Date: Sat, 29 Jul 2017 21:27:06 +0200 From: Tijl Coosemans <tijl@FreeBSD.org> To: Mark Millard <markmi@dsl-only.net> Cc: Dimitry Andric <dim@FreeBSD.org>, toolchain@FreeBSD.org Subject: Re: [package - head-amd64-default][games/simutrans] Failed for simutrans-120.2.2 in build Message-ID: <20170729212706.33841c4c@kalimero.tijl.coosemans.org> In-Reply-To: <D780E949-FB27-4A84-8E10-75D1E7A8EA06@dsl-only.net> References: <201707272142.v6RLg1G4099900@beefy12.nyi.freebsd.org> <20170728135510.2c6de57f@kalimero.tijl.coosemans.org> <F47E0976-759A-45A0-8421-8FD4402A9980@FreeBSD.org> <20170729015914.184c2660@kalimero.tijl.coosemans.org> <D780E949-FB27-4A84-8E10-75D1E7A8EA06@dsl-only.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 29 Jul 2017 00:34:39 -0700 Mark Millard <markmi@dsl-only.net> wrote:
> On 2017-Jul-28, at 4:59 PM, Tijl Coosemans <tijl at FreeBSD.org> wrote:
>> On Fri, 28 Jul 2017 19:54:04 +0200 Dimitry Andric <dim at FreeBSD.org> wrote:
>>> On 28 Jul 2017, at 13:55, Tijl Coosemans <tijl@freebsd.org> wrote:
>>>>
>>>> On Thu, 27 Jul 2017 21:42:01 +0000 pkg-fallout@FreeBSD.org wrote:
>>> ...
>>>>> In file included from squirrel/squirrel/sqvm.cc:5:
>>>>> In file included from /usr/include/c++/v1/math.h:310:
>>>>> /usr/include/c++/v1/limits:149:85: error: expected expression
>>>>> _LIBCPP_INLINE_VISIBILITY static _LIBCPP_CONSTEXPR type max() _NOEXCEPT {return type();}
>>>>> ^
>>>>> squirrel/squirrel/sqobject.h:131:24: note: expanded from macro 'type'
>>>>> #define type(obj) ((obj)._type)
>>>>> ^
>>>>
>>>> Simutrans code defines 'type' as a macro. Shouldn't libc++ headers use
>>>> _type or __type or something?
>>>
>>> No, the member name 'type' is used in many classes in the C++ standard
>>> library, for example all the traits in <type_traits>. Programs should
>>> not attempt to redefine this, at least not as a macro.
>>>
>>> Note that this also doesn't work with libstdc++, e.g.:
>>>
>>> $ cat boom.cpp
>>> #define type "nope, this will not work"
>>> #include <type_traits>
>>>
>>> and then:
>>>
>>> $ g++ -c boom.cpp
>>> boom.cpp:1:14: error: expected unqualified-id before string constant
>>> #define type "nope, this will not work"
>>> ^
>>> boom.cpp:1:14: error: expected class-name before string constant
>>> #define type "nope, this will not work"
>>> ^
>>> boom.cpp:1:14: error: expected '{' before string constant
>>> boom.cpp:1:14: error: expected class-name before string constant
>>> #define type "nope, this will not work"
>>> ^
>>> boom.cpp:1:14: error: expected '{' before string constant
>>> boom.cpp:1:14: error: expected class-name before string constant
>>> #define type "nope, this will not work"
>>> ^
>>> boom.cpp:1:14: error: expected '{' before string constant
>>> boom.cpp:1:14: error: expected class-name before string constant
>>> #define type "nope, this will not work"
>>> ^
>>> boom.cpp:1:14: error: expected '{' before string constant
>>> boom.cpp:1:14: error: expected unqualified-id before string constant
>>> #define type "nope, this will not work"
>>> ^
>>> In file included from boom.cpp:3:0:
>>> /usr/local/lib/gcc6/include/c++/type_traits:212:60: error: template argument 1 is invalid
>>> : public __is_void_helper<typename remove_cv<_Tp>::type>::type
>>> ^
>>> /usr/local/lib/gcc6/include/c++/type_traits:212:61: error: expected '{' before '::' token
>>> : public __is_void_helper<typename remove_cv<_Tp>::type>::type
>>> ^~
>>> [...and lots more errors like this...]
>>
>> The code does not include <type_traits> or any of that C++11 stuff. It
>> includes <math.h>. This works with libstdc++ because it doesn't have
>> <math.h>, but it would also work when <cmath> was included, because
>> libstdc++ uses __type everywhere (and __enable_if and __is_arithmetic,
>> etc. where libc++ headers use enable_if and is_arithmetic). The
>> libstdc++ way makes more sense. You cannot expect C++98 code to know
>> about reserved identifiers in C++11 or C++11 code to know about reserved
>> identifiers in later standards.
>
> I'll first note that Annex D D.5 C standard library
> headers says:
>
> "the C++ standard library provides the 25 C headers,
> as shown in table 154"
>
> and table 154 lists: <math.h> . That is relevant
> for the below.
>
> ISO/IEC 14882:2011(E) 17.6.4.3.1 Macro Names
> says:
>
> "A translation unit that include a standard library
> header shall not #define or #undef names declared
> in any standard library header."
>
> I'll note that the standard has sections with titles
> like "Type names", "Class names", "Nested type names",
> "Names of template specializations", and "Predefined
> macro names". My understanding is that the earlier
> quote spans avoiding matching all such names.
>
>
>
> ISO/IEC 14882:2011(E) mandates such things as:
>
> template <bool, class T> struct is_arithmetic;
> . . .
> template <bool, class T = void> struct enable_if;
> . . .
> template <class T, T v> typedef integral_constant {
> . . .
> typedef integral_constant<T,v> type;
> . . .
> };
But none of this should be exposed to C++98 code. These names were not
reserved in the C++98 standard so C++98 code is free to use them. If
libc++ cannot compile such valid C++98 code it is simply not compliant
with that standard. Note that in this case we were lucky to see a
diagnostic. C++98 code may use these names in a way that doesn't cause
an error. Who's going to review our 27000 ports to make sure they are
still compiled correctly?
> For targeting -std=c++11 or later in compiles
> __enable_if and __is_arithemtic and __type
> would be wrong in these places and require
> code using the standard to use the names
> that have the __ prefixes, in violation of
> the standard's specifications. That includes
> having no explicit -std= but depending on a
> default that happens to end up with c++11 or
> later as the version to target.
Of course things like __enable_if are for internal use only. In C++11
mode enable_if needs to be made available.
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20170729212706.33841c4c>