The assert() in tt_pthraed_error()
would always fail, since it was checking for e == 0, and
e could be equal to ETIMEDWAIT.

Produce a more verbose error message if failed to restart
from a snapshot.

Failure recovery: don't confirm rolled back LSNs.
Transactions are rolled back in the order, reverse to the initiation
order. Thus we first confirm the greatest LSN, then the one
preceding it, and so on. If we confirm LSNs of these
transactions, it can be done only out of order.

At better diagnostics to the failure of partial write. Rewrite
save_snapshot() to produce more diagnostics.

Truncate the log file in case of a partial write:
simple lseek() to the last good offset is not enough,
since, if we choose to close the file, we may still have
garbage at the end of the file.

Fix a coding bug in the patch which suppresses unnecessary warnings.
Add a comment.

When there was a partial write, rollback would
first confirm LSNs of requests which are only trying
to enter the queue, and only then LSNs of committed
or rolled back events.

Always put requests into the queue, even if we know
that they will be rolled back. This guarantees
a correct LSN order in confirm_lsn().

Split the rollback queue around the first not written
request, not around the first request in the batch.
In case of a partial write we may have written
a few requests at the head of the batch.

A fix and a test case for Bug#1043858
"server crash on lua stack overflow on CentOS 5.4"

Workaround a bug in libobjc2 which would mistakenly
try to catch a foreign exception in @catch (id e) frame,
when there is a @finally down the stack.

A test case is added to the test suite.

A refined test case is sent to the libobjc2 maintainer.

console client log flush prior to close.

Make sure we don't crash on NULL argument of error().

The wal writer thread could write a wrong fiber
name into the log, when logging about WAL rotation.
vsay() takes the current fiber name from a global
variable, and it's changed dynamically by the concurrent
thread (the front end thread).

Conflicts:
	test/box/lua.result

Fix coding style, add comments, remove
an unused variable (int raise).

A fix and a test case for Bug#1042738
"Tree iterator over entire range gives a spurious error"

We didn't look at part_count in num32_key_node_cmp,
and thus could produce wrong comparison result if
part_count was 0.

To be on the safe side, zero key parts even if
part count is 0 (this is what we do elsewhere as well).

Cleanup: make sure you can run box_big/lua.test many times.

Fix a memory leak in Lua stored procedures introduced
when cherry-picking changes from the old net-split
tree.

Make a palloc savepoint *before* we allocate txn
and port, to make sure txn and port memory is freed
when rolling back to the palloc savepoint.
Otherwise a long-running Lua procedure (e.g. background
fiber) can keep allocating palloc memory indefinitely.