DBus signals about change of locale property for given address book are
deferred until all address books will update their locale.
Depending on the number of address books and their size delay between
changing address book's locale property and receiving signal about that
takes about one or two minutes.
Flushing DBus properties after changing locale property of address book
causes immediate emission of DBus signal.

(cherry picked from commit 69e94310)

	Commit "addressbook: fix localed race condition" brought from
	upstream master checks if the environment variable EDS_TESTING
	is set, while the original code used EDS_LOCALED_TESTING.

	This patch reverts checking to branch specific name.

Fixed crash that occurred when changing locale by migrating locale handling
to EDataBookFactory, which is the server side singleton. Crash happened because when
locale changed, there was an attempt to access an EDataBook instance while it was in
course of being freed in another thread.

Cherry picked from commit in master:
EDataBookFactor / EDataBook: Migrated locale handling to EDataBookFactory
c8914fc7

Fixes BGO #723276

This reverts commit 129901a6.

This reverts commit 2b61817f.

Modified system-address-book.source.in to configure sort keys
by default.

Sort keys specified for E_CONTACT_FAMILY_NAME, E_CONTACT_GIVEN_NAME
and E_CONTACT_FULL_NAME

No search expression is invalid anymore.

Conflicts:

	tests/libebook/client/test-client-cursor-operations.c

x-evolution-any-field with a NULL value (or byte) means 'get all contacts'.

After much discussion on sqlite-users mailing list, I found
the real, correct technique for querying our multi attribute
values on the auxiliary tables.

  o Removed the crack query optimizer

  o Added the index to aux_table.uid (this is important
    to optimize contact inserts, as we have to delete
    all aux_table.uid entries in the case of a replace).

  o Drop the LEFT JOIN statements in favor of

    JOIN 'folder_id_email_list' AS 'email_list' ON +email_list.uid = summary.uid

    This does not create exponentially huge data sets because of the ON
    statement, the '+' ensures that email_list.uid is ineligable for
    indexing, since we want to use email_list.value indexes later on.

When finalizing a cursor, it is important to have access
to the direct backend available so that it can be asked
to delete the direct access cursor instance.

For this reason EBookClientCursor must keep the EBookClient alive,
EBookClient now exposes a hidden API (not in the header file)
to allow the EBookClientCursor to ask the EBookClient to delete
the cursor on the EBookClientCursor's behalf.

Ugly, but works.

Conflicts:

	addressbook/libebook/e-book-client.c

There is no such thing anymore as an invalid query for the cursor.

Fixed the above function to not pass NULL to e_collator_get_index()

Now we aren't filtering by fields of interest for contact changes,
so disabled the part of the test which adds a contact.

This can be reverted if we so decide to filter it manually
in EDataBookView

Conflicts:

	tests/libebook/client/test-client-uid-only-view.c

Initial changes are filtered by fields-of-interest, however
change notifications are not filtered, this is because it's actually
faster to just send the vcard as is.

Conflicts:

	tests/libebook/client/test-client-revision-view.c

  o Added test for the underscore

    Before we used to only escape the '%' character in LIKE statements,
    but SQLite also recognizes the '_' as a single char wildcard and
    needs escaping.

    This test ensures that the underscore is properly handled.

  o Added query test which ORs the same field

    This helps to test the query optimizer in EBookBackendSqlite, which
    will optimize a query which ORs two multi attribute fields (like
    E_CONTACT_EMAIL).

  o Added tests for the NOT queries

  o Added tests for NOT, AND and OR and as well as some nested AND / OR
    tests which trigger the query optimizer in EBookBackendSqlite

  o Fix expectations for phone number matches, now that the phone number
    comparisons are fixed

  o Added --filter option.

    Pass a regular expression to filter which test cases should run

  o Added test which searches by UID (can trigger bugs if normalization
    is mistakenly done on the UIDs).

Conflicts:

	tests/libebook/client/test-client-custom-summary.c

This could not be backported from master, manually done.

Conflicts:

	addressbook/libebook/e-book-client.c

Refactored the whole test suite so that it's isolated to
libedata-book, use the EBookSqlite directly.

Now for the majority of test cases, we test with and without
fields in the summary (with an empty summary configuration)
and we also test the EBookSqlite with a callback to fetch
vcards and without a callback.

We also now test for e164 number change notifications in locale changes.

Conflicts:

	tests/libebook/data/vcards/sorted-4.vcf
	tests/libebook/data/vcards/sorted-5.vcf
	tests/libedata-book/Makefile.am

Conflicts:

	addressbook/libedata-book/e-book-backend.h

This is a complete rewrite for the EBookBackendSqliteDB API,
some improvements here include:

  o Support for e_book_query_exists() and e_book_query_not()

  o ESexp parsing only happens once, least possible processing
    while generating a query

  o Query preflighting stage checks for the validity of a query,
    and also optimizes it, so that the generated statement will
    be as fast as possible

  o Implementation of queries are now much more straigh forward,
    we have a single array of constraint generator functions, one
    for each type of EBookQueryTest

  o General interactions with SQLite is improved, functions like
    ebsql_string_append_printf() and book_backend_sqlite_exec_printf()
    reduce the amount of redundant code

  o Inserts are done with prepared SQLite statements, which should
    buy us optimization on batch inserts

  o Phone number queries no longer require the E_BOOK_INDEX_PHONE
    to work, however they are quicker with the index

  o Auxiliary tables are now created for each multi attribute contact
    field in the summary separately, reducing the amount of rows
    needed to traverse for a match (also adding more flexibility to
    how we handle these multi-attribute fields)

  o Special care has been taken to ensure that we begin write transactions
    with BEGIN IMMEDIATE, allowing us to fail gracefully if and when
    SQLITE_BUSY is returned

  o Schema upgrades are more failsafe, Just add columns needed in the
    summary if they are not already declared, even though SQLite does
    not support "ADD COLUMN IF NOT EXISTS".

  o When searching contacts with a search expression that cannot
    be satisfied by the summary columns, use a fallback built into
    the SQLite instead of loading everything into memory

  o Allow usage of the cursor with fields that are not in the summary,
    both for cursor ordering and filtering.

  o While relocalizing the whole database, care is taken to perform
    the operation in batches, this way EBookBackendSqlite does not
    ever try to load the whole database in memory at once.

  o Region code for phone number parsing is now tied into the locale
    setting and will change as a result of locale changes.

  o While relocalizing and reformatting the e164 phone number attributes
    in vcards, we send notifications for vcards which changed as a
    result of relocalization via a EbSqlVCardCallback which can
    be passed to e_book_backend_sqlite_set_locale()

  o Error codes are refined to be more meaningful in general than the
    traditional E_BOOK_SQLITE_ERROR_OTHER

Conflicts:

	addressbook/libedata-book/Makefile.am
	po/POTFILES.in

Also declare the first member with "= 0", just because we now
use it to index an array, and I'm paranoid/unsure about the C spec
in this case.

Conflicts:

	addressbook/libebook-contacts/e-book-query.c
	addressbook/libebook-contacts/e-book-query.h