svn commit: r335278 - head/bin/pwd

[Eitan Adler]

On 19 June 2018 at 12:57, Pedro Giffuni <pfg at freebsd.org> wrote:
>
>
> On 19/06/2018 11:25, John Baldwin wrote:
>>
>> On 6/18/18 10:26 PM, Eitan Adler wrote:
>>>
>>> On 18 June 2018 at 10:57, John Baldwin <jhb at freebsd.org> wrote:
>>>>
>>>> On 6/16/18 10:14 PM, Eitan Adler wrote:
>>>>>
>>>>> Author: eadler
>>>>> Date: Sun Jun 17 05:14:50 2018
>>>>> New Revision: 335278
>>>>> URL: https://svnweb.freebsd.org/changeset/base/335278
>>>>>
>>>>> Log:
>>>>>    pwd: mark usage as dead
>>>>
>>>> You keep committing changes like this and ignoring e-mails about them.
>>>
>>> I replied both the first time and this time. I may have
>>> (accidentally?) ignored similar emails though. The question I have is
>>> other than the mild code churn what's the harm?
>>
>> It adds clutter.  Also, fixing the tool means you fix all the places at
>> once rather than slowly adding workarounds one by one.
>>
>>>> What broken compiler are you using that doesn't properly inherit __dead2
>>>> from the call to exit()?
>>>
>>> In this case, scan-build50 was getting annoyed.
>>
>> Does scan-build from LLVM 6.0 handle this correctly?  If so, I'd say to
>> just mark this warning as broken (and thus ignore it) for scan-build50
>> just as we ignore certain warnings from GCC 4.2.1 because they are
>> broken-as-implemented.
>>
> FWIW, clang's scan-build is made to even more false positives and general
> noise than the regular compiler warnings.
> It is better to just ignore it unless it finds something real.

I don't consider this a real harm, but I'll try and remember to ignore
these in the future.

-- 
Eitan Adler
Source, Ports, Doc committer
Bugmeister, Ports Security teams