Attn: Maintainer of boost/bind.hpp
In boost/bind.hpp there is injection of a boost namespace contents into the global namespace: ```c++ #include <boost/bind/bind.hpp> #ifndef BOOST_BIND_NO_PLACEHOLDERS #if defined(BOOST_CLANG) # pragma clang diagnostic push # if __has_warning("-Wheader-hygiene") # pragma clang diagnostic ignored "-Wheader-hygiene" # endif #endif using namespace boost::placeholders; #if defined(BOOST_CLANG) # pragma clang diagnostic pop #endif #endif // #ifndef BOOST_BIND_NO_PLACEHOLDERS ``` The fact that Boost actually disables the relevant compiler warning makes this particularly insidious. No modern C++ library should ever do "using namespace" into the global namespace in its public headers. Please fix this in the next release of Boost. Niall
Niall Douglas wrote:
In boost/bind.hpp there is injection of a boost namespace contents into the global namespace:
Including <boost/bind/bind.hpp> doesn't bring the placeholders into the global namespace; if you don't want them there, just include <boost/bind/bind.hpp>. <boost/bind.hpp> does for backward compatibility.
Please fix this in the next release of Boost.
Breaking code without any warning is not a good practice, but I'll mark <boost/bind.hpp> as deprecated to allow people time to migrate.
In boost/bind.hpp there is injection of a boost namespace contents into the global namespace:
Including <boost/bind/bind.hpp> doesn't bring the placeholders into the global namespace; if you don't want them there, just include <boost/bind/bind.hpp>. <boost/bind.hpp> does for backward compatibility.
Unfortunately, some of the Boost libraries include <boost/bind.hpp>, and thus pollute the global namespace.
Please fix this in the next release of Boost.
Breaking code without any warning is not a good practice, but I'll mark <boost/bind.hpp> as deprecated to allow people time to migrate.
Personally speaking, I think the macro `BOOST_BIND_NO_PLACEHOLDERS` needs to become `BOOST_BIND_ENABLE_GLOBAL_PLACEHOLDERS`. People can still use <boost/bind.hpp>, but they'll need to explicitly opt into the old behaviour. (This will also encourage the maintainers of those libraries depending on global placeholders to fix their libraries) Niall
Niall Douglas wrote:
Breaking code without any warning is not a good practice, but I'll mark <boost/bind.hpp> as deprecated to allow people time to migrate.
Personally speaking, I think the macro `BOOST_BIND_NO_PLACEHOLDERS` needs to become `BOOST_BIND_ENABLE_GLOBAL_PLACEHOLDERS`. People can still use <boost/bind.hpp>, but they'll need to explicitly opt into the old behaviour.
That's true in principle, and will be my next step, at some point. But what's also true is that people don't appreciate their code - that has worked for decades - to be broken. Just grep for <boost/bind.hpp> within Boost, and extrapolate.
On 13/12/2019 17:58, Peter Dimov via Boost wrote:
Niall Douglas wrote:
Breaking code without any warning is not a good practice, but I'll mark > <boost/bind.hpp> as deprecated to allow people time to migrate.
Personally speaking, I think the macro `BOOST_BIND_NO_PLACEHOLDERS` needs to become `BOOST_BIND_ENABLE_GLOBAL_PLACEHOLDERS`. People can still use <boost/bind.hpp>, but they'll need to explicitly opt into the old behaviour.
That's true in principle, and will be my next step, at some point.
But what's also true is that people don't appreciate their code - that has worked for decades - to be broken. Just grep for <boost/bind.hpp> within Boost, and extrapolate.
Ok, how about this: 1. Both `BOOST_BIND_NO_PLACEHOLDERS` and `BOOST_BIND_ENABLE_GLOBAL_PLACEHOLDERS` have the effect on placeholders injection. 2. The deprecation message says something like "In Boost v1.XX, the macro BOOST_BIND_NO_PLACEHOLDERS shall cease to have effect, and the macro BOOST_BIND_ENABLE_GLOBAL_PLACEHOLDERS will default to undefined. This will cause ::placeholders to no longer be injected into the global namespace by default. Please stop using <boost/bind.hpp>, or define BOOST_BIND_ENABLE_GLOBAL_PLACEHOLDERS to suppress this message.". How about Boost v1.76 as the switch over point? That's a year's notice. Niall
participants (2)
-
Niall Douglas -
Peter Dimov