Skip unstable tests of `replication-luatest/linearizable_test.lua` due
to tarantool-qa#277.

NO_DOC=testing stuff
NO_TEST=testing stuff
NO_CHANGELOG=testing stuff

The fixed tests often failed with

  fail | 2023-03-01 15:54:30.550 [3724975] main/103/server_instance.lua
  F> can't initialize storage: unlink, called on fd 63,
  aka unix/:(socket), peer of unix/:(socket): Address already in use

We fixed a similar issue in commit 3d3e9dea ("test: fix flaky
box-luatest/gh_7917_log_row_on_recovery_error_test") by using unique
instance names. Let's do the same here.

NO_DOC=testing stuff
NO_TEST=testing stuff
NO_CHANGELOG=testing stuff

Server API from the luatest has been changed: `server:clean()` method
has been removed. Use `server:drop()` instead.

Follows up tarantool/luatest#296

NO_DOC=test fix
NO_TEST=test fix
NO_CHANGELOG=test fix

Bump test-run to new version with the following improvements:

- Bump luatest to 0.5.7-33-g8523e5c [1]

[1] tarantool/test-run@7db594d

Part of tarantool/luatest#296

NO_DOC=testing stuff
NO_TEST=testing stuff
NO_CHANGELOG=testing stuff

Tuples that have been inserted into system spaces after the _space
definition has been inserted into _space have the same format as the
space into which they were inserted. However, previously inserted tuples
have an incomplete tuple format with parts missing. One piece that is
missing information is the JSON token, which is used to determine the
number of fields.

The tarantoolsqlIdxKeyCompare() function contains the correct checks for
the case when the number of fields from the format is equal to or less
than the fieldno field, but uses tuple_format_field() before this check,
resulting in an assertion.

This patch forces tarantoolsqlIdxKeyCompare() to call
tuple_format_field() only after checking that fieldno is less than
field_count.

Closes #8418

NO_DOC=Bugfix in debug.

This may result in invalid statistics if there's more than 4 GB
allocated for index extents.

Fixes commit a75f4b7e ("memtx: introduce read view statistics").
Follow-up #8501
Coverity report 1537026, 1537027.

NO_DOC=bug fix
NO_CHANGELOG=unreleased
NO_TEST=complicated; checked by coverity

Allow to register tweaks for variables of type double. Such tweak can be
set to any integer or double value.

Follow-up #7883

NO_DOC=internal
NO_TEST=internal
NO_CHANGELOG=internal

To make changelog preparation less stressful let's pass each and every
change to changelogs/ directly through the doc team.

NO_CHANGELOG=no code changes
NO_TEST=no code changes
NO_DOC=no code changes

Fix grammar, punctuation, and wording

NO_CHANGELOG=changelog
NO_DOC=changelog
NO_TEST=changelog

* Add an extra step that should be considered as a workaround for the
  runner agent (ChristopherHX/github-act-runner) that is used on FreeBSD
  machines. Without it, env.ARTIFACT_EXISTS will be undefined for some
  reason.

* In the s3-upload-artifact action change `if: env.ARTIFACT_EXISTS` to
  `if: env.ARTIFACT_EXISTS == 'true'` to be compatible with the FreeBSD
  runner agent.

Follows up tarantool/multivac#116

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

Add a new possible value for `bootstrap_strategy` configuration option -
"config". When bootstrap_strategy is "config", the user may use
`bootstrap_leader` configuration option to specify the URI or UUID of
the desired bootstrap leader.

Closes #7999

@TarantoolBot document
Title: new configuration option - `bootstrap_leader`.

Configuration receives a new parameter - `bootstrap_leader`. The
parameter specifies the desired node to bootstrap from.

The parameter is valid only when another option - `bootstrap_strategy` -
is set to `'config'`.
The user may pass either the bootstrap leader's URI or its UUID string.

If bootstrap leader is passed as a UUID, the node will bootstrap from
the peer whose UUID matches the given one.

If bootstrap leader is passed as a URI, the URI must contain the same
host:port pair (or unix socket path) used for this node in
`box.cfg.replication`. In this case the node will bootstrap from the
remote peer listed in the corresponding replication entry.

For example, this call will return an error:
```lua
box.cfg{
    bootstrap_strategy = 'config',
    bootstrap_leader = 'localhost:3301',
    replication = {'127.0.0.1:3301'},
}
```
But this call will succeed:
```lua
box.cfg{
    bootstrap_strategy = 'config',
    bootstrap_leader = 127.0.0.1:3301',
    replication = {'127.0.0.1:3301'},
}
```
In the example above, setting either or both of `bootstrap_leader` and
`replication` to `"user:password@127.0.0.1:3301` will work just fine,
i.e. only the bootstrap leader's host and port are matched against
replication entries.

If `bootstrap_leader` points at the instance being configured, the same
URI entry for the instance must still be present in `replication`
configuration parameter, like this:
```lua
box.cfg{
    listen = 'localhost:1234',
    bootstrap_leader = 'localhost:1234',
    replication = {'localhost:1234', ...},
}
```
The only exception to this is setting `bootstrap_leader` to the same
UUID as instance_uuid, like this:
```lua
box.cfg{
    listen = 'something',
    bootstrap_leader = '11111111-1111-1111-1111-111111111111',
    instance_uuid = '11111111-1111-1111-1111-111111111111',
```
In this case the node will bootstrap the cluster on its own.

The test will include test cases for other possible bootstrap
strategies, so rename it properly. Doing this in a separate commit
helps to minimize the diff.

In-scope-of #7999

NO_DOC=test rename
NO_TEST=test rename
NO_CHANGELOG=test rename

replicaset_connect() sleeps in a loop while waiting for the connections
to be established and checks a bunch of conditions in between the loop
cycles to check if it can proceed (with either a success or a failure).

The conditions are already quite complicated and multiple, but it's
going to get worse with the addition of bootstrap_strategy "config" and
"supervised": during bootstrap replicaset_connect() will proceed as soon
as the configured bootstrap leader is connected.

Factor all these checks out to a helper function for simplicity.

Part-of #7999

NO_DOC=refactoring
NO_TEST=refactoring
NO_CHANGELOG=refactoring

Add uri_is_nil() and uri_addr_is_equal() helpers.

In-scope-of #7999

NO_DOC=not user-visible
NO_TEST=tested implicitly in the next commit
NO_CHANGELOG=not user-visible

Some time ago we decided that new checkers added to box_check_config()
shouldn't throw. Instead they return -1 and box_check_config() raises an
error.

box_check_uuid() will be used in one of the upcoming checkers, so make
it return -1 and set diag in case of error.

Since it's used by box_check_instance_uuid() and
box_check_replicaset_uuid(), fix them as well.

In-scope-of #7999

NO_DOC=refactoring
NO_TEST=refactoring
NO_CHANGELOG=refactoring

Checking a single uri will be needed in upcoming commits.

In-scope-of #7999

NO_DOC=refactoring
NO_TEST=refactoring
NO_CHANGELOG=refactoring

We manage using sanitizers with CMake flags: ENABLE_ASAN and
ENABLE_UB_SANITIZER. sql_fuzzer enables Address Sanitizer by passing
-fsanitize=address directly to CFLAGS. It works fine on OSS Fuzz, but
breaks building fuzzers for Sydr [1], because Sydr requires building
with disabled sanitizers, see documentation [2].

1. https://github.com/ispras/oss-sydr-fuzz/pull/164
2. https://sydr-fuzz.github.io/docs/#%D0%BE%D0%B1%D0%B5%D1%80%D1%82%D0%BA%D0%B0-sydr

Fixes #8529

NO_CHANGELOG=<fuzzing testing>
NO_DOC=<fuzzing testing>
NO_TEST=<fuzzing testing>

NO_DOC=ci
NO_TEST=ci
NO_CHANGELOG=ci

This patch adds the step to all workflows with test-run tests to save
test artifacts to S3 for providing artifact link from Multivac Grafana
dashboard.

Resolves tarantool/multivac#116

NO_DOC=CI
NO_TEST=CI
NO_CHANGELOG=CI

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

This patch adds a local GitHub action to save test artifacts to S3.
Print a warning if the provided artifact path doesn't exist.

Part of tarantool/multivac#116

NO_DOC=CI
NO_TEST=CI
NO_CHANGELOG=CI

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

Patch adds a grammar-based SQL fuzzer that based on libprotobuf-mutator
[1] and libFuzzer [2]. Tarantool's SQL grammar [3][4] is described in
Protobuf format in a file sql_query.proto, fuzzing engine generates SQL
query in Protobuf format and then serializes it to a string
(std::string) in sql_query_proto_to_string.* files. Resulted string is
passed to SQL engine in Tarantool.

NOTE: Current version of SQL grammar and serializers implements only
`CREATE TABLE` and `SELECT` queries with all supported options.

1. https://github.com/google/libprotobuf-mutator
2. https://llvm.org/docs/LibFuzzer.html
3. https://www.tarantool.io/en/doc/latest/reference/reference_sql/sql_statements_and_clauses/
4. https://www.tarantool.io/en/doc/latest/reference/reference_sql/sql_user_guide/

Part of #4826

NO_DOC=<fuzzing testing of SQL>

Seed corpus was generated by fuzzing test itself.

Needed for #4826

NO_CHANGELOG=<fuzzing testing of SQL>
NO_DOC=<fuzzing testing of SQL>
NO_TEST=<fuzzing testing of SQL>

SQL queries generated by fuzzing test could trigger a code that want
access to a space cache. Usually this cause an abort, because fuzzing
test doesn't perform a complete initialization of tarantool instance. To
avoid this we isolated problem parts of code with macro
FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION. More appropriate solution will
be implemented later.

With introduced macro FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION we change
behaviour of SQL engine and this breaks SQL regression tests. Therefore
FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION is disabled in
test-release-asan.

Needed for #4826

NO_CHANGELOG=internal
NO_DOC=internal
NO_TEST=internal

SQL fuzzing test requires an entrypoint for SQL engine and the most
suitable function for such purpose is sql_stmt_compile(). However, this
function is not in a public API, and we don't want to expose it as
public function. As a solution function sql_fuzz() is introduced, it is
defined under a macro FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION.

Needed for #4826

NO_CHANGELOG=internal
NO_DOC=internal
NO_TEST=internal

Patch introduce a macro FUZZING_BUILD_MODE_UNSAFE_FOR_PRODUCTION. It is
a common build macro recommended in libFuzzer's documentation [1].

1. https://llvm.org/docs/LibFuzzer.html#fuzzer-friendly-build-mode

Needed for #4826

NO_CHANGELOG=build
NO_DOC=build
NO_TEST=build

Follows-up #4823

NO_CHANGELOG=internal
NO_DOC=internal
NO_TEST=internal

Function `datetime_strptime` decodes string with datetime according to
specified format, it accepts a datetime struct, buffer with datetime and
string with format in arguments. Fuzzing test used static string
"iso8601" as a format and it blocked fuzzing test to cover functions
used by datetime_strptime under the hood. Fuzz introspector shows that
code coveraged by a test is quite low.

Patch updates the test to make it more effective: buffer with datetime
and format string are generated using FDP (Fuzzing Data Provider).

Test file extension was changed to .cc, because FuzzingDataProvider is
used and we need building it by C++ compiler.

Function `tnt_strptime` uses assert, that triggered by fuzzing tests.
Therefore it was replaced with to if..then.

1. https://storage.googleapis.com/oss-fuzz-introspector/tarantool/

Fixes #8490

NO_CHANGELOG=fuzzing test
NO_DOC=fuzzing test
NO_TEST=fuzzing test

Fixes #8502
Needed for #8490

NO_DOC=bugfix
NO_TEST=covered by fuzzing test

NO_DOC=gdb extension
NO_TEST=gdb extension
NO_CHANGELOG=gdb extension

Closes #8317

NO_DOC=gdb extension
NO_TEST=gdb extension
NO_CHANGELOG=gdb extension

Rearrange code to simplify support of multiple types of lists

NO_DOC=gdb extension
NO_TEST=gdb extension
NO_CHANGELOG=gdb extension

Follow-up commit 4eb0a0dd ("gdb: check if address refers to symbol
with 'info symbol' command").

NO_DOC=gdb extension
NO_TEST=gdb extension
NO_CHANGELOG=gdb extension

Original test was flaky and passed even if the problem was not fixed in
Release build and more rarely in Debug. Let's improve this test to be more
reliable.

Closes tarantool/tarantool-qa#203

NO_CHANGELOG=test
NO_DOC=test

Currently test can fail if in developer environment .inputrc is custom.

Follow-up #4317

NO_TEST=test fix
NO_DOC=test fix
NO_CHANGELOG=test fix

Remove parts that used before flightrec reader API is added.

- fix test_max_record_size case of prbuf unit test to use reader from
  file
- drop prbuf reader from buffer in memory
- drop prbuf Lua FFI interface

Follow up: https://github.com/tarantool/tarantool-ee/issues/319

NO_DOC=internal
NO_CHANGELOG=internal

Found by EE CI in release asan in flightrec test which uses prbuf.
Change prbuf test so that the issue is tested in CE where the prbuf code
resides.

Follow-up https://github.com/tarantool/tarantool-ee/issues/319

NO_DOC=bug fix
NO_CHANGELOG=unreleased bug fix

Closes #8501

@TarantoolBot document
Title: Document memtx read view statistics

New entries have been added to `box.stat.memtx()` output:

```
tarantool> box.stat.memtx().data
---
- garbage: 0
  total: 24986
  read_view: 0
...

tarantool> box.stat.memtx().index
---
- read_view: 0
  total: 933888
...
```

`data` shows how much memory is allocated for memtx tuples:
 - `data.total` - total amount of memory allocated for data tuples.
   This includes `data.read_view` and `data.garbage` plus tuples that
   are actually stored in memtx spaces.
 - `data.read_view` - memory held for read views.
 - `data.garbage` - memory that is unused and scheduled to be freed
   (freed lazily on memory allocation).

`index` shows how much memory is allocated for memtx index extents:
 - `index.total` - total amount of memory allocated for indexing data.
   This includes `index.read_view` plus memory used for indexing tuples
   that are actually stored in memtx spaces.
 - `index.read_view` - memory held for read views.

All numbers are given in bytes.

`data.read_view` and `index.read_view` include memory allocated both for
system read views (snapshot, replication) and user read views (EE-only).
They should be non-zero only if there are open read views. To list all
open read views, use `box.read_view.list()`.

The commit adds new read view methods needed for pagination. Also, the
check that tuple is not nil is dropped for the sake of consistency - we
will omit this check in read view because it is incomplete anyway and
the error message looks fine without it.

Part of tarantool/tarantool-ee#285

NO_CHANGELOG=in EE
NO_DOC=in EE

Since we are going to implement pagination for read_view, we need to
factor out iterator_position logics to avoid duplication.

Part of tarantool/tarantool-ee#285

NO_TEST=refactoring
NO_CHANGELOG=refactoring
NO_DOC=refactoring

Currently, we check that tuple passed as position fits space format.
However, for pagination purposes, it's enough to validate only parts of
tuple that are used in cmp_def. And, it allows not to use space format -
we will need it in read view. So the patch replaces tuple validation with
validation of its cmp_def parts only.

Closes #8511
Part of tarantool/tarantool-ee#285

NO_DOC=bugfix