This information has been added to the main README.md file, so is
now redundant.

This is not needed now that PropertyValueAliases is used for script
mapping.

The mapping of the script labels in the UCD data to ISO 15924
script tags is now done using the sc property map in the
PropertyValueAliases data.

This has the following benefits:

    1.  It removes the dependency on the IANA subtag registry.

    2.  It ensures the scripts are correct as specified in the
        UCD data files.

Conflicts:
	dictsource/dict_phonemes
	phsource/compile_report

Changes to:
en, pt,
it,  implement ordinal numbers.
eu, stress positions.
es,  revert change to [t] and [k]
de, change to sounds "pr" and "kr".


git-svn-id: https://espeak.svn.sourceforge.net/svnroot/espeak/trunk@387 d46cf337-b52f-0410-862d-fd96e6ae7743

Conflicts:
	dictsource/dict_phonemes
	phsource/compile_report

Language changes: an, de, en, it, pt, ta, gu, hi, pa, ur.


git-svn-id: https://espeak.svn.sourceforge.net/svnroot/espeak/trunk@386 d46cf337-b52f-0410-862d-fd96e6ae7743

This reverts commit 6ac54860a1.

This change was preventing the voice data from being installed.

The Android 5.0 Text-to-speech output settings dialog checks that
the languages from CheckVoiceData use the ISO3 language and
country codes. This change matches that expectation.

This improves the performance of looking up a voice by name.

As of Android 5.0, TextToSpeech.setLanguage calls the new
TextToSpeechService.onGetDefaultVoiceNameFor(lang,country,variant)
and uses the specified voice name.

The default Android 5.0 implementation of onGetDefaultVoiceNameFor
does not work with eSpeak. This provides a working implementation.

This fixes the TextToSpeechTest.testLanguages tests on Android 5.0.

The default Android 5.0 implementation enumerates the available
locales. This enumerates the voices returned by eSpeak.

This changes the getLanguage helper to return the result from
getLanguage on all platform versions, disabling the deprecated
message on API 21 and above.

This adds method declarations for the new voice APIs, forwarding
them to the default implementations.

This creates a custom hamcrest matcher that maps the TextToSpeech
LANG_* constants to strings. This makes it easier to read the
text output when the tests fail.