[cfe-dev] [clang-tidy] Excluding individual files/lines from clang-tidy checks

[Mehdi Amini via cfe-dev]

> On Nov 2, 2016, at 4:42 PM, Nikita via cfe-dev <cfe-dev at lists.llvm.org> wrote:
> 
> I think I'll explain why I need this functionality and why I think it
> would be useful for others.
> 
> Currently clang-tidy doesn't provide a practical way to suppress
> diagnostics from headers you don't have control over. If using a
> function defined in such header causes a false positive, there is no
> easy way to deal with this issue. Since you cannot modify a header,
> you cannot add NOLINT to it. And adding NOLINT to each invocation of
> such function is either impossible or hugely impractical if
> invocations are ubiquitous.
> 
> Using the '-header-filter' doesn't help to address this issue as well.
> Unless your own headers are strategically named, it is virtually
> impossible to craft a regex that will filter out only the third-party
> ones.
> 
> The '-line-filter' can be used to suppress such warnings, but it's not
> very convenient. Instead of excluding a line that produces a warning
> you have to include all other lines. Plus, it provides no way to
> suppress only certain warnings from a file.
> 
> With all this mentioned, I think that clang-tidy would benefit from a
> mechanism that is specifically designed to suppress certain warnings
> from certain lines in a convenient way.

That causes some maintainability issues though: let say I filter something between line 1000 and 1005 today, what if some code is later added before?
Do you see a way of handling this?

— 
Mehdi


> 
> I don't see how existing options can be changed to provide this
> functionality, so I propose adding a new option -
> '--exclude-from-checks' - that will take a list of locations to
> suppress warnings from. It can additionally take a list of checks to
> suppress. The format of its' argument can be a JSON array similar to
> the one '-line-filter' takes:
> [{"name":"header.h","lines":[[1,5]],"checks":"cert-*"}].
> 
> I would gladly appreciate any feedback on this idea and look forward
> to the discussion.
> 
> Regards,
> Nikita
> 
> On Sat, Oct 29, 2016 at 12:11 AM, Nikita <nikita.kakuev at gmail.com> wrote:
>> Well, I'm trying to mimic the behavior of '-line-filter'. But I also
>> think that JSON is cumbersome. It requires putting arguments in
>> quotes, plus the format itself seems a bit redundant. Specifying
>> '[{"name":"file.h"}]' without the line range seemingly does nothing.
>> Diagnostics for "file.h" will be emitted even without this filter.
>> 
>> On Fri, Oct 28, 2016 at 11:52 PM, Richard via cfe-dev
>> <cfe-dev at lists.llvm.org> wrote:
>>> 
>>> [Please reply *only* to the list and do not include my email directly
>>> in the To: or Cc: of your reply; otherwise I will not see your reply.
>>> Thanks.]
>>> 
>>> In article <CAGJfN2yb=3D-oCyTycsULN7F=Gn=KxFW_O4r4ndzCd8bFZ2muw at mail.gmail.com>,
>>>    Nikita via cfe-dev <cfe-dev at lists.llvm.org> writes:
>>> 
>>>>> clang-tidy '-exclude-from-checks=[{"name":"exclude.c","lines":[[5,7]]}]'
>>> 
>>> Why are we specifying command-line arguments as JSON?
>>> 
>>> It seems unnecessarily cumbersome to me.
>>> --
>>> "The Direct3D Graphics Pipeline" free book <http://tinyurl.com/d3d-pipeline>
>>>            The Terminals Wiki <http://terminals-wiki.org>
>>>     The Computer Graphics Museum <http://ComputerGraphicsMuseum.org>
>>>  Legalize Adulthood! (my blog) <http://LegalizeAdulthood.wordpress.com>
>>> _______________________________________________
>>> cfe-dev mailing list
>>> cfe-dev at lists.llvm.org
>>> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev
> _______________________________________________
> cfe-dev mailing list
> cfe-dev at lists.llvm.org
> http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev