AlertDialog.Builder.show() will create an AlertDialog before it show. This is the source code snippet:
/**
* Creates a {@link AlertDialog} with the arguments supplied to this builder and
* {@link Dialog#show()}'s the dialog.
*/
public AlertDialog show() {
AlertDialog dialog = create();
dialog.show();
return dialog;
}
github: close#96
When a plugin throws an unchecked exception, we're not catching it
anywhere and so the error callback is not being called.
This change adds a try/catch to catch such exceptions.
The online bridge toggles between on & off. Turns out that starting with
Android 4.3, navigation resets the online state, so the code had a 50/50
chance of being caught in the wrong state.
We now forcefully reset the online state on page transitions.
- fixed grammer in log message in Config.java
- updated the class javadoc in CordovaActivity to catch up to reality: use config.xml instead of set*Property() methods.
- added deprecation message for SetFullscreen
- add javadoc target to ant script. It must be invoked explicitly to run.
- ignore the generated javadoc html directories.
- clean up javadoc errors in source files.
- upon invoking 'clean' target, erase generated jar and javadoc
- Found stale cordova.js in the test project. Changed ant script to copy in
a fresh one in the pre-build stage. Removed stale copies from git, added
to .gitignore. This is what was causing the test failure when sending
javascript to the webview.
- Found almost-stale copy of device plugin, did the same as above.
- Fixed spacing in log messages in CordovaWebView.
- Updated README.md with current information.
- Removed dollar signs that were previously accidently appended to the license.
- Added cordova version to html pages, next to the other metadata.
- Fix incorrect separator in the viewport meta tag.
- Changed old-style <log> tag to <preference name="loglevel"> in config.xml.
- In BackButtonMultiPageTest, increased the TIMEOUT so the deviceready event
can trigger the referenced javascript, improved the setup() method to
explicitly load the starting point.
Turns out our Quota logic has been wrong all along. If we were to
actually track the quota needed, we'd need to store a map of
dbName->quota for all dbNames. Instead, we just set a really high quota
since we don't ever want to decline storage.
I *think* this fixes all of the SECURITY_ERR exceptions we've been
seeing. Even those on Honeycomb / ICS.
Andrew Grieve