Includes fixes for:
mysql/mysqli: Fix ability for MetaTables to filter by table name,
broken since 5.15.
See http://phplens.com/lens/lensforum/msgs.php?id=19359

postgres8: New class/type with correct behavior for _insertid().

postgres9: Fixed assoc problem.
See http://phplens.com/lens/lensforum/msgs.php?id=19296



Change-Id: If54f95815015ab9b7f58e102fb6c65cd086b8ac2
Signed-off-by: Robert Lyon <robertl@catalyst.net.nz>

Bug 1238424
Signed-off-by: Aaron Wells <aaronw@catalyst.net.nz>
Change-Id: Id87addb9cbd42a7b56757d49788331a578ca4e4c

Upgrade our ADODB driver from the long-obsolete mysql driver to the actively developed mysqli driver

Bug 1187964: The biggest improvement is that the ADODB mysqli driver actually supports transactions!

Change-Id: I8e55ea28cd6e6b8ed5c883646c1172d318af5554

On the principle that it's good secur1ty to remove executables you're not likely
to use, clearing out all the bits of ADOdb we're currently not using. This includes
drivers and data dictionaries for other DBs, tests, docs, icons, PHP4 session code,
pear auth container, and non-English error messages.

Change-Id: Ibf059ed1d0d7dafbbce16589ead2c0a006e3fcdd
Signed-off-by: Aaron Wells <aaronw@catalyst.net.nz>

Bug #1180625

ADOdb 5.18, downloaded form http://sourceforge.net/projects/adodb/files/adodb-php5-only/adodb-518-for-php5/
I've cherry-picked the following Mahara commits onto it:

1. adodb: fix MySQL bug on installing Mahara
This was reported in the forums by Bruce Ellis:
  http://mahara.org/interaction/forum/topic.php?id=131
and corresponds to Eduforge #3159 and LP #547934.
(cherry picked from commit e7029e5f)

2. adodb: RecordSet_postgres7->MoveNext() - remove expensive is_array()
 implicit count() of the array keys/columns. This shaves a good 5%
of the exec time of get_records('config')
(cherry picked from commit 04ef8a3b)

3. adodb: check for Postgres 8.1
See Eduforge #2318 or LP #547951.
(cherry picked from commit c6dfccad)

4. Check for valid columns in adodb postgres64 driver MetaIndexes fn (bug #796102)
Indexes on functions of columns have a 0 in the indkey column of
pg_catalog.pg_index, which doesn't refer to a column.
Don't return any columns for the index, rather than returning an empty
column name.
(cherry picked from commit b87d9961)

5. Bug #841711: ADODB modification: add __wakeup to connection object
This reconnects to the db after wakeup, as resource types cannot be
serialized and unserialized.
This is necessary for unit tests, phpunit backups and restore globals
between each test.
(cherry picked from commit 41707745

)

Change-Id: I888bfc0cfb6acfc58adf1b416580c4b5ddd70732
Signed-off-by: Aaron Wells <aaronw@catalyst.net.nz>

This reconnects to the db after wakeup, as resource types cannot be
serialized and unserialized.

This is necessary for unit tests, phpunit backups and restore globals
between each test.

Change-Id: I160b4c68ac1db0c3c0a3e1aec1b0dc8584763bf9
Signed-off-by: Penny Leach <penny@mjollnir.org>
Signed-off-by: Andrew Robert Nicols <andrew.nicols@luns.net.uk>

Indexes on functions of columns have a 0 in the indkey column of
pg_catalog.pg_index, which doesn't refer to a column.

Don't return any columns for the index, rather than returning an empty
column name.

Change-Id: I1464b8a9fda8bd137dcabf1145da289ccf32c500
Signed-off-by: Richard Mansfield <richard.mansfield@catalyst.net.nz>

This is what was run:

  find . -iname '*.png' -exec optipng -o7 -q {} \;
  find . -iname '*.gif' -exec gifsicle -O2 -b {} \;
  find . -iname '*.jpg' -exec jpegoptim -p --strip-all {} \;
  find . -iname '*.jpeg' -exec jpegoptim -p --strip-all {} \;

Change-Id: I6a516782e673167f04ef0945a04a417514ad5cf2
Signed-off-by: Francois Marier <francois@catalyst.net.nz>

We don't use this and the directory is full of executable PHP files.
Signed-off-by: Francois Marier <francois@catalyst.net.nz>

Signed-off-by: Francois Marier <francois@catalyst.net.nz>

See Eduforge #2318 or LP #547951.

(cherry picked from commit 4bdafccb

)
Signed-off-by: Francois Marier <francois@catalyst.net.nz>

And implicit count() of the array keys/columns. This shaves a good 5%
of the exec time of get_records('config')

(cherry picked from commit 4652e561

)
Signed-off-by: Francois Marier <francois@catalyst.net.nz>

This was reported in the forums by Bruce Ellis:

  http://mahara.org/interaction/forum/topic.php?id=131

and corresponds to Eduforge #3159 and LP #547934.

(cherry picked from commit d2b9a10c

)
Signed-off-by: Francois Marier <francois@catalyst.net.nz>

The only security issue fixed since 4.92 has to do with those sites
where magic_quotes_sybase is turned ON, so it's not an issue for us.

This will most likely help with LP #572189.
Signed-off-by: Francois Marier <francois@catalyst.net.nz>

Signed-off-by: Francois Marier <francois@catalyst.net.nz>

Signed-off-by: Francois Marier <francois@catalyst.net.nz>

http://mahara.org/interaction/forum/topic.php?id=131
(cherry picked from commit 4c2a22677c3ae886f0a0f76aea15a0a15eb02eb4)

And implicit count() of the array keys/columns. This shaves a good 5%
of the exec time of get_records('config')