This patch introduces explicit cast from MAP to INTERVAL.

Follow-up #6773

NO_DOC=INTERVAL has already been introduced.
NO_CHANGELOG=INTERVAL has already been introduced.

This patch introduces explicit cast from MAP to DATETIME.

Follow-up #6773

NO_DOC=DATETIME has already been introduced.
NO_CHANGELOG=DATETIME has already been introduced.

This patch introduces explicit cast from STRING to DATETIME.

Follow-up #6773

NO_DOC=DATETIME has already been introduced.
NO_CHANGELOG=DATETIME has already been introduced.

When applier ack writer was moved to applier thread, it was overlooked
that it would start sharing replicaset.vclock between two threads.

This could lead to the following replication errors on master:

 relay//102/main:reader V> Got a corrupted row:
 relay//102/main:reader V> 00000000: 81 00 00 81 26 81 01 09 02 01

Such a row has an incorrectly-encoded vclock: `81 01 09 02 01`.
When writer fiber encoded the vclock length (`81`), there was only one
vclock component: {1: 9}, but at the moment of iterating over the
components, another WAL write was reported to TX thread, which bumped
the second vclock component {1: 9, 2: 1}.

Let's fix the race by delivering a copy of current replicaset vclock to
the applier thread.

Also add a perf test to the perf/ directory.

Closes #7089
Part-of tarantool/tarantool-qa#166

NO_DOC=internal fix
NO_TEST=hard to test

We don't rebuild the tuples stored in the space when a space format is
updated, just check that they are compatible with the new format. For
the old tuple fields to be accessible by the new format field names, we
point the old format dictionary to the new format dictionary (a field
dictionary is a reference counted object so the same dictionary can be
used by multiple formats).

This isn't necessary for online space upgrade, because in this case
we upgrade each tuple stored in the space to the new format before
returning it to the user. Moreover, using the new dictionary for the old
tuples would even be harmful in this case, because the old tuples may be
incompatible with the new format, while the space upgrae function may
use the old field names.

Needed for https://github.com/tarantool/tarantool-ee/issues/112

NO_DOC=ee
NO_TEST=ee
NO_CHANGELOG=ee

iproto_write_error() may reset diag (for example, if the client closes
the socket), thus invalidating the error we are going to log and leading
to a use after free bug. We must log the error before trying to send it
to the client.

The bug was introduced by commit 9b4ab9fe ("iproto: use iostream
abstraction"), which switched iproto_write_error() from plain write() to
the iostream_write(), which may set diag.

Closes #6890

NO_DOC=bug fix
NO_CHANGELOG=unreleased

Add the ubuntu_20_10.yml and ubuntu_20_10_aarch64.yml workflow files to
build Tarantool packages for x86_64 and aarch64 systems.

Actually, it was already added in 980d7676 (2.9.0-100-g980d7676b) in the
scope of #5824 and removed then in 19a161b5 (2.10.0-beta2-4-g19a161b5f)
and dc19be40 (2.10.0-beta2-5-gdc19be406).

But, eventually, it was decided to add this workflow again to release
Tarantool packages for Ubuntu Groovy as well.

Follows-up tarantool/tarantool#5824

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

Add the ubuntu_21_04_aarch64.yml and ubuntu_21_10_aarch64.yml workflow
files to build Tarantool packages for the aarch64 system.

Follows-up tarantool/tarantool#5825
Follows-up tarantool/tarantool#6566

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

Add the ubuntu_22_04.yml and ubuntu_22_04_aarch64.yml workflow files to
build Tarantool packages for x86_64 and aarch64 systems.

Closes tarantool/tarantool-qa#237

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

Co-authored-by: Yaroslav Lobankov <y.lobankov@tarantool.org>

Consider the following case a safe `select` call (i.e., do not issue a
warning):
* offset + limit <= 1000 (default value of offset is 0, default value of
limit is 4294967295);

Add new dangerous `select` call case:
* 'ALL', 'GE', 'GT', 'LE', 'LT' iterator even with key present.

Because of the additional logic, it was decided to refactor
`check_select_args` and call it after resolving options in
`check_select_opts`.

Closes #7129

@TarantoolBot document
Title: improve check for dangerous `select` calls

Calls with `offset + limit <= 100`
are now considered safe (i.e., a warning is not issued), e.g.:
box.space.s:select(nil, {offset = 1000})
box.space.s:select(nil, {limit = 1000})
box.space.s:select(nil, {offset = 500, limit = 500})

'ALL', 'GE', 'GT', 'LE', 'LT' iterators are now considered dangerous by
default even with key present, e.g.:
box.space.s:select({0})
box.space.s:select({0}, {iterator = 'GE'})

But not that these calls are still safe:
box.space.s:select({0}, {limit = 1000})
box.space.s:select({0}, {limit = 1000, iterator = 'GE'})
box.space.s:select({0}, {iterator = 'EQ'})

Constructor date.new() and modifier date:set() should always produce same
result for all attributes combinations. Fixed the problem for
`timestamp` with `tzoffset`.

Fixes #6793

@TarantoolBot document
Title: datetime :set{} with tzoffset

Constructor `date.new()` and modifier `date:set()` should always produce
same result for all attributes combinations. Fixed the problem for
`timestamp` with `tzoffset`.

```
tarantool> date.new{tzoffset = '+0800', timestamp = 1630359071}
---
- 2021-08-30T21:31:11+0800
...

tarantool> date.new():set{tzoffset = '+0800', timestamp = 1630359071}
---
- 2021-08-30T21:31:11+0800
...
```

The unauthenticated git protocol on port 9418 is no longer supported.

NO_DOC=doc
NO_TEST=doc
NO_CHANGELOG=doc

This patch introduces a new syntax that allows to create MAP values in
an SQL query.

Part of #4763

@TarantoolBot document
Title: Syntax for MAP in SQL

The syntax for creating document values is now available in SQL. You can
use `{`, `:` and `}` to create a MAP value. Only INTEGER, STRING, and
UUID values can be keys in a MAP value. If there are two or more values
for the same key, the last one will be used for that key.

Examples:
```
tarantool> box.execute("SELECT {1 : 'a', 'asd' : 1.5, uuid() : true};")
---
- metadata:
  - name: COLUMN_1
    type: map
  rows:
  - [{1: 'a', 91ca4dbb-c6d4-4468-b4a4-ab1e409dd87e: true, 'asd': 1.5}]
...
```

```
tarantool> box.execute("SELECT {'h' : ['abc', 321], 7 : {'b' : 1.5}};")
---
- metadata:
  - name: COLUMN_1
    type: map
  rows:
  - [{7: {'b': 1.5}, 'h': ['abc', 321]}]
...
```

After this patch, variable names will have to follow the rules defined
for identifiers in SQL. Essentially, this means that a digit can no
longer be used as the first character of a bind variable name.

Part of #4763

NO_DOC=Will be added later.
NO_CHANGELOG=Will be added later.

The `${{ github.ref_name }}` expression is always evaluated to 'master'
because the workflow is running on the 'master' branch. But it is wrong
in the case of the '2.10' branch. Now it's fixed.

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

The '2.8' branch is no longer supported. So let's switch to '2.10'.

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

The perf_sysbench.yml workflow has the 'workflow_dispatch' trigger
that is often used to test some workflow changes. To not litter in
the InfluxDB prod bucket, MyTeam chat and mailing list, the special
option is added (Debug mode). When checking this option the InfluxDB
test bucket, MyTeam chat, and mailing list will be used.

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

Since analyzing the performance test results via the workflow artifacts
is quite inconvenient (we need to download the artifact, unpack it, look
through results), we need a more human approach to see the test report.
Moreover, it would be cool to see the performance difference between the
commits.

So this patch adds reporting the performance test results to MyTeam chat
and mailing list. Furthermore, the comparison of test results is added
to the report. Seeing the test report in the chat or mail is very simple
and convenient, plus forever results history as a bonus.

The test report looks something like this:

    # Curr:
    #   branch:  add-feature-y
    #   build:   1.2.3-42-g6c71c3901
    #   summary: Add feature Y
    #   machine: x86_64
    #   distrib: ce
    #   gc64:    false
    # Prev:
    #   branch:  add-feature-x
    #   build:   1.2.3-41-g5271240ea
    #   summary: Add feature X
    #   machine: x86_64
    #   distrib: ce
    #   gc64:    false
    +----------------------+----------+----------+----------+
    |Sysbench              | Curr(rps)| Prev(rps)|     Ratio|
    +----------------------+----------+----------+----------+
    |oltp_delete           | 11408.000| 11532.000|     0.989|
    |oltp_insert           | 31455.000| 31322.000|     1.004|
    |oltp_point_select     | 65031.000| 66486.000|     0.978|
    |oltp_read_only        |  1732.000|  1735.000|     0.998|
    |oltp_read_write       |  1211.000|  1196.000|     1.013|
    |oltp_update_index     | 15580.000| 15570.000|     1.001|
    |oltp_update_non_index | 16129.000| 16093.000|     1.002|
    |oltp_write_only       |  4482.000|  4434.000|     1.011|
    |select_random_points  | 18525.000| 18517.000|     1.000|
    |select_random_ranges  |  4364.000|  4331.000|     1.008|
    +----------------------+----------+----------+----------+
    |gmean                 |  9223.593|  9220.182|     1.000|
    +----------------------+----------+----------+----------+

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

Since recently we partially support timezone names (i.e. as
abbreviations) so we may modify tz attribute support for
datetime constructors or :set() operations.

Closes #7076
Relates to #7007

@TarantoolBot document
Title: datetime tz attribute

Now `tz` attribute is properly handled in datetime value
constructors or `:set{}` method modifiers.

```
tarantool> T = date.new{year = 1980, tz = 'MSK'}
---
...

tarantool> T.tzoffset
---
- 180
...

tarantool> T.tz
---
- MSK
...
tarantool> T = date.new{year = 1980, tzoffset = 180}
---
...

tarantool> T.tzindex
---
- 0
...

tarantool> T.tz
---
-
...

tarantool> T.tzoffset
---
- 180
...

tarantool> T:set{tz = 'MSK'}
---
...

tarantool> T.tz
---
- MSK
...

```

It looks like we have some files in the source tree that are not used
anymore. So there is no sense to keep them.
Files to be deleted: .appveyor.yml, Jenkinsfile, snapcraft.yaml.

NO_DOC=files removal
NO_TEST=files removal
NO_CHANGELOG=files removal

It doesn't work, because fio.rmtree doesn't accept wildcards.
Let's delete the whole server working tree on cleanup.

NO_DOC=test
NO_CHANGELOG=test

Similar to RLIST_HEAD_INITIALIZER and RLIST_HEAD. Needed to initialize
a global fiber_cond variable without calling fiber_cond_create.

NO_DOC=internal
NO_TEST=internal
NO_CHANGELOG=internal

 - Add func_id_by_name to box.internal. It will be used in EE.
 - Don't call box.internal.space.upgrade from box.schema.space.upgrade -
   we will redefine box.schema.space.upgrade in EE instead.

Follow-up commit 38b25832 ("box: add space upgrade stubs").

NO_DOC=internal
NO_TEST=internal
NO_CHANGELOG=internal

This change contains the following improvements:

1. Prettify view in the job list.

   Before:
     - centos_7 / centos_7 (GC64=OFF)
     - centos_7 / centos_7 (GC64=ON)

   After:
     - centos_7 / centos_7
     - centos_7 / centos_7 (gc64)

2. Change possible values for the ${GC64} env variable to
   `true/false` instead of `ON/OFF` which is more traditional.

3. Use `-DLUAJIT_ENABLE_GC64=ON` flag in the `debian/rules` and
   `rpm/tarantool.spec` files directly instead of providing it
   in the workflow files.

Follows-up tarantool/tarantool-qa#159
Follows-up tarantool/tarantool-qa#161

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

 - Cleanup the unreleased directory: leave only fix-string-dequoting.md
   there, which was added after last branching (2.10).
 - Remove 2.10.0-beta1.md, 2.10.0-beta2.md, 2.8.1.md, because they are
   available at the corresponding branches.

NO_DOC=changelogs
NO_TEST=changelogs

Sometimes we would like to run only integration testing and not touch
other tests. Now it can be done by setting the 'integration-ci' label
on a pull request.

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

main.cc includes some headers from Tarantool Enterpise (to be more precise -
flightrec.h). To build it we should include proper extra directories.

NO_DOC=<Build fix>
NO_CHANGELOG=<Build fix>
NO_TEST=<Build fix>

ffi.typeof() returns cdata, which in turn can't be concatenated with string
without explicit call of `tostring`. Moreover, passing data of native Lua
type would also generate wrong error. So let's simply raise general error
without mentioning passed type.

NO_DOC=<Internal fix>
NO_CHANGELOG=<Internal fix>

Previously we decided to pass `msg->p_inbuf->rpos` to flight recorder in
order to dump request. However, this is wrong way to do it since `rpos`
is moved only request has been processed, i.e. at the same moment several
requests may have the same `rpos` value. To fix this let's save the start
of unparsed request to `iproto_msg`. There's one detail that should be
clarified to understand that in this case we'll always get valid
pointer to ibuf. Imagine following state of ibuf:
```
+_____________________
|   |  |          |
+---------------------
    ^  ^          ^
  RPOS R2       WPOS
   R1
```
R1 is the first request in the buffer (i.e. rpos points to it);
R2 is the second. One can argue that if R2 is processed faster than R1
than during the dump of R1 its `reqstart` may point to the garbage.
However, dump of request takes place at the bery beginning of request
execution in TX thread. As far as messages are started to be processed
exactly in the same way as they were received, then R1 will be always
dumped before R2 and its further processing.

Follow-up 247515e9

NO_DOC=<No user visible changes>
NO_TEST=<No functional changes>
NO_CHANGELOG=<No functional changes>

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

It looks like the simplest solution to deliver gc64 tarantool builds
is to have a separate repo for it. So this patch removes old stuff
related to the 'tarantool-gc64' package and adds the new logic to
store gc64 packages in a separate repo.

Closes: tarantool/tarantool-qa#161

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

Co-authored-by: Yaroslav Lobankov <y.lobankov@tarantool.org>

space_upgrade_check_alter() must be called after the new space
definition is updated.

Follow-up commit 38b25832 ("box: add space upgrade stubs").

NO_DOC=internal
NO_TEST=internal
NO_CHANGELOG=internal

Box read operations (select, min, max, random, pairs) may call a Lua
function via Lua C API if space upgrade (EE feature) is in progress.
This is forbidden if the original function was called via FFI:

https://github.com/tarantool/luajit/commit/4f4fd9ebeb10d53c8e67f45170938f052818e64e

Let's add helper functions to disable FFI for box read operations.
We will use them in EE code.

NO_DOC=internal
NO_TEST=internal
NO_CHANGELOG=internal

In Lua C code 'struct iterator&' and 'struct iterator*' are handled in
exactly the same way. However, if we create an iterator object in Lua by
calling box_index_iterator via FFI, and pass the object to Lua C, it
will have 'struct iterator*' type, not 'struct iterator&'. In the next
patch we need Lua C functions to work with such objects so let's replace
'struct iterator&' with 'struct iterator*' in C code.

NO_DOC=refactoring
NO_TEST=refactoring
NO_CHANGELOG=refactoring

clang 13 includes a check for subtraction from NULL pointer which
is considered UB: historically, we had an alignment checking macro
which is affected by this. It seems like the intention of the
macro's author was to implicitly cast the pointer being checked to
uintptr_t without including stddef.h — replace this subtraction
with an explicit cast.

There is no way to set SQL_4_BYTE_ALIGNED_MALLOC, so the corresponding
part of the code was removed. Now there are only 8 byte alignment
assertions.

NO_CHANGELOG=UB fix
NO_DOC=UB fix
NO_TEST=UB fix

Co-authored-by: Georgiy Lebedev <curiousgeorgiy@gmail.com>

Previously,

select "t1"."a" from (select "a" from "t") as "t1";

returned a result column name `t1` instead of `t1.a` because of
incorrect work of a dequoting function. The reason was that
previously sqlDequote() function finished its work when found the
first closing quote.

Old logic worked for simple selects where the column name doesn't
contain an explicit scan name ("a" -> a).
But for the sub-queries results sqlDequote() finished its work right
after the scan name ("t1"."a" -> t1). Now the function continues its
deqouting till it gets the null terminator at the end of the string.

Closes #7063

NO_DOC=don't change any public API, only a bug fix

Co-authored-by: Mergen Imeev <imeevma@gmail.com>

It will be used to stall background space upgrade in tests.

While we are at it, move ERRINJ_TX_DELAY_PRIO_ENDPOINT to restore
the sorted order of the error injection list.

Follow-up commit 38b25832 ("box: add space upgrade stubs").

NO_DOC=internal
NO_TEST=internal
NO_CHANGELOG=internal

We need the new format to apply the upgrade function to a tuple.

Follow-up commit 38b25832 ("box: add space upgrade stubs").

NO_DOC=internal
NO_TEST=internal
NO_CHANGELOG=internal

Follow-up commit 38b25832 ("box: add space upgrade stubs").

NO_DOC=test
NO_CHANGELOG=test