[Tarantool-patches] [PATCH v1 1/1] schema: remove assert on wrong insert into _priv

Mergen Imeev imeevma at tarantool.org
Mon Sep 27 09:41:54 MSK 2021


Hi!

On 09.09.2021 14:07, Mergen Imeev wrote:
> Hi! Thank you for the review! My answers, diff and new patch below.
>
> On Tue, Sep 07, 2021 at 11:59:58PM +0200, Vladislav Shpilevoy wrote:
>> Hi! Thanks for the patch!
>>
>> See 4 comments below.
>>
>> On 06.09.2021 10:11, imeevma at tarantool.org wrote:
>>> Prior to this patch, an assertion was throwed if a tuple with an
>> 1. throwed -> thrown.
>>
> Fixed.
>
>>> incorrect id was inserted into the _priv system space. This bug
>>> appeared only in the debug build.
>>>
>>> Closes #6295
>>> ---
>>> https://github.com/tarantool/tarantool/issues/6295
>>> https://github.com/tarantool/tarantool/tree/imeevma/gh-6295-assert-on-insert-with-wrong-id
>> 2. Could you please add a changelog?
>>
> Added.
>
>>> diff --git a/test/box-tap/gh-6295-assert-on-wrong-id.test.lua b/test/box-tap/gh-6295-assert-on-wrong-id.test.lua
>>> new file mode 100755
>>> index 000000000..e4822f395
>>> --- /dev/null
>>> +++ b/test/box-tap/gh-6295-assert-on-wrong-id.test.lua
>>> @@ -0,0 +1,34 @@
>>> +#!/usr/bin/env tarantool
>>> +
>>> +local tap = require('tap')
>>> +local test = tap.test('gh-6295-assert-on-wrong-id')
>>> +
>>> +test:plan(5)
>>> +
>>> +local ok, res
>>> +
>>> +box.cfg{}
>>> +
>>> +-- Should be an error, not an assertion.
>>> +local _priv = box.space._priv
>>> +local errmsg = "Function '1000000' does not exist"
>>> +ok, res = pcall(_priv.replace, _priv, {1, 2, 'function', 1000000, box.priv.A})
>>> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Function exists")
>> 3. Shouldn't the message be like a comment? Describing the correct
>> behaviour. It is printed always along with ok/no ok in the tap output.
>> So it should be about what is expected.
>>
> Thanks, fixed.
>
>>> diff --git a/test/box-tap/suite.ini b/test/box-tap/suite.ini
>>> index b09d7db4f..fd55d5d24 100644
>>> --- a/test/box-tap/suite.ini
>>> +++ b/test/box-tap/suite.ini
>>> @@ -3,7 +3,7 @@ core = app
>>>   description = Database tests with #! using TAP
>>>   is_parallel = True
>>>   use_unix_sockets_iproto = True
>>> -release_disabled = errinj_set_with_enviroment_vars.test.lua
>>> +release_disabled = errinj_set_with_enviroment_vars.test.lua, gh-6295-assert-on-wrong-id.test.lua
>> 4. Why? The test does not pass in release?
>>
> It does, however in release build nothing will change. At least at the first
> glance.
>
>>>   config = suite.cfg
>>>   fragile = {
>>>       "retries": 10,
>>>
>
> Diff:
>
>
> diff --git a/changelogs/unreleased/gh-6295-assert-on-wrong-insert-into-_priv.md b/changelogs/unreleased/gh-6295-assert-on-wrong-insert-into-_priv.md
> new file mode 100644
> index 000000000..058d218d7
> --- /dev/null
> +++ b/changelogs/unreleased/gh-6295-assert-on-wrong-insert-into-_priv.md
> @@ -0,0 +1,5 @@
> +## bugfix/sql
> +
> +* Now inserting a tuple with the wrong "id" field into the \_priv space will
> +  return the correct error (gh-6295).
> +
> diff --git a/test/box-tap/gh-6295-assert-on-wrong-id.test.lua b/test/box-tap/gh-6295-assert-on-wrong-id.test.lua
> index e4822f395..add2df429 100755
> --- a/test/box-tap/gh-6295-assert-on-wrong-id.test.lua
> +++ b/test/box-tap/gh-6295-assert-on-wrong-id.test.lua
> @@ -13,22 +13,22 @@ box.cfg{}
>   local _priv = box.space._priv
>   local errmsg = "Function '1000000' does not exist"
>   ok, res = pcall(_priv.replace, _priv, {1, 2, 'function', 1000000, box.priv.A})
> -test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Function exists")
> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Proper error is returned")
>   
>   errmsg = "Sequence '1000000' does not exist"
>   ok, res = pcall(_priv.replace, _priv, {1, 2, 'sequence', 1000000, box.priv.A})
> -test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Sequence exists")
> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Proper error is returned")
>   
>   errmsg = "Space '1000000' does not exist"
>   ok, res = pcall(_priv.replace, _priv, {1, 2, 'space', 1000000, box.priv.A})
> -test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Space exists")
> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Proper error is returned")
>   
>   errmsg = "User '1000000' is not found"
>   ok, res = pcall(_priv.replace, _priv, {1, 2, 'user', 1000000, box.priv.A})
> -test:is_deeply({ok, tostring(res)}, {false, errmsg}, "User exists")
> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Proper error is returned")
>   
>   errmsg = "Role '1000000' is not found"
>   ok, res = pcall(_priv.replace, _priv, {1, 2, 'role', 1000000, box.priv.A})
> -test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Role exists")
> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Proper error is returned")
>   
>   os.exit(test:check() and 0 or 1)
> diff --git a/test/box-tap/suite.ini b/test/box-tap/suite.ini
> index fd55d5d24..b09d7db4f 100644
> --- a/test/box-tap/suite.ini
> +++ b/test/box-tap/suite.ini
> @@ -3,7 +3,7 @@ core = app
>   description = Database tests with #! using TAP
>   is_parallel = True
>   use_unix_sockets_iproto = True
> -release_disabled = errinj_set_with_enviroment_vars.test.lua, gh-6295-assert-on-wrong-id.test.lua
> +release_disabled = errinj_set_with_enviroment_vars.test.lua
>   config = suite.cfg
>   fragile = {
>       "retries": 10,
>
>
> New patch:
>
>
> commit a3542e0082c480f166796281d3996f6edfffcaf7
> Author: Mergen Imeev <imeevma at gmail.com>
> Date:   Sat Aug 7 13:11:12 2021 +0300
>
>      schema: remove assert on wrong insert into _priv
>      
>      Prior to this patch, an assertion was thrown if a tuple with an invalid
>      id was inserted into the _priv system space. This bug appeared only in
>      the debug build.
>      
>      Closes #6295
>
> diff --git a/changelogs/unreleased/gh-6295-assert-on-wrong-insert-into-_priv.md b/changelogs/unreleased/gh-6295-assert-on-wrong-insert-into-_priv.md
> new file mode 100644
> index 000000000..058d218d7
> --- /dev/null
> +++ b/changelogs/unreleased/gh-6295-assert-on-wrong-insert-into-_priv.md
> @@ -0,0 +1,5 @@
> +## bugfix/sql
> +
> +* Now inserting a tuple with the wrong "id" field into the \_priv space will
> +  return the correct error (gh-6295).
> +
> diff --git a/src/box/alter.cc b/src/box/alter.cc
> index 3bd56feb9..e87fbb847 100644
> --- a/src/box/alter.cc
> +++ b/src/box/alter.cc
> @@ -3954,6 +3954,8 @@ priv_def_check(struct priv_def *priv, enum priv_type priv_type)
>   		return -1;
>   	}
>   	const char *name = schema_find_name(priv->object_type, priv->object_id);
> +	if (name == NULL)
> +		return -1;
>   	if (access_check_ddl(name, priv->object_id, grantor->def->uid,
>   			     priv->object_type, priv_type) != 0)
>   		return -1;
> diff --git a/src/box/schema.cc b/src/box/schema.cc
> index 1970871cc..cf1e531e3 100644
> --- a/src/box/schema.cc
> +++ b/src/box/schema.cc
> @@ -701,36 +701,51 @@ schema_find_name(enum schema_object_type type, uint32_t object_id)
>   	case SC_SPACE:
>   		{
>   			struct space *space = space_by_id(object_id);
> -			if (space == NULL)
> -				break;
> -			return space->def->name;
> +			if (space != NULL)
> +				return space->def->name;
> +			diag_set(ClientError, ER_NO_SUCH_SPACE,
> +				 tt_sprintf("%d", object_id));
> +			break;
>   		}
>   	case SC_FUNCTION:
>   		{
>   			struct func *func = func_by_id(object_id);
> -			if (func == NULL)
> -				break;
> -			return func->def->name;
> +			if (func != NULL)
> +				return func->def->name;
> +			diag_set(ClientError, ER_NO_SUCH_FUNCTION,
> +				 tt_sprintf("%d", object_id));
> +			break;
>   		}
>   	case SC_SEQUENCE:
>   		{
>   			struct sequence *seq = sequence_by_id(object_id);
> -			if (seq == NULL)
> -				break;
> -			return seq->def->name;
> +			if (seq != NULL)
> +				return seq->def->name;
> +			diag_set(ClientError, ER_NO_SUCH_SEQUENCE,
> +				 tt_sprintf("%d", object_id));
> +			break;
>   		}
>   	case SC_ROLE:
> -	case SC_USER:
>   		{
>   			struct user *role = user_by_id(object_id);
> -			if (role == NULL)
> -				break;
> -			return role->def->name;
> +			if (role != NULL)
> +				return role->def->name;
> +			diag_set(ClientError, ER_NO_SUCH_ROLE,
> +				 tt_sprintf("%d", object_id));
> +			break;
> +		}
> +	case SC_USER:
> +		{
> +			struct user *user = user_by_id(object_id);
> +			if (user != NULL)
> +				return user->def->name;
> +			diag_set(ClientError, ER_NO_SUCH_USER,
> +				 tt_sprintf("%d", object_id));
> +			break;
>   		}
>   	default:
> -		break;
> +		unreachable();
>   	}
> -	assert(false);
> -	return "(nil)";
> +	return NULL;
>   }
>   
> diff --git a/test/box-tap/gh-6295-assert-on-wrong-id.test.lua b/test/box-tap/gh-6295-assert-on-wrong-id.test.lua
> new file mode 100755
> index 000000000..add2df429
> --- /dev/null
> +++ b/test/box-tap/gh-6295-assert-on-wrong-id.test.lua
> @@ -0,0 +1,34 @@
> +#!/usr/bin/env tarantool
> +
> +local tap = require('tap')
> +local test = tap.test('gh-6295-assert-on-wrong-id')
> +
> +test:plan(5)
> +
> +local ok, res
> +
> +box.cfg{}
> +
> +-- Should be an error, not an assertion.
> +local _priv = box.space._priv
> +local errmsg = "Function '1000000' does not exist"
> +ok, res = pcall(_priv.replace, _priv, {1, 2, 'function', 1000000, box.priv.A})
> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Proper error is returned")
> +
> +errmsg = "Sequence '1000000' does not exist"
> +ok, res = pcall(_priv.replace, _priv, {1, 2, 'sequence', 1000000, box.priv.A})
> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Proper error is returned")
> +
> +errmsg = "Space '1000000' does not exist"
> +ok, res = pcall(_priv.replace, _priv, {1, 2, 'space', 1000000, box.priv.A})
> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Proper error is returned")
> +
> +errmsg = "User '1000000' is not found"
> +ok, res = pcall(_priv.replace, _priv, {1, 2, 'user', 1000000, box.priv.A})
> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Proper error is returned")
> +
> +errmsg = "Role '1000000' is not found"
> +ok, res = pcall(_priv.replace, _priv, {1, 2, 'role', 1000000, box.priv.A})
> +test:is_deeply({ok, tostring(res)}, {false, errmsg}, "Proper error is returned")
> +
> +os.exit(test:check() and 0 or 1)


More information about the Tarantool-patches mailing list