| Commit message (Collapse) | Author | Age | Files | Lines |
|
|
|
| |
Change-Id: I750654c28cb3d9f9aa67bd56e4d8d770dbfde4b4
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The download manager uses threading in a simple way -- it launches two
threads, UpdateThread and DownloadThread, and both are "fire and
forget". This is fortunate for testing, since it means we can
eliminate multithreading and simply run each thread in order, and
everything still works.
This change does just that, abstracting Thread.start() behind
SystemFacade and making FakeSystemFacade put new threads into a queue
and then run through them serially. This simplifies much of the test
code and makes it all much more predictable.
I've simplified the test code as much as possible here and moved a few
more tests over to PublicApiFunctionalTest, leaving only a minimum in
DownloadManagerFunctionalTest, which will eventually be deleted
altogether. I've also improved testing in some areas -- for example,
we can now test that running notifications get cancelled after the
download completes in a robust way.
There is one test case that checks for race conditions and requires
multithreading. I've moved this into a new ThreadingTest class, which
uses a custom FakeSystemFacade that allows multithreading. I've
extracted AbstractPublicApiTest for the newly shared code.
Change-Id: Ic1d5c76bfa9913fe053174c3d8b516790ca8b25f
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
* Three new DB fields, one indicating whether a download was initiated by the public API or not, two to hold connectivity control info. DB migration to add these fields and code in DownloadProvider.insert() to handle them.
* Change broadcast intent code to match public API spec, for public API downloads only. (Legacy code can go away once existing clients are converted over to the new API.)
* Introduce SystemFacade methods for sending broadcasts and checking package ownership; this facilitates new tests of broadcast code.
* Change DownloadInfo.canUseNetwork() to obey new connectivity controls available in public API, but again, retain legacy behavior for downloads initiated directly through DownloadProvider
* New test cases to cover the new behavior
Also made a couple changes to reduce some test flakiness I was observing:
* in tearDown(), wait for any running UpdateThread to complete
* in PublicApiFunctionalTest.setUp(), if the test directory already exists, remove it rather than aborting
DB changes for broadcast + roaming support
Change-Id: I60f39fc133f678f3510880ea6eb9f639358914b4
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Motivation: I need to fix the handling of 302s, so that after a
disconnect, subsequent retries will use the original URI, not the
redirected one. Rather than store extra information in the DB, I'd
like to just keep the redirected URI in memory and make the redirected
request within the same DownloadThread. This involves working with
the large-scale structure of DownloadThread.run(). Since run() was a
~700 line method, I didn't feel comfortable making such changes.
So this change refactors run() into a ~80 line method which calls into
a collection of ~20 other short methods. The state previously kept in
local variables has been pulled into a couple of state-only inner
classes. The error-handling control flow, formerly handled by "break
http_request_loop" statements, is now handled by throwing a
"StopRequest" exception. The remaining structure of run() has been
simplified -- the outermost for loop, for example, could never
actually repeat and has been removed for now. Some other bits of code
have been cleaned up a bit, but the functionality has not been
modified.
There are many good next steps to this refactoring. Besides various
other cleanup bits, a major improvement would be to consolidate the
State/InnerState classes, move some functionality to this new class
(there are many functions of the form "void foo(State)" which would be
good candidates), and promote it to a top-level class. But I want to
take things one step at a time, and I think what I've got here is a
major improvement and should be enough to allow me to safely implement
the changes to redirection handling.
In the process of doing this refactoring I added many test cases to
PublicApiFunctionalTest to exercise some of the pieces of code I was
moving around. I also moved some test cases from
DownloadManagerFunctionalTest. Over time I'd like to move everything
over to use the PublicApiFunctionalTest approach, and then I may break
that into some smaller suites.
Other minor changes:
* use longs instead of ints to track file sizes, as these may be
getting quite large in the future
* provide a default DB value of -1 for COLUMN_TOTAL_BYTES, as this
simplifies some logic in DownloadThread
* small extensions to MockResponse to faciliate new test cases
Change-Id: If7862349296ad79ff6cdc97e554ad14c01ce1f49
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
In previous refactoring I had combined the code for starting a
download (from DownloadService.insertDownload()) and restarting a
download (from DownloadService.updateDownload()). I'd missed the fact
that the former checks DownloadInfo.isReadyToStart() while the latter
checks DownloadInfo.isReadyToRestart(). This cause a race condition
where multiple startService() calls during a download would cause the
service to try to spawn a second thread for the same running download.
I've fixed the bug and added a test case to exercise this.
Change-Id: Ia968331a030137daac7c8b5792a01e3e19065b38
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This change introduces support for a maximum download size that may go
over a mobile connection. Downloads above this limit will wait for a
wifi connection.
To accomplish this, I moved a lot of the logic for checking
connectivity info into DownloadInfo itself. I then moved the code to
call these checks from DownloadService, where it would call the checks
before spawning a DownloadThread, into DownloadThread itself. This
makes it simpler to check connectivity after we get Content-Length
info. It also eliminates the risk of a race condition where
connectivity changes between the check and the actual request
execution.
I realize this change reduces efficiency, because we now call into
ConnectivityManager/TelephonyManager twice per DownloadThread, rather
than once per DownloadService "tick". I feel that it's OK since its a
small amount of computation running relatively infrequently. If we
feel that it's a serious concern, and that the efficiency issues
outweigh the race problem, I can go easily back to the old approach.
I've left out the code to actually fetch the limit. I think this will
come from system settings, but I want to double-check, so I'll put it
in a separate change.
Other changes:
* simplify SystemFacade's interface to get connectivity info - rather
than returning all connected types, just return the active type,
since that should be all we care about
* adding @LargeTest to PublicApiFunctionalTest
Change-Id: Id1faa2c45bf2dade9fe779440721a1d42cbdfcd1
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Provider changes:
* new many-to-one DB table holding headers for each download. since
there was no real migration logic in DownloadProvider, I implemented
some.
* DownloadProvider.insert() reads request headers out of the
ContentValues and puts them into the new table
* DownloadProvider.query() supports a new URI form,
download/#/headers, to fetch the headers associated with a download
* DownloadProvider.delete() removes request headers from this table
Service changes:
* made DownloadInfo store request headers upon initialization. While
I was at it, I refactored the initialization logic into DownloadInfo
to get rid of the massive 24-parameter constructor. The right next
step would be to move the update logic into DownloadInfo and merge
it with the initialization logic; however, I realized that headers
don't need to be updatable, and in the future, we won't need the
update logic at all, so i didn't bother touching the update code.
* made DownloadThread read headers from the DownloadInfo and include
them in the request; merged the custom Cookie and Referer logic into
this logic
Also added a couple new test cases for this stuff.
Change-Id: I421ce1f0a694e815f2e099795182393650fcb3ff
|
|
|
|
|
|
|
|
|
|
| |
The public API is getting deeply reworked for forward compatibility,
but since the Download Manager and the Browser need to continue using
the old API, a separate copy is being kept on the side.
Bug: 2245521
Change-Id: I85eff6ba9efc68600aa80e8dffa6720b0f2ed155
|
|
|
|
|
|
|
|
|
| |
The restart time is used multiple times during the same pass,
and it needs to be consistent across calls. Otherwise, it's
possible for a download to not be restarted immediately and
to not be scheduled for a future restart.
BUG=2055624
|
| |
|
|
|
|
|
|
| |
This fixes a number of style violations that weren't caught by automated
tools and brings those files closer to compliance with the official style
guide for this language.
|
| |
|
|
|