Added --exclude option (#54).

(cherry picked from commit c17c10a4)

This allows to overcome problems when CMake chooses one toolchain to
build tarantool, but a library (libluajit.a or libcurl.a) is built using
another (incompatible) toolchain.

Fixes #4587.

(cherry picked from commit 1eead75e)

FreeBSD has OpenSSL as part of the base system: libraries are located in
/usr/lib, headers are in /usr/include. However a user may install the
library into /usr/local/{lib,include} from ports / pkg. In this case
tarantool did choose /usr/local version, while libcurl will pick up a
base system library. This is fixed by passing --with-ssl option with an
argument (/usr/local or /usr if custom -DOPENSSL_ROOT_DIR=<...> is not
passed).

Now the behaviour is the following. If -DOPENSSL_ROOT_DIR=<...> is
passed, then try to use OpenSSL from it. Otherwise find the library in
/usr/local and then in /usr. This is right as for tarantool's crypto
module as well as for libcurl submodule.

There is a flaw here: a user is unable to choose a base system library
if a ports / pkg version of OpenSSL is installed. The reason here is
that tarantool's crypto module depends on other libraries and
-I/usr/local/include may be added to build options. I have no good
solution for that, so `cmake . -DOPENSSL_ROOT_DIR=/usr` will give a
warning on FreeBSD and `gmake` likely will fail if libraries are of
different versions (see cmake/os.cmake comments for more information).
See also a [discussion][1] in FreeBSD community about all those /usr and
/usr/local problems.

There were two other problems that may fail tarantool build on FreeBSD:
they are fixed in this commit and described below.

First, libcurl's configure script chooses GCC by default if it exists
(say, installed from ports / pkg). It is unexpected behaviour when
tarantool sources itself are built with clang. Now it is fixed by
passing a compiler explicitly to the libcurl's configure script: the
library will use base system clang by default or one that a user pass to
tarantool's cmake.

Side note: GCC has /usr/local/include in its default headers search
paths; libcurl's configure script chooses GCC as a compiler and OpenSSL
from a base system by default (when CC and --with-ssl=<...> are not set)
that leads to OpenSSL header / library mismatch. It is the primary
reason of the build fail that was fixed in
1f2338bd ('build: FreeBSD packages
installation'). It is not much relevant anymore, because we don't try to
link with a base system OpenSSL if /usr/local one exists (however if it
is asked explicitly with -DOPENSSL_ROOT_DIR=<...> we'll do, but will
give a warning). Anyway, it is important to know such details if we'll
change build scripts in a future.

Second, backtraces are not supported on FreeBSD, but were enabled if
libunwind headers is found. This leads to an error on cmake stage,
because of inability to find a right library (this is a bug). Now we
disable backtraces on FreeBSD by default even if libunwind is found. See

When CC is passed to libcurl's configure script, the new problem opens
on Mac OS. CMake chooses XCode toolchain by default (at least on a
particular system where I tried it), which requires -isysroot=<SDK_PATH>
option to be passed to a preprocessor and a compiler in order to find
system headers. See [2] for more information.

[1]: https://wiki.freebsd.org/WarnerLosh/UsrLocal
[2]: https://developer.apple.com/documentation/xcode_release_notes/xcode_10_release_notes#3035623

Follows up #4490.

(cherry picked from commit ea5929db)

Before the patch there was a race in replication
password configuration. It was possible that a replica
connects to a master with a custom password before
that password is actually set. The replica treated the
error as critical and exited.

But in fact it is not critical. Replica even can
withstand absence of a user and keeps reconnecting.
Wrong password situation arises from the same problem
of non atomic configuration and is fixed the same -
keep reconnect attempts if the password was wrong.

Closes #4550

(cherry picked from commit aa2e2c56)

The previous patch introduced a way to set box.cfg options
in a strict order, even on a reconfiguration. It was used to set
listen before replication.
The same order problem existed for replication settings. A user
could do

    box.cfg{
        replication_connect_quorum = 0,
        replication = {...}
    }

and expect, that due to quorum 0 the cfg() will return
immediately. But actually the behaviour was undefined - due to
arbitrary order of keys in a Lua table, replication could be
applied before quorum.

The patch makes all replication settings be applied before
replication.

Follow up #4433
Part of #3760

(cherry picked from commit 00c6c437)

Before the patch the nil UUID was ignored and a new random one
was generated. This was because internally box treats nil UUID
as its absence.

Now a user will see an explicit message that nil UUID is a
reserved value.

Closes #4282

(cherry picked from commit a8ebd334)

Added Ubuntu 19.10 Eoan Ermine into CI.

Close #4583

Reviewed-by: Alexander Turenko <alexander.turenko@tarantool.org>
(cherry picked from commit c6cd2e62)

Before this patch RUN_TESTS condition in Dockerfile.staticbuild was
ignored and always was true.

However adding of brackets solves only part of problem. If RUN_TESTS is
empty `sh -c` returns 1 and build fails.

However if we run tests we should fail build if tests are not passed.
Ternary logic was rewritten to fair if-else.

This patch fixes it and allows build tarantool statically without
running tests.

@Totktonada: Fixed .gitlab.mk to pass RUN_TESTS environment variable to
docker build arguments.

Reviewed-by: Alexander Turenko <alexander.turenko@tarantool.org>
(cherry picked from commit 8c85dbc7)

After we started using bundled version of libyaml by default (see commit
47b91e90), we can remove it from
building dependencies for RPM and DEB packages.

Closes #4442

Reviewed-by: Alexander Turenko <alexander.turenko@tarantool.org>
(cherry picked from commit 1d4e584a)

Is_nullable JSON-path indexes used to lose is_nullable property
being defined on a space having format. This patch fixes this
problem.

Closes #4520

(cherry picked from commit 3d01b3af)

Functional index extractor code used to raise an invalid
error message when the user-defined extractor function
returns a scalar instead of a table.

Closes #4553

(cherry picked from commit 260a3b3a)

Wrap throwing lua_newthread in luaT_newthread using luaT_cpcall
to process arising error properly.

Closes #4556

(cherry picked from commit 54e23d6d)

End recovery (which means building secondary indexes) just after
last known log file was read. This allows fast switch to hot standby
instance without any delay for secondary index to be built.
Due to engine_end_recovery carryover, xdir_collect_inprogress,
previously being called from it, is now moved to garbage collector.

Closes #4135

(cherry picked from commit 5aa243de)

Console client's eval() method in case of an error at
reading from a socket was trying to return a variable
declared in a different view scope. Instead, the
error should be raised to drop the connection.

Reviewed-by: Alexander Turenko <alexander.turenko@tarantool.org>
(cherry picked from commit 89ec1d97)

When we handle unix console connection we should setup
eos if only we meet "\set output" command in a stream.
For example there could be "\set language" which should
not affect eos settings.

Closes #4568

Reported-by: Alexander Turenko <alexander.turenko@tarantool.org>
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Alexander Turenko <alexander.turenko@tarantool.org>
(cherry picked from commit 092cd675)

This will allow us to reuse this map in other routines
and same time optimize memory allocation a bit.

Part-of #4568

Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
Reviewed-by: Alexander Turenko <alexander.turenko@tarantool.org>
(cherry picked from commit 44bfb4f1)

Rows_per_wal option was deprecated because it can be covered by
wal_max_size. In order not to complicate WAL code with that
option's support this commit drops it completely.

In some tests the option was used to create several small xlog
files. Now the same is done via wal_max_size. Where it was
needed, number of rows per wal is estimated as wal_max_size / 50.
Because struct xrow_header size ~= 50 not counting paddings and
body.

Note, file box/configuration.result was deleted here, because it
is a stray result file, and it contained the rows_per_wal option
mentioning. Its test was dropped much earlier in
fdc3d1dd.

Closes #3762

(cherry picked from commit c6012920)

In patch c6bea65f I
added a bug - replication/misc leaves a bad value in
box.cfg.replication. Before that patch the test was
resetting this to empty replication. In my patch I
forgot about that, and left there the value

    {box.cfg.listen, "12345"}

This patch cleans it up.

Follow up #3760

(cherry picked from commit 399899a1)

test_run:wait_upstream() and test_run:wait_downstream() now wait until
an upstream / a downstream appears. This prevents an attempt to index a
nil value when one of those functions are called before a record about a
peer appears in box.info.replication. It was observed on
replication/show_error_on_disconnect test after commit
c6bea65f ('replication: recfg with 0
quorum returns immediately').

Fixes #4563.

(cherry picked from commit 18864835)

Replication quorum 0 not only affects orphan status, but also,
according to documentation, makes box.cfg() return immediately
regardless of whether connections to upstreams are established.

It was not so before the patch. What is worse, even with non 0
quorum the instance was blocked on reconfiguration for connect
timeout seconds, if at least one node is not connected.

Now quorum is respected on reconfiguration. On a bootstrap it is
still impossible to return earlier than
replication_connect_timeout, because nodes need to choose some
cluster settings. Too early start would make it impossible -
cluster's participants will just start and choose different
cluster UUIDs.

Closes #3760

(cherry picked from commit c6bea65f)

A successfully fetched remote instance ballot isn't updated during
bootstrap procedure. This leads to a case when different instances
choose different masters as their bootstrap leaders.

Imagine such a situation.
You start instance A without replication set up. Instance A successfully
bootstraps.
You also have instances B and C both with replication set up to {A, B,
C} and replication_connect_quorum set to 3
You first start instance B. It doesn't proceed to choosing a leader
until one of the events happens: either the replication_connect_timeout
runs out, or instance C is up and starts listening on its port.
B has established connection to A and fetched its ballot, with some
vclock, say, {1: 1}.
B retries connection to C every replication_timeout seconds.
Then you start instance C. Instance C succeeds in connecting to A and B
right away and bootstraps from instance A. Instance A registers C in its
_cluster table. This registration is replicated to instance C.
Meanwhile, instance C is trying to sync with quorum instances (which is
3), and stays in orphan mode.
Now replication_timeout on instance B finally runs out. It retries a
previously unsuccessful connection to C and succeeds. C sends its ballot
to B with vclock = {1: 2, 2:0} (in our example), since it has already
incremented it after _cluster registration.
B sees that C has a greater vclock than A, and chooses to bootstrap from
C instead of A. C is orphan and rejects B's attempt to join. B dies.

To fix such ungentlemanlike behaviour of C, we should at least include
loading status in ballot and prefer fully bootstrapped instances to the
ones still syncing with other replicas.
We also need to use a separate flag instead of ballot's already existent
is_ro, since we still want to prefer loading instances over the ones
explicitly configured to be read-only.

Closes #4527

(cherry picked from commit dc1e4009)

(cherry picked from commit 8749b3e1)

During rework of the console lua mode series
the declaration of variable has been lost and
this cause test case for remote unix console
connection to fail.

Fixes issue from c358398c

Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
(cherry picked from commit df821d0f)

To be consistent in naming scheme, for example
we already have parse_operators.

Part-of #3834

Reviewed-by: Konstantin Osipov <kostja.osipov@gmail.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
(cherry picked from commit af9ef211)

The other modules would be able to find out which eos marker
is currently active. For example when reading replies from
remote server via text based console protocol.

@TarantoolBot document
Title: > require('console').eos()

Returns a string with currently active end of string marker.

Part-of #3834

Reviewed-by: Konstantin Osipov <kostja.osipov@gmail.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
(cherry picked from commit 9d83b5f6)

If we change output mode on remote machine via
text-based session

 node a (server)
 ------
 require('console').listen('unix/:/tmp/X.sock')
 ...

 node b (client)
 ------
 require('console').connect('unix/:/tmp/X.sock')
 connected to unix/:/tmp/X.sock
 ...
 unix/:/tmp/X.sock> \set output lua

the client get stuck forever, it is because the text
wire protocol waits for yaml eos terminator which of
course never hit the peer, because lua mode uses own
one.

Thus to fix this problem we have to preprocess the text
we're passing to the server, just like we do in local
console. So we reuse command parsing code and remember
current output terminator in text_connection_mt instance.

Another problem is that named default output mode. There
could be a mixed environment where server operates in
default lua mode but client connects with yaml mode. To
break a tie we yield "\set output" command with current
output mode when establishing a connection. Since the
output format is per-session bound this doesn't affect
any other connections on a server.

Part-of #3834

Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
(cherry picked from commit c358398c)

We will need to parse and verify "\set" commands
in remote console text-wire protocol thus to not
duplicate the code lets factor helper out for
reuse sake.

Part-of #3834

Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
(cherry picked from commit 96dbc49d)

This will help us to distinguish end of string/stream
in text protocols (such as remote console connection).

Note that we start printing ";" terminator for every
lua output. Actually current yaml output does the
same but inside console.c module.

And since lua output is yet a new feature in stabilization
phase we're safe to make such changes without breaking api.

Part-of #3834

Reviewed-by: Konstantin Osipov <kostja.osipov@gmail.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
(cherry picked from commit d67808fa)

This allows lua to not create them every call since
they are constant speeding up processing a bit.

Part-of #3834

Reviewed-by: Konstantin Osipov <kostja.osipov@gmail.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
(cherry picked from commit 840a7646)

Sole symbols (such as box.NULL) are not processed
by serpent "custom" symbols feature, since they
are not in table.

Thus we should process them separately. Without it
we have

 > require('console').set_default_output("lua,block")
 > ;
 > box.NULL
 > "cdata<void %*>: NULL";

instead of

 > box.NULL
 > box.NULL;

as it should be.

Part-of #3834

Reviewed-by: Konstantin Osipov <kostja.osipov@gmail.com>
Signed-off-by: Cyrill Gorcunov <gorcunov@gmail.com>
(cherry picked from commit 6847367f)

fiber_stack_destroy() doesn't depend on cord() helper but
rather accepts slab cache pointer in arguments, lets
make the same in fiber_stack_create() helper, otherwise
it looks quite imbalanced.

Signed-off-by: Kirill Yukhin <kyukhin@tarantool.org>
(cherry picked from commit fa6d78b397d375d4c18875ac72b3798b2e06a1e8)

The mprotect() syscall may fail due to heavy memory
pressure (say there is no enough resources to split
VMAs and etc).

In this case we must not proceed without guard page
but refuse to create a new fiber.

Closes #4541

Signed-off-by: Kirill Yukhin <kyukhin@tarantool.org>
(cherry picked from commit e58dfe07189134297ea956eb70cd7d96e20a6c6e)

This will allow us to call it from inside fiber_stack_create
as we need to clean up resources on error.

Part-of #4541

Signed-off-by: Kirill Yukhin <kyukhin@tarantool.org>
(cherry picked from commit 8f1251d3611aa30d5bb6098716e965fd8dc08551)

(cherry picked from commit 0b9de586)

Count lines in the json parsing structure. It is needed to print
the number of line and column where a mistake was made.

Closes #3316

(cherry picked from commit 9f9bd3eb2d064129ff6b1a764140ebef242d7ff7)
(cherry picked from commit 53d43160)

Code to run main script (passed via command line args, or
interactive console) has a footer where it notifies systemd,
logs a happened error, and panics.

Before the patch that code was unreachable in case of any
exception in a main script, because panic happened earlier. Now a
happened exception is correctly carried to the footer with proper
error processing.

A first and obvious solution was replace all panics with diag_set
and use fiber_join on the script runner fiber. But appeared, that
the fiber running a main script can't be joined. This is because
normally it exits via os.exit() which never returns and therefore
its caller never dies = can't be joined.

The patch solves this problem by passing main fiber diag to the
script runner by pointer, eliminating fiber_join necessity.

Closes #4382

(cherry picked from commit 157a2d88)

Closes #4434
Follow-up #4366

@TarantoolBot document
Title: json/msgpack.cfg.encode_deep_as_nil option

Tarantool has several so called serializers to convert data
between Lua and another format: YAML, JSON, msgpack.

YAML is a crazy serializer without depth restrictions. But for
JSON, msgpack, and msgpackffi a user could set encode_max_depth
option. That option led to crop of a table when it had too many
nested levels. Sometimes such behaviour is undesirable.

Now an error is raised instead of data corruption:

    t = nil
    for i = 1, 100 do t = {t} end
    msgpack.encode(t) -- Here an exception is thrown.

To disable it and return the old behaviour back here is a new
option:

    <serializer>.cfg({encode_deep_as_nil = true})

Option encode_deep_as_nil works for JSON, msgpack, and msgpackffi
modules, and is false by default. It means, that now if some
existing users have cropping, even intentional, they will get the
exception.

(cherry picked from commit d7a8942a)

Tuple is a C library exposed to Lua. In Lua to translate Lua
objects into tuples and back luaL_serializer structure is used.

In Tarantool we have several global serializers, one of which is
for msgpack. Tuples store data in msgpack, and in theory should
have used that global msgpack serializer. But in fact the tuple
module had its own private serializer because of tuples encoding
specifics such as never encode sparse arrays as maps.

This patch makes tuple Lua module use global msgpack serializer
always. But how does tuple handle sparse arrays now? In fact,
the tuple module still has its own serializer, but it is updated
each time when the msgpack serializer is changed.

Part of #4434

(cherry picked from commit 676369b1)

Msgpack Lua module is not a simple set of functions. It is a
global serializer object used by plenty of other Lua and C
modules. Msgpack as a serializer can be configured, and in theory
its configuration updates should affect all other modules. For
example, a user could change encode_max_depth:

    require('msgpack').cfg({encode_max_depth = <new_value>})

And that would make tuple:update() accept tables with <new_value>
depth without a crop.

But in fact msgpack configuration didn't affect some places, such
as this one. And all the others who use msgpackffi.

This patch fixes it, for encode_max_depth option. Other options
are still ignored.

Part of #4434

(cherry picked from commit 4bb253f7)

There are some objects called serializers - msgpack, cjson, yaml,
maybe more. They are global objects affecting both Lua and C
modules.

A serializer have settings which can be updated. But before the
patch an update changed only C structure of the serializer. It
made impossible to use settings of the serializers from Lua.

Now any update of any serializer is reflected both in its C and
Lua structures.

Part of #4434

(cherry picked from commit fe4a8047)