Re: [boost] [1.34.0beta] many, many warnings... :(

From: Boris Gubenko (Boris.Gubenko_at_[hidden])
Date: 2007-05-02 13:06:25

Vladimir Prus wrote:
> We probably can use warnings-as-errors only for gcc and msvc, to reduce
> portability burden.

cxx (Tru64, VMS) has -msg_error option and aCC (HP-UX) has +We and
+Weargs options. They all increase message(s) severity to error.

When running boost tests on HP-UX/aCC6, I suppress 53 (!) warnings
(via CXXOPTS env. variable).

Thanks,
  Boris

----- Original Message ----- 
From: "Vladimir Prus" <ghost_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Wednesday, May 02, 2007 12:39 PM
Subject: Re: [boost] [1.34.0beta] many, many warnings... :(

> Stefan Seefeld wrote:
>
>>>>>> <pick your favorite color>.
>>>>> The problem is that the current regression reporting tools don't count
>>>>> warnings (previous version use to), so there's nothing
>>>>> nagging developers
>>>>> about warnings introduced in their code.
>>>> If you treat warnings as errors
>>>
>>> Fine we me, but not necessary fine with everybody ;-). And I suspect
>>> Boost.Build's warnings-as-errors=on work only with few compilers.
>>> But few is better than none.
>>
>> The problem is that some warnings can't be avoided, at least not if you
>> try to be portable.
>> (Example: As discussed in a different thread: putting in a return
>> statement to satisfy some compilers may trigger a 'unreachable code'
>> warning on others.)
>
> I think BOOST_AVOID_WARNING_XXX macro can be used to suppress a warning
> in compiler-specific way.
>
> We probably can use warnings-as-errors only for gcc and msvc, to reduce
> portability burden.
>
> - Volodya
>
>
> _______________________________________________
> Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost
>
>