[tarantool-patches] Re: [PATCH 4/6] swim: wrap sio_strfaddr()

[Vladislav Shpilevoy <v.shpilevoy@tarantool.org>]

On 24/04/2019 19:40, Konstantin Osipov wrote:
> * Vladislav Shpilevoy <v.shpilevoy@tarantool.org> [19/04/24 18:50]:
>> SIO provides a function sio_strfaddr() to obtain string
>> representation of arbitrary struct sockaddr. And it uses
>> singleton buffer to store results, because some of them can't
>> fit into tt_static_buf() according to POSIX name limits.
> 
> This is minor, but I would make sio_strfaddr() use the same set of
> buffers as provided by tt_static_buf() instead, with a static
> assert that the buffer size is large enough.

As I said, the tt_static_buf is not big enough. SIO uses NI_MAXHOST
constant which on Linux is 1025 > size of tt_static_buf.

http://man7.org/linux/man-pages/man3/getnameinfo.3.html

If you still think we will not break anything, I can do that. Just
confirm that you are sure.

It will not eliminate swim_inaddr_str() though, because most of
the problems were about huge cast to 'const struct sockaddr *'
and passing sizeof(struct sockaddr_in) each time - it made code
unreadable.

> 
>> SWIM uses only AF_INET addresses, they are short enough to fit
>> into tt_static_buf(). Also SWIM is going to call sio_strfaddr()
>> on more than 1 address in a row in subsequent patches, and
>> singleton buffer does not work here - each call will overwrite
>> result of the previous. Besides, SWIM never uses struct sockaddr
>> type. All these reasons are sufficient to wrap sio_strfaddr()
>> with encapsulated cast to struct sockaddr *, and copying on
>> tt_static_buf().
>>
> 
> -- 
> Konstantin Osipov, Moscow, Russia, +7 903 626 22 32
> http://tarantool.io - www.twitter.com/kostja_osipov
>