Re: [Tarantool-patches] [PATCH v6 08/25] Fix luacheck warnings in test/app-tap

[Igor Munkin <imun@tarantool.org>]

Sasha,

On 01.06.20, Alexander Turenko wrote:
> > > > -local function execute_and_verify(test, client, input, exp_output, name)
> > > > -    test:test(name, function(test)
> > > > +local function execute_and_verify(testcase, client, input, exp_output, name)
> > > > +    testcase:test(name, function(test)
> > > 
> > > Please, don't forbid redefinitions. I often use it for 'standard' names
> > > like 'opts', 'ok', 'err', 'test'. It allows to write code blocks in the
> > > same style:
> > > 
> > >  | local ok, err = <...>
> > >  | <...>
> > >  | local ok, err = <...>
> > > 
> > 
> > I have no objections regarding this proposal, only two nits:
> > * Those suppressions should be enabled source-wide *only and only when*
> >   all unintentional occurences or the ones masking bugs (if we found
> >   such) are fixed. Otherwise I see no sence in enabling luacheck.
> 
> Aren't real bugfixes (that actually changes present behaviour) separated
> from style changes?

Those found by Vlad[1] were separated[2][3] and already merged to stable
branches. The one found by me[4] in tarantoolctl tests is still mixed in
the corresponding patch[5].

> 
> Don't get what you meant with 'unintentional occurences': if it is not a
> bug and not a code style violation, than it it does not look as
> something that should be fixed. But if there is an example, where
> luacheck reveals a place that is worth to change for, say, readability,
> then okay.

I was talking about the following scenario: a suppression is enabled
source-wide without checking the root cause and masked bugs or
'unintentional occurences' are left in the source code. This is exactly
the case we faced with the bugs I mentioned above. Thereby I propose to
investigate errors prior to suppress them. Otherwise, as I said before,
enabling luacheck looks like cargo cult.

I expect no bugs produced by redefenition but can't say it for sure.

> 
> I bet a beer that this certain warning does not give us any real bug.

Agree with you here.

> 
> The main advantage of luacheck is preventing some kinds of typos: say, a
> typo in a function name in some failure branch, which is not covered by
> tests. The tool is definitely useful without the redefinition warning.

Precisely.

> 
> WBR, Alexander Turenko. 

[1]: https://lists.tarantool.org/pipermail/tarantool-patches/2020-April/015930.html
[2]: https://lists.tarantool.org/pipermail/tarantool-patches/2020-April/015933.html
[3]: https://lists.tarantool.org/pipermail/tarantool-patches/2020-April/015934.html
[4]: https://lists.tarantool.org/pipermail/tarantool-patches/2020-April/016066.html
[5]: https://lists.tarantool.org/pipermail/tarantool-patches/2020-May/017245.html

-- 
Best regards,
IM