Re: [Tarantool-patches] [PATCH v2] iproto: add an empty body to the unprepare response

[Kirill Yukhin <kyukhin@tarantool.org>]

Hello,

On 15 мар 16:34, Vladislav Shpilevoy wrote:
> 
> 
> On 06/03/2020 18:27, Alexander Turenko wrote:
> > On Thu, Mar 05, 2020 at 08:41:35AM +0000, Nikita Pettik wrote:
> >> On 05 Mar 08:41, Kirill Yukhin wrote:
> >>> Hello,
> >>>
> >>> On 03 мар 19:16, Chris Sosnin wrote:
> >>>> Absence of the body in the unprepare response forces users to perform
> >>>> additional checks to avoid errors. Adding an empty body fixes this problem.
> >>>>
> >>>> Closes #4769
> >>>> ---
> >>>> branch: https://github.com/tarantool/tarantool/tree/ksosnin/gh-4769-unprepare-response-body
> >>>> issue: https://github.com/tarantool/tarantool/issues/4769
> >>>>
> >>>> As Nikita suggested, I created box/iproto.test.lua, and basically
> >>>> inserted wrappers for requests testing from box-py for future usage.
> >>>
> >>> Could you please rename the test to be not so generic?
> >>> Like box/gh-4769-iproto-unprep-body or whatever.
> >>
> >> Kirill, this test is going to assemble all iproto-related tests
> >> which don't rely on net.box module. Setting up all preparations
> >> required for raw iproto communication results in duplicating ~30-40
> >> lines of code in each test file.
> > 
> > Technically there are two ways to extract helpers from a 'core =
> > tarantool' test:
> > 
> > * Add it to, say, test/box/box.lua and to _G.protected_globals.
> > * Add it to a separate Lua file in test/box/lua and to 'lua_libs' field
> >   in test/box/suite.ini. After this you can use `require` for this
> >   module in a test.
> > 
> > So technically you're not blocked here. Both ways are available and
> > don't lead to much code duplication, but the process (SOP) requires to
> > add a test for a bug to a separate file. (Personally I still don't sure
> > it is good, but anyway.)
> > 
> > NB: 'receive', not 'recieve'. Very often typo.
> > 
> > WBR, Alexander Turenko.
> 
> The whole purpose of the 'one issue - one file' was to simplify
> reproducibility in a console. When you need to extract some helpers
> into a second file, the idea does not work anymore, but just complicates
> life, when you need to invent how to make resuable and abstract
> something, which is not needed to be reusable and abstract really.

Also, the purpose was to separate cases so they're not interfere.
I guess, we might have something like 'gcc -E' to preprocess
the cases.

--
Regards, Kirill Yukhin