When a URL is malformed because the path does not start with a slash,
we assume that the path starts with the first character that is not
valid in the host and insert a leading slash. This is the reason why
the regex for the path component does not force a leading slash.
Bug: 1011602
Change-Id: I8efe46c058d2ee2d1a6a4406ee25dc021315222b
Merge commit 'f17bd19acbb88429b50a1497bfec0752b456d5c6'
* commit 'f17bd19acbb88429b50a1497bfec0752b456d5c6':
Use new nativeLibraryPath field for NativeActivity
Merge commit '2cb3e83654c99e202c170d9d0237d8d1f4054354' into gingerbread-plus-aosp
* commit '2cb3e83654c99e202c170d9d0237d8d1f4054354':
Use new nativeLibraryPath field for NativeActivity
Instead of being naughty and guessing the native library path from
dataDir, use the new nativeLibraryPath field because it can be on SD
card now.
Change-Id: I284bde42e0b0114366e412eb7212af22b47208d8
The WebAddress class provides a lenient parser for URLs. Currently, it
identifies the host portion with the regex [<chars>]+(\.[<chars>]+)* where
<chars> is the set of characters valid for the host name. This pattern excludes
the case where the host ends with a dot, which is valid possibility. As a
result, any trailing dot is pushed into the path component. Since we add a
leading slash to the path if one is missing, the result is a path that begins
with '/.'.
This fix changes the host regex to [<chars>]+[<chars>\.]* which allows trailing
dots and fixes the problem.
Bug: 2337042
Change-Id: I310512531787e0f742988f5d815ad944fd39e059
There was a deadlock when ConnectionHandler exited the loop in SocketPipeThread, and would call the onFinishedCallback, which called the synchronized method in Forwarder that would
deadlock. Changing the Forwarder class solved the issue and made it more efficient.
Change-Id: I947450a19573f2e88274b1ebc7b77d4df6afffa7
Merge commit 'a060b5f73863f4b5f161674a9c5ac0a0d8d76072'
* commit 'a060b5f73863f4b5f161674a9c5ac0a0d8d76072':
Reduce footprint of Signature from ~7000 to ~1448
CameraSourceTimeLapse now decides whether to use still or video
camera automatically. It checks if the passed in size is a valid
preview size and if it is, then uses the video camera else uses
the still camera.
Removed from StagefrightRecorder the support to set parameter
useStillCameraForTimeLapse.
Change-Id: I71f5b0fc7080ca524792381efe918d22e41a7f36
Merge commit '6dc003a73f7f01aa5dd5831e777e2aa44f65b032' into gingerbread-plus-aosp
* commit '6dc003a73f7f01aa5dd5831e777e2aa44f65b032':
Reduce footprint of Signature from ~7000 to ~1448
Merge commit 'd6fd133d18c189ab112daef9dc649d0796df5d99'
* commit 'd6fd133d18c189ab112daef9dc649d0796df5d99':
Not all audio source has the drift time information
Merge commit 'acc4e041d27c24b45fb190c493717fda987b52c2'
* commit 'acc4e041d27c24b45fb190c493717fda987b52c2':
DO NOT MERGE. Some stuff that is Not Ready.
Merge commit '3326f8043858152ea2a407a9a4c20fdcdc859e73' into gingerbread-plus-aosp
* commit '3326f8043858152ea2a407a9a4c20fdcdc859e73':
Fix the last remaining gaps in our Java 6 API.
Merge commit '7627c221c6ce8bf433bf207b002eabe7236c4757' into gingerbread-plus-aosp
* commit '7627c221c6ce8bf433bf207b002eabe7236c4757':
DO NOT MERGE Tweak overscroll
Signature had lazy initialization of the mString member when needed, but
it would stick around forever when initialized. Each package had one or
more Signatures that would be ~7000 bytes each. At a couple hundred
packages, that's over 1.3MB just for signatures.
Whenever packages.xml was written out, it would write the Signature for
each program as well which happens at boot thereby initializing the
mString member pretty much immediately.
Change-Id: Idb882ffeca2861b3e87437fc83f001710d6c0441
