Re: [Boost-users] [format] warning...warning...warning

Subject: Re: [Boost-users] [format] warning...warning...warning
From: Steven Watanabe (watanabesj_at_[hidden])
Date: 2017-03-23 15:43:55

AMDG

On 03/23/2017 02:45 AM, Arnaud RICHARD via Boost-users wrote:
> Hello,
> I'm a complete Boost newbie. I've been told very often that boost is the way to go for serious C++ coding, and I was not surprised to see such emphasis on the home page:
> "We aim to establish "existing practice" and provide reference implementations so that Boost libraries are suitable for eventual standardization"
> 
> So when I used the format library on my existing (non-trivial) project and first compiled... I was very very surprised to get my neat compilation console flooded with hundreds of warnings !
> 
> I acknowledge I activate many many warnings:
> CXXFLAGS+=-Wall -Wextra -Wcast-qual -Wctor-dtor-privacy -Wdisabled-optimization \
>           -Wformat=2 -Winit-self -Wlogical-op -Wmissing-include-dirs \
>           -Wnoexcept -Woverloaded-virtual -Wredundant-decls -Wshadow \
>           -Wsign-conversion -Wsign-promo -Wstrict-null-sentinel -Wstrict-overflow=5 \
>           -Wswitch-default -Wundef -Wunused
> 
> The warnings were as trivial as (for example):
> ../ext/boost/boost/type_traits/is_default_constructible.hpp:16:22: warning: "BOOST_GCC_VERSION_WORKAROUND_GUARD" is not defined [-Wundef]
> #if BOOST_WORKAROUND(BOOST_GCC_VERSION, < 40700)
> 

  If you compile with higher-than-default warning settings,
it's simplest to include Boost with -isystem instead of -I.
- At high warning settings, most additional warnings are spurious.
  Fixing them is a lot of effort for little gain.
- Every compiler warns about different issues.  Just because
  the code compiles without warnings on one compiler doesn't
  mean that other compilers will be equally happy.  There have
  even been cases in the past where compiler A warned if
  you did X, while compiler B warned if you /didn't/ do X.
- Sometimes "fixing" spurious warnings can create new bugs.

With that being said, I personally try to explicitly suppress
(with #pragma's) any known warnings that I don't care about.

> ../ext/boost/boost/format/parsing.hpp:435:67: warning: conversion to 'std::vector<boost::io::detail::format_item<char, std::char_traits<char>, std::allocator<char> >, std::allocator<boost::io::detail::format_item<char, std::char_traits<char>, std::allocator<char> > > >::size_type {aka unsigned int}' from 'int' may change the sign of the result [-Wsign-conversion]
>              string_type & piece = (cur_item==0) ? prefix_ : items_[cur_item-1].appendix_;
> 
> I understand they are only warning and the code is functionally correct and standard-wise correct.
> Personally when I face such false warnings in my code, I change the code to remove the warning so that if a warning does make sense, I can notice it.
> 
> I thought a reference implementation would be written with such policy.

You really shouldn't expect even those who are pedantic about
warnings to enable more than -Wall -Wextra.

> There's no way I'm gonna read each of the hundreds of messages to understand if it is a mistake in my source code or a "shortcut" in boost library.
> 
> So this is my first user experience. Not impressed at all. Boost will not be my reference.
> 

In Christ,
Steven Watanabe