while we are reading the
contents of a rendered page. As a result the reader was blocking on I/O
and the print spooler was getting into a locked state that required a
restart of its process. Now the remote renderer is destroyed
asynchronously.
bug:18498626
Change-Id: I1312bf808f30430728b4038dd4be43c55d2be825
1. Available printers are provided by a loader driven by the activity
lifecycle. After a rotation of an app that does not handle rotation
the print activity is destroyed in a delayed fashion. When another
print task is started the loader of the destroed activity gets two
cancellations, one when the activity is stopped and one when it is
destroyed but when the second cancellation happens the loader has a
bad state that it is loading while it is not and a NPE occurs.
2. If a rotation happens early before the print activity is fully initialized
(there are asyncrconous operations that take place) we get a NPE as
the print preview controller is not created yet.
3. A lockup of the print preview process occurs if the print print
activity is destroyed after we initiated binding to the rendering
service but before we receive the remote interface. In this case no
remote service is received as exepcted since we already unbound from
it resulting in a missed signal.
bug:18238590
Change-Id: I81817d81702f649ded97dfbab7d7bba28f22fa91
If apps are writing malformed content (typically not a PDF file) or if the
PDF content they provide to the print system is password protected, are now
crashed as both of these are app bugs.
bug:17636435
Change-Id: Ifce6a3199e587448dd38f6a84290a965c24b698b
When rendering a PDF file for print preview we take into account
the selected print options such as paper size, orientation, etc
without modifying the document. To print we send the doc in its
original form and the print options so the print service can apply
the necessary transforms in addition to the optional custom options
it supports. When saving to PDF we have to actually change the
document as we act as a print service.
bug:13545980
Change-Id: Icdcecf962bec6ff742cc6015df5af9d9086ce760
When printing from two apps at the same time the second print UI is
getting stuck. There were a couple of issues here:
AdapterView was not notifying for item selection if the data changes
after scheduling a dalayed selection notification and the notification
execution. The code assumed that a layout pass will occur and posponed
the notification after the layout pass but it is not guaranteed that
such a layout pass will occur. Now we delay only if a layout pass is
being scheduled.
Also when binding to the PDF rendering service the print spooler was
using the same intent and as a result two print activites were getting
the same renderer instance while they should get separate ones. Now
we use different data in the intent to ensure we get separate renderer
instances.
Change-Id: I6aa7c7b041957804b4273549dd837a6d70064efc
If the printing app with the print UI on top is killed from recents we get
a crash because: 1) the remote print document was not transitioned to a
failed state if the printing app dies (this is an unrecoverable failure);
2) the print preview controller was destroyed asyncronosly during which it
also asynchronously disconnects from the rendering service which however
happens after the system has already cleaned up all connections of the print
UI activity as it is being destoryed.
bug:18109386
Change-Id: If6200b14a8aa90622228bbb659e9c4962226f561
We order printers based on past usage. In some cases the ordering
does not work. The reason for that was an incorrect assumption that
the entries in an ArrayMap are ordered in the order they are added.
bug:18109283
Change-Id: Ie367e4ca5e6bd79f335a060074c9211054e3a931
One can save only one copy to PDF. Showing more that one copy when saving
to PDF enven in a disabled input box is confusing to the user. We now set
the copies to one when the PDF printer is selected.
bug:17631303
Change-Id: Ia04d144efc899e7f3836c19349d9c87f3ae8f0f0
When saving to PDF we bring up the file picker UI which covers
the print UI. In this case the printing app may get killed as
it is not in the forground (the system is bound to the print
spooler so it cannot get killed). If that happens we are ending
up with an empty file. However, if the file chooser UI is up
we already have everything from the app we need and it dying
does not matter. This change takes care if ignoring printing
app death while saving to PDF.
bug:17922948
Change-Id: I10d808e3c3e93e850dbc2a948f2482381a887928
The first issue is that the save to pdf button is enabled
before the first layout completed and at this point we do
not know anything about the document. The second is that
if a layout is in progress and we select save to pdf we
end up starting the documents UI twice and finishing the
second instance carshes the spooler.
bug:17676878
Change-Id: Id0b1f06e4d377347286c2dde09669d431441234b
The custom settings activity of a print serivce can change the selected
pages and the spooler was not checking whether the returned pages were
valid leading to a crash.
bug:17678553
Change-Id: Id8ea3d482d2909b97a06752d61b416dfe5948483
It is possible that a buggy app breaks the contract and provides
content to be printed in format other than PDF. This was leading
to a crash in the print spooler. This change fixes the crash and
shows a user friendly error message.
bug:17642690
Change-Id: I5a4acb06080a152562655da6851467b3e71d8658
It is possible that two providers for the same page are assigned
to two preview views despite temporariliy while rebinding. We were
however releasing the provider from the view which is being bound
if not for the same page which as a result was cancelling rendering
of this page which is provided to another view via another provider.
Hence, having an empty page. Removed the unnecessary code to release
a provider when binding the view - release is happening only if the
view holder is recycled.
bug:17515670
Change-Id: I197438c16cfdb363f521cd1320f154da54eea5f5
Saving to PDF a subset of the pages requires trimming the
undesired pages and when this is done writing the ready doc
to a URI. To write the file to URI we have to obtain it but
it is aquired by the shredding task and never released. Now
we are releasing the document immediately after trimming it.
bug:17631301
Change-Id: I4db7966c65c75f0f14c3cb52fd83b4d8fd5a4687
If the user presses back before the remote print document
update has completed we get a crash. The reason is that
after the update completion we were trying to finish the
activity if we are in a cancelled state but this is not
needed as we aready handled the back key so we will finish.
Handling finish twice was creating the problem as classes
with lifecycles are not designed to be used after being
finished. In particular, we were calling doFinish() twice.
bug:17630561
Change-Id: If418f237a2def7c8e4a072ac8826283f4dd7fc85
In the spooler we have the renderer reading a file to visualize
content and the app writing a file to produce the content. Since
we have to swap the file under the renderer we have a mutex file
provider that both parties can request, use when released, and
release when required. This enables us to request the file which
closes the renderer and when the renderer is closed ask the app
to write some more pages, then open the renderer, and so on. The
mutex file provider was throwing of a thread that does not own
the file thries to relase it which is not needed, this should be
just a nop.
bug:17607134
Change-Id: Id6a2ce92d70077f57978b95315648faf02c13c68
The spooler communicates with he remote PDF renderer asynchronously.
When print is confirmed we close the renderer, destroy it, and unbind
from its service. If we unbind from the service after the print activiy
is finished we get a crash. The bug was that we did not wait until we
disconnect from the remote renderer before finishing the print activity.
bug:17583115
Change-Id: I55b0135f9c5658b3a4fda2901b8b3bdef044e211
When user chooses to save to PDF we open the document chooser UI.
Now if the user now backs to the print UI without choosing a file,
the print UI has all options disabled and the back button does
not work - the user is stuck. This is because we were not
transitioning the print UI in the correct state.
bug:17557454
Change-Id: I385129e106ed41564b89137ca54d409127ba8ff0
Removed the restriction for the page content providers to
be singletons as the RecyclerView may hold on views with
providers and ask us to bind other views for the same page.
Note that the views that RecyclerView is holding will be
soon rebound or recycled just they are not at the time we
bind other views for the same position. We were guarding
against this but caused some pages not appearing sometimes
on scrolling. Also now keeping the media size and magins
of recycled page view (will be updated if needed on
rebinding) or avoid changes of the background to null as
setting it to not null later causes an undesired layout pass.
bug:17537922
Change-Id: If029f6c413e15d4f5c1ba1ac5e4d52ad2cad5904
The operations of the remote renderer have to be performed in order, open,
render some pages, close. One of the tasks was executed on the wrong executor
resuling in a race and an occassional bad state.
Also fixed a NPE if the preview list is flinged and then the user presses
back.
bug:17537922
Change-Id: I5048078ba2b875a2a8335f3a4324afaa34d014a2
We were not clearing the tracked printer when the print activity is
paused and then not setting the tracked printer when the activity
is resumed.
bug:17525271
Change-Id: Ib0fb29006bfcf5ad03dce24b3407cd96dc421442
It is possible that we try to switch framents after the print
activity finish method is called. In this state the activity
is going away and trying to perform a fragment operation throws.
Now if we are finishing no fragment opration is performed.
bug:17520007
Change-Id: I0dd935d3cdaeddca31fccf49bd5de97e878abb9d
The bound pages in the adapter were not properly tracked, more
specifically we did not remove pages from the bound set ending
up in a bad state. Also when the printer changes and if the
new one has different page size the content disappears due to
us not notifying for the data set change.
bug:17515670
Change-Id: Iac74e864609012f6804584fa133e87c27dc186a5
The code executed in the shredder completion callback must
be called on the main thread but instead it was called on
another one. This led to a crash.
bug:17514533
Change-Id: Id9e86d38a90fedadc60f967b193470fd83eb3362
We laizily request printer capabilities and when they are reported
we were not updating the preview and as a result the content was not
rendered, hence pages showing an empty state.
bug:17514117
Change-Id: Icfee7eaa1ab63f5e4d7655d4057ef2110c27cd43
If the printer capabilities are reported with a delay which is pretty
common we were trying to evaluate the preview page dimensions without
having a media size resulting in a NPE.
bug:17501953
Change-Id: I6765137f61b8fd060e9eabff07e632875d934963
A recent bug fix exposed another one where we were not updating
the internal state of the remote print document after it finishes
updating. This resulted in a crash on every print.
bug:16966145
Change-Id: I398ff7976533241e3d4cb6cd27f69cdc25e67be9
The logic when a page view to request the content had a logical
error and it was getting in a bad state where it never requested
the content anymore. This change simplifies the logic and minimizes
the kept state.
Removed unnecessary calls to recycle for a bitmap as according
to the graphics folks we should not be calling this API anymore.
bug:16966145
Change-Id: Idda8314b43343136d3b867167261790481116fac
