[Tarantool-patches] [PATCH 02/15] Check data_offset overflow in struct tuple

Aleksandr Lyapunov alyapunov at tarantool.org
Mon Jul 6 16:39:50 MSK 2020


Thank for the review.

On 05.07.2020 20:03, Vladislav Shpilevoy wrote:
> +i1 = s:create_index('pk')
>> + | ---
>> + | ...
>> +i2 = s:create_index('mk', {parts={{'[2][*]', 'uint'}}})
> 1. Why do you need a multikey index? This test lacks a comment.
Mutlikey index creates a variadic-length offset table in a tuple.
That is a simplest way to overflow it.
Sure, I'll add a comment.
>> diff --git a/test/box/huge_field_map_long.test.lua b/test/box/huge_field_map_long.test.lua
>> new file mode 100644
> 3. Why do you need a 'long' test? The first test seems to be good enough.

It's just another way to overflow offset table. For example 1.10 does not
have multikey indexes, so only long test will crash it. And perhaps 
we'll merge
into 1.10.
There's also a chance that we'll fix 
https://github.com/tarantool/tarantool/issues/5085
and multikey test will not work.



More information about the Tarantool-patches mailing list