Investigation of GNU libunwind problems on the aarch64-linux-gnu
platform drive us to the conclusion that libunwind-1.2.1 provided by
major distribution packages is broken. Not to mention that its test
suite fails with SEGFAULTs.

Last but not least, some distributions, e.g. CentOS 8 (see #4611) do
not provide a libunwind package.

Hence, bundle libunwind: bundling is enabled by default on all
platforms, except for macOS — a system package can be used if its
version is greater or equal than 1.3.0 (minimal version that does not
seem to be broken on aarch64-linux-gnu).

* Add new submodule: bump it to current master.
* Refactor libunwind package search logic out of compiler.cmake.
* Add CMake script for building bundled libunwind.
* Add CMake script for extracting version of libunwind.
* Re-enable backtrace for all RHEL distributions by default.
* Remove libunwind from static build.

Needed for #4002
Closes #4611

NO_DOC=build system
NO_TEST=build system

It looks like we just missed the fuzzing.yml workflow when worked on
adding this feature to our CI process in #6446.

Follows up tarantool/tarantool-qa#100

NO_DOC=ci
NO_CHANGELOG=ci
NO_TEST=ci

It looks like this trigger was added beforehand for some purposes
but never used. So there is no sense to keep it in our workflows.

NO_DOC=ci
NO_CHANGELOG=ci
NO_TEST=ci

We have recently moved all development from local branches to forks and
disabled running workflows in forks. So there is no sense to keep logic
for running full testing in forks if a branch has the 'full-ci' suffix.

NO_DOC=ci
NO_CHANGELOG=ci
NO_TEST=ci

We have recently moved all development from local branches to forks.
Most workflows cannot be run in forks because many of them need access
to private runners and/or secrets.

Workflows in forks can be disabled manually, but it's hard to remember
and it needs an extra action from a developer. Instead, we can detect
the repo in workflows and run them if the repo is tarantool/tarantool.

Closes #6913

NO_DOC=ci
NO_CHANGELOG=ci
NO_TEST=ci

Information messages have been moved from the critical logging level
to info to avoid false alarms when reading logs.

NO_DOC=commit does not add any new visible functionality

NO_TEST=testing will require more resources than the result it will bring

Added support to netbox's `stream` and `future` objects using the
__autocomplete metamethod.

Closes #6305
NO_DOC=internal

This patch updates the logic in lua_rl_complete() where __index from
a metatable apparently expected to be a table, not a function. It is
rather simple for a table object, but some of them are userdata, so
an __autocomplete method is introduced to be sure it returns a table
with all names suitable for tabcompletion.

Part of #6305
NO_DOC=internal
NO_CHANGELOG=part of a fix

When the deferred DELETE optimization was introduced, it was enabled for
all Vinyl spaces. Though the optimization  should improve write
performance, at the same time it may result in significant degradation
of reads. Let's disable this optimization by default and add a space
option to enable it.

Closes #4501

@TarantoolBot document
Title: Document defer_deletes space option

The new option is a boolean flag that only makes sense for Vinyl spaces
that have secondary indexes. Setting the flag results in deferring
generation of DELETE statements for secondary indexes till compaction of
the primary index. It should speed up writes, because it eliminates a
lookup in the space for REPLACE and DELETE operations. At the same time,
it may result in degradation of reads from secondary indexes, because it
entails an additional lookup in the primary index per each "phantom"
tuple (a tuple that was deleted in the primary index, but not yet in the
secondary index, and hence should be skipped on read).

Example:

```lua
box.schema.space.create('test', {
    engine = 'vinyl',
    defer_deletes = true,
})
```

If omitted on space creation, this option is set to the value of
`box.cfg.vinyl_defer_deletes`, which is false by default.

Tarantool console quits if you type Ctrl+C.
This patch fixes console behavior on sending SIGINT.
Console discards the current input on typing Ctrl+C
and invites user to the new line.

In daemon mode the process will exit after receiving SIGINT.

Test gh-2717 should be skipped on release build, cause it
uses error injection which is enabled only on debug mode.

Fixes #2717

@TarantoolBot document
Title: Use Ctrl+C to discard the input in console

The signal SIGINT discards the input in console mode.
When tarantool executes with -e flag or runs as a daemon, SIGINT
kills the process and tarantool complains about it in log.

During DDL operations triggers of old space and new created space
are swapped. This leads to crash in case if this swap occurs from
space on_replace triggers. This patch banned all DDL operations
from on_replace triggers.

Closes #6920

@TarantoolBot document
Title: Ban DDL operations from space on_replace triggers

Previously user can set function for space on_replace trigger,
which performs DDL operation. This may leads to tarantool crash,
that's why this patch bans DDL operations from on_replace triggers.
All this operations fails with error: "Space on_replace trigger
does not support DDL operations".

This patch fixes an assertion or segmentation fault when getting the
value of MP_EXT via netbox.

Closes #6766

NO_DOC=Bugfix

netbox_perform_request throws a general ClientError exception on
time-out error: it should throw the more suitable and POSIX-compliant
TimedOut exception which sets errno to ETIMEDOUT.

Closes #6144

NO_DOC=There is no mention of what kind of error is returned on timeout
in the net.box documentation.

There was no zeroing of the number of compressed tuples in the case of
space truncation. Also there was no update of compressed tuples count
in case of using memtx mvcc engine. This patch fix these problems.

Follow up #2695

NO_CHANGELOG=bugfix
NO_DOC=bugfix
NO_TEST=test in ee version

Run `docker run` with `--rm` to remove contaners after running
a container-based build.
These containers are never reused, so there's no reason to keep them.
Instead, a new container is created for each build.
Not removing them results in wasting disk space on CI runners.

NO_DOC=ci
NO_CHANGELOG=ci
NO_TEST=ci

Part of #6773

NO_DOC=Doc-request will be added in another commit.
NO_CHANGELOG=Changelog will be added in another commit.

This patch introduces basic DATETIME support in SQL. After this patch,
it will be allowed to select DATETIME values from spaces, insert them
into spaces, and use them in functions. CAST() from DATETIME to STRING,
SCALAR and ANY is also supported.

Part of #6773

NO_DOC=Doc-request will be added in another commit.
NO_CHANGELOG=Changelog will be added in another commit.

NO_CHANGELOG=feature was not released

Closes #6823

@TarantoolBot document
Title: Document msgpack object iterator take_array method

The new method copies the given number of msgpack values starting from
the iterator cursor position to a new msgpack array object. On success
it returns the new msgpack object and advances the iterator cursor. If
there isn't enough values to decode, it raises a Lua error and leaves
the iterator cursor unchanged.

This function could be implemented in Lua like this:

```lua
function take_array(iter, count)
    local array = {}
    for _ = 1, count do
        table.insert(array, iter:take())
    end
    return msgpack.object(array)
end
```

Usage example:

```lua
local mp = msgpack.object({10, 20, 30, 40})
local it = mp:iterator()
it:decode_array_header()       -- returns 4
it:decode()                    -- returns 10
local mp2 = it:take_array(2)
mp2:decode()                   -- returns {20, 30}
it:decode()                    -- returns 40
```

Normalization was unnecessary in negative interval values
close to 0. This led to extra second in result, like:

```
tarantool> date.now() - date.now()
---
- -1.000026000 seconds
...
```

We do need to normalize when assign to datetime values, but not for
intermediate intervals.

Closes #6882

NO_DOC=bugfix

The problem is memtx_tuple_new() says address of memtx_tuple, while
memtx_tuple_delete() says address of an actual tuple (part of memtx_tuple).
Different addresses in log may be confusing so the patch fixes the problem.

Closes #6634

NO_CHANGELOG=no visible changes
NO_DOC=no visible changes

See https://github.com/tarantool/checkpatch

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

 - Remove lz4 from link libraries, because it can be linked via
   EXTRA_CORE_LINK_LIBRARIES. It'd be more correct, because the
   open-source doesn't use lz4. Follow-up commit 55f968b4 ("msgpack:
   add stubs for compressed data support").

 - Add OpenSSL include directories unconditionally, because they are
   used in the open-source build since commit e5ec324d ("Move
   OpenSSL initialization to lib/core/ssl").

NO_DOC=cmake
NO_CHANGELOG=cmake

Limbo terms tracking is shared between appliers and when
one of appliers is waiting for write to complete inside
journal_write() routine, an other may need to access read
term value to figure out if promote request is valid to
apply. Due to cooperative multitasking access to the terms
is not consistent so we need to be sure that other fibers
read up to date terms (ie written to the WAL).

For this sake we use a latching mechanism, when one fiber
takes a lock for updating other readers are waiting until
the operation is complete.

For example here is a call graph of two appliers

applier 1
---------
applier_apply_tx
  (promote term = 3
   current max term = 2)
  applier_synchro_filter_tx
  apply_synchro_row
    journal_write
      (sleeping)

at this moment another applier comes in with obsolete
data and term 2

                              applier 2
                              ---------
                              applier_apply_tx
                                (term 2)
                                applier_synchro_filter_tx
                                  txn_limbo_is_replica_outdated -> false
                                journal_write (sleep)

applier 1
---------
journal wakes up
  apply_synchro_row_cb
    set max term to 3

So the applier 2 didn't notice that term 3 is already seen
and wrote obsolete data. With locking the applier 2 will
wait until applier 1 has finished its write.

We introduce the following helpers:

1) txn_limbo_begin: which takes a lock
2) txn_limbo_commit and txn_limbo_rollback which simply release
   the lock but have different names for better semantics
3) txn_limbo_process is a general function which uses x_begin
   and x_commit helper internally
4) txn_limbo_apply to do a real job over processing the
   request, it implies that txn_limbo_begin been called

Testing such in-flight condition won't be easy so we introduce
"box.info.synchro.queue.busy" field to report if limbo is
currently latched and processing a sync request.

@TarantoolBot document
Title: synchronous replication changes

`box.info.synchro.queue` gets a new `busy` field. It is set to
`true` when there is a synchronous transaction is processing
but not yet complete. Thus any other incoming synchronous
transactions will be delayed until active one is finished.

Part-of #6036

Acked-by: Serge Petrenko <sergepetrenko@tarantool.org>
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>

To test if latch is locked.

Part-of #6036

NO_DOC=Internal helper for next patch
NO_CHANGELOG=Internal helper for next patch

Acked-by: Serge Petrenko <sergepetrenko@tarantool.org>
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>

To build EE tests, we need to add some new CMake variables:

 - Set EXTRA_TEST_SUITES_{BINARY,SOURCE}_DIR to the binary/source directory
   that contains extra test suites.
 - Set EXTRA_TEST_SUITES to the list of extra test suites to build.

NO_DOC=cmake
NO_CHANGELOG=cmake

When a request is redirected, old headers are of no use anymore and
should be dropped. We can detect a redirect by checking the value of
CURLINFO_REDIRECT_COUNT.

Note about the test: we change the redirect reply to 'redirecting...',
so that its length differs from 'hello world'.

Close #6101

NO_DOC=bug fix

NO_DOC=test
NO_CHANGELOG=test

NO_DOC=test
NO_CHANGELOG=test

In some cases, when calling a user-defined function, some of the memory
allocated earlier could be lost. This patch fixes this problem.

Closes #6789

NO_DOC=No need for doc in this case

* ci: introduce GitHub Actions

NO_DOC=LuaJIT submodule bump
NO_CHANGELOG=no visible changes

This commit add stub functions for tuple compression. Now after new
tuple allocation, we check if space format contains compression for
some fields and if it so, we compress new allocated tuple. When we
return tuple from space to user (from box_select or using C API), we
check if tuple is compressed and if it is so decompress tuple.
In opensource version user unable to set compression for tuple fields,
so all stub functions are unreachable.

Closes #2695

NO_CHANGELOG=stubs for enterprise version
NO_DOC=stubs for enterprise version

Currently `index_build` function used only for build secondary
indexes in memtx engine. Move and rename it to simplify subsequent
implementation of tuple compression.

Part of #2695

NO_CHANGELOG=refactoring
NO_DOC=refactoring

Currently `index_get` and `iterator_next` functions return
result tuple as is, without any transformation, because
tuple is stored in the space in the same format in which
it is returned to the user. Later after tuple compression
implementation (and maybe some other features) tuple should
be transformed before return it to user. Rework `index_get`
and `iterator_next` functions - now they are expected to
return the tuple in format appropriate for user. Implement
`index_get_raw` and `iterator_next_raw` functions, which
return tuple as is.

Part of #2695

NO_CHANGELOG=internal changes
NO_DOC=internal changes

When implementing tuple compression, we need to change memtx
indexes vtab dynamically, when we change space format. For
this purpose we implement helpful functions for memtx index
vtab selection.

Part of #2695

NO_CHANGELOG=refactoring
NO_DOC=refactoring

When implementing tuple compression, it is planned
to use templates to select index vtab. Templates
enabled only in c++, so we convert all associated
files from *.c to *.cc.

Part of #2695

NO_CHANGELOG=refactoring
NO_DOC=refactoring

Currently memtx use stmt->old_tuple and stmt->new_tuple
for rollback, but later when tuple compression will be
implemented we need special place to store tuples, which
was inserted or deleted during transaction, because
stmt->old_tuple and stmt->new_tuple are used in on_replace
triggers and should be decompressed.

Part of #2695

NO_CHANGELOG=internal changes
NO_DOC=internal changes

The purpose of this commit is code refactoring, such that all
the work which is common to insert/replace/delete/update/upsert
functions in memtx space is done in one place.

Part of #2695

NO_CHANGELOG=refactoring
NO_DOC=refactoring

Implement ability to set compression for tuple fields. Compression type
for tuple fields is set in the space format, and can be set during space
creation or during setting of a new space format.
```lua
format = {{name = 'x', type = 'unsigned', compression = 'none'}}
space = box.schema.space.create('memtx_space', {format = format})
space:drop()
space = box.schema.space.create('memtx_space')
space:format(format)
```
For opensource build only one compression type ('none') is
supported. This type of compression means its absence, so
it doesn't affect something.

Part of #2695

NO_CHANGELOG=stubs for enterprise version
NO_DOC=stubs for enterprise version

Library `tuple` depends on the `coll` library, since `key_def.c` file
from the `tuple` library use functions from the `coll` library. This
usually does not cause problems, since the `coll` library links with
tarantool and resolves symbols in `tuple` library. For tests it is not
a problem, because functions from the `tuple` library, which depends
from the `coll` library are `static inline` and they they are not used
in tests. But adding not `static inline` function in `tuple` library
which depends on `coll` library leads to unresolved symbols in tests.
Fix this explicitly adding `coll` library as a target link library
for `tuple` library. Also remove `${ICU_LIBRARIES}` from target link
libraries of `tuple` library because `tuple` library use ICU only via
`coll` library.

NO_CHANGELOG=cmake
NO_DOC=cmake

This commit add stub functions for compressed data
support. Add new MP_EXT type MP_COMPRESSION. Despite
the fact that this type is added to open source version
all functions for working with it are implemented only
in enterprise version of tarantool. In open source version
add stubs for appropriate functions.

Part of #2695

NO_CHANGELOG=stubs for enterprise version
NO_DOC=stubs for enterprise version