boost.function with CW 8.2 on W2k

Bertolt Mildner

9 Sep 2002 9 Sep '02

12:21 p.m.

I get the following warnings when I use boost::function with CW 8.2 on W2k: Warning : variable / argument 'deprecated' is not used in function (included from: function10.hpp:28 function.hpp:37 Test.cpp:8) function_template.hpp line 388 BOOST_FUNCTION_DEPRECATED_INNER Warning : variable / argument 'deprecated' is not used in function function.hpp line 874 BOOST_FUNCTION_DEPRECATED_INNER Warning : variable / argument 'deprecated' is not used in function function.hpp line 867 BOOST_FUNCTION_DEPRECATED_INNER I'm using a relatively new CVS snapshot (<2 days old). -- Free Crypto Token - FCT www.FreeCryptoToken.org

Show replies by date

Douglas Gregor

9 Sep 9 Sep

12:47 p.m.

New subject: [Boost-Users] boost.function with CW 8.2 on W2k

On Monday 09 September 2002 08:21 am, Bertolt Mildner wrote:

...

Warning : variable / argument 'deprecated' is not used in function (included from: function10.hpp:28 function.hpp:37 Test.cpp:8) function_template.hpp line 388 BOOST_FUNCTION_DEPRECATED_INNER

The 'set' method is deprecated and will disappear some time after 1.29.0. Please use the assignment operator instead. Doug

Bertolt Mildner

4:55 p.m.

"Douglas Gregor" schrieb im Newsbeitrag news:200209090847.39513.gregod@cs.rpi.edu...

...

On Monday 09 September 2002 08:21 am, Bertolt Mildner wrote:

...
Warning : variable / argument 'deprecated' is not used in function (included from: function10.hpp:28 function.hpp:37 Test.cpp:8) function_template.hpp line 388 BOOST_FUNCTION_DEPRECATED_INNER

The 'set' method is deprecated and will disappear some time after 1.29.0. Please use the assignment operator instead.

Hmm... I'm not using a set method!? -- Free Crypto Token - FCT www.FreeCryptoToken.org

Douglas Gregor

10 Sep 10 Sep

3:19 a.m.

New subject: [Boost-Users] Re: boost.function with CW 8.2 on W2k

On Monday 09 September 2002 12:55 pm, Bertolt Mildner wrote:

...

"Douglas Gregor" schrieb im Newsbeitrag news:200209090847.39513.gregod@cs.rpi.edu...

...
On Monday 09 September 2002 08:21 am, Bertolt Mildner wrote:

...
Warning : variable / argument 'deprecated' is not used in function (included from: function10.hpp:28 function.hpp:37 Test.cpp:8) function_template.hpp line 388 BOOST_FUNCTION_DEPRECATED_INNER

The 'set' method is deprecated and will disappear some time after 1.29.0. Please use the assignment operator instead.

Hmm... I'm not using a set method!?

Now I'm completely confused... does code like the following cause a warning? (I don't have access to this compiler, unfortunately, but there aren't any warnings in the Boost regression tests...) template<typename T> struct foo { void bar() { int deprecated; } }; int main() { return sizeof(foo); } You may also compile with BOOST_FUNCTION_NO_DEPRECATED defined, which will completely remove all deprecated functionality. Doug

Bertolt Mildner

11:21 a.m.

[...]

...

Now I'm completely confused... does code like the following cause a warning? (I don't have access to this compiler, unfortunately, but there aren't any warnings in the Boost regression tests...)

template<typename T> struct foo { void bar() { int deprecated; } };

int main() { return sizeof(foo); }

At least with my compiler settings it will! Warning : variable / argument 'deprecated' is not used in function hello.cpp line 51 void bar() { int deprecated; } Error : illegal type hello.cpp line 74 return sizeof(foo);

...

You may also compile with BOOST_FUNCTION_NO_DEPRECATED defined, which will completely remove all deprecated functionality.

Thanks, that helped a lot! BTW: I also have the problem that I can't use Boost.Function together with precompiled headers because of the following error: Error : illegal use of precompiled header (included from: function.hpp:22 KeyStorage.h:12 HostMemoryKeyStorage.h:9 DesKey.h:11 MWPrefixWin32.pch++:14) function_base.hpp line 288 void nonnull() {}; I have tried to make nonnull() inline and/or static but that didn't help. I think the root of the problem is that the address of nonnull() is taken. I wonder if it is really needed to return a real (member) function pointer or if it would be OK to just return (safe_bool)1 ? Bertolt -- Free Crypto Token - FCT www.FreeCryptoToken.org

Bertolt Mildner

12:29 p.m.

[...]

...

I have tried to make nonnull() inline and/or static but that didn't help. I think the root of the problem is that the address of nonnull() is taken. I wonder if it is really needed to return a real (member) function pointer or if it would be OK to just return (safe_bool)1 ?

Don't know if this is useful but at least it makes my compiler more happy ... diff -r1.33 function_base.hpp 287,291c287 < struct dummy { < void nonnull() {}; < }; < < typedef void (dummy::*safe_bool)(); ---

...

typedef void (*safe_bool)();

295c291 < { return (this->empty())? 0 : &dummy::nonnull; } ---

...

{ return (this->empty())? 0 : (safe_bool)~0; }

298c294 < { return (this->empty())? &dummy::nonnull : 0; } ---

...

{ return (this->empty())? (safe_bool)~0 : 0; }

-- Free Crypto Token - FCT www.FreeCryptoToken.org

Douglas Gregor

1:49 p.m.

New subject: [Boost-Users] Re: Re: boost.function with CW 8.2 on W2k

On Tuesday 10 September 2002 08:29 am, Bertolt Mildner wrote:

...

[...]

...
I have tried to make nonnull() inline and/or static but that didn't help. I think the root of the problem is that the address of nonnull() is taken. I wonder if it is really needed to return a real (member) function pointer or if it would be OK to just return (safe_bool)1 ?

Don't know if this is useful but at least it makes my compiler more happy [snip]

...
{ return (this->empty())? 0 : (safe_bool)~0; } { return (this->empty())? (safe_bool)~0 : 0; }

Unfortunately, that invokes implementation-defined behavior, because (safe_bool)~0 may or may not be the null pointer. Perhaps if we made nonnull a function template it would make precompiled headers possible again... (I _really_ wish I had this compiler to try things on, but thanks for helping out!) Doug

Bertolt Mildner

2:56 p.m.

"Douglas Gregor" schrieb im Newsbeitrag news:200209100949.40416.gregod@cs.rpi.edu...

...

On Tuesday 10 September 2002 08:29 am, Bertolt Mildner wrote:

...
[...]

...
I have tried to make nonnull() inline and/or static but that didn't help. I think the root of the problem is that the address of nonnull() is taken. I wonder if it is really needed to return a real (member) function pointer or if it would be OK to just return (safe_bool)1 ?

Don't know if this is useful but at least it makes my compiler more happy [snip]

...
{ return (this->empty())? 0 : (safe_bool)~0; } { return (this->empty())? (safe_bool)~0 : 0; }

Unfortunately, that invokes implementation-defined behavior, because (safe_bool)~0 may or may not be the null pointer. [...]

Is really only a valid (member) function pointer guaranteed to be non-null? I think it is very strange that ~0 may not be non-null! I have tried this one: diff -r1.33 function_base.hpp 267a268,275

...

namespace detail { inline void nonnull() { } }

274c282 < bool empty() const { return !manager; } ---

...

inline bool empty() const { return !manager; }

284c292 < operator bool () const { return !this->empty(); } ---

...

inline operator bool () const { return !this->empty(); }

287,291c295 < struct dummy { < void nonnull() {}; < }; < < typedef void (dummy::*safe_bool)(); ---

...

typedef void (*safe_bool)();

294,295c298,299 < operator safe_bool () const < { return (this->empty())? 0 : &dummy::nonnull; } ---

...

inline operator safe_bool () const { return (this->empty())? 0 : &detail::nonnull; }

297,298c301,302 < safe_bool operator!() const < { return (this->empty())? &dummy::nonnull : 0; } ---

...

inline safe_bool operator!() const { return (this->empty())? &detail::nonnull : 0; }

*but*, now I get the following errors: Error : call of non-function (instantiating: 'boost::function1>>>, std::vector>, std::allocatorboost::function_base>::function1>>>, std::vector>, boost::shared_ptrFct::Crypto::Engine, std::allocatorboost::function_base>>(boost::function2>>>, std::vector>, boost::shared_ptrFct::Crypto::Engine, std::allocatorboost::function_base>)') (instantiating: 'boost::function1>>>, std::vector>, std::allocatorboost::function_base>::assign_to>>>, std::vector>, boost::shared_ptrFct::Crypto::Engine, std::allocatorboost::function_base>>(boost::function2>>>, std::vector>, boost::shared_ptrFct::Crypto::Engine, std::allocatorboost::function_base>)') (instantiating: 'boost::function1>>>, std::vector>, std::allocatorboost::function_base>::assign_to>>>, std::vector>, boost::shared_ptrFct::Crypto::Engine, std::allocatorboost::function_base>>(boost::function2>>>, std::vector>, boost::shared_ptrFct::Crypto::Engine, std::allocatorboost::function_base>, boost::detail::function::function_obj_tag)') (instantiating: 'boost::detail::function::function_obj_invoker1>>>, std::vector>, boost::shared_ptrFct::Crypto::Engine, std::allocatorboost::function_base>, boost::shared_ptr>>>, std::vector>>::invoke(boost::detail::function::any_pointer, std::vector>)') function_template.hpp line 110 return (*f)(BOOST_FUNCTION_ARGS); Error : call of non-function (instantiating: 'boost::detail::function::function_obj_invoker1>>>, std::vector>, boost::shared_ptrFct::Crypto::Engine, std::allocatorboost::function_base>, boost::shared_ptr>>>, std::vector>>::invoke(boost::detail::function::any_pointer, std::vector>)') function_template.hpp line 110 return (*f)(BOOST_FUNCTION_ARGS); Bertolt -- Free Crypto Token - FCT www.FreeCryptoToken.org

Douglas Gregor

3:09 p.m.

New subject: [Boost-Users] Re: Re: Re: boost.function with CW 8.2 on W2k

On Tuesday 10 September 2002 10:56 am, Bertolt Mildner wrote:

...

Is really only a valid (member) function pointer guaranteed to be non-null? I think it is very strange that ~0 may not be non-null!

Yes, it is strange, but it should be possible. For instance, this isn't guaranteed to give a null function pointer: int x = 1; return (safe_bool)(1-x); Only an integral constant expression that is '0', when converted to a pointer, is guaranteed to become the null pointer. You also can't rely on: T* p = 0; int p_int = (int)p; assert(p_int == 0); // p_int need not be zero! Except for the syntactic oddity that '0' can be converted to a null pointer, there need not be a relationship between an integer 0 and the null pointer :)

...

I have tried this one:

...
typedef void (*safe_bool)();

You don't want to do that :) safe_bool is a pointer-to-member-function so that it is _only_ useful in boolean contexts. Perhaps this would appease your compiler: struct dummy { template<typename T> void nonnull() {}; }; typedef void (dummy::*safe_bool)(); operator safe_bool () const { return (this->empty())? 0 : &dummy::nonnull<int>; } safe_bool operator!() const { return (!this->empty())? 0 : &dummy::nonnull<int>; } If it works, I'll put the workaround into CVS. Doug

Bertolt Mildner

3:35 p.m.

"Douglas Gregor" schrieb im Newsbeitrag news:200209101109.28788.gregod@cs.rpi.edu...

...

On Tuesday 10 September 2002 10:56 am, Bertolt Mildner wrote:

[...]

...

struct dummy { template<typename T> void nonnull() {}; };

typedef void (dummy::*safe_bool)();

operator safe_bool () const { return (this->empty())? 0 : &dummy::nonnull<int>; }

safe_bool operator!() const { return (!this->empty())? 0 : &dummy::nonnull<int>; }

If it works, I'll put the workaround into CVS.

Now I get: Error : ambiguous access to overloaded function (included from: function.hpp:22 KeyStorage.h:12 HostMemoryKeyStorage.h:9 DesKey.h:10 MWPrefixWin32.pch++:14) function_base.hpp line 298 { return (!this->empty())? 0 : &dummy::nonnull<int>; } -- Free Crypto Token - FCT www.FreeCryptoToken.org

Peter Dimov

3:57 p.m.

New subject: [Boost-Users] Re: Re: Re: boost.function with CW 8.2 on W2k

From: "Douglas Gregor"

...

safe_bool operator!() const { return (!this->empty())? 0 : &dummy::nonnull<int>; }

This is not related to the problem you are discussing, but operator! can (and probably should) return an ordinary bool.

Bertolt Mildner

4:20 p.m.

"Peter Dimov" schrieb im Newsbeitrag news:004301c258e2$cc559850$1d00a8c0@pdimov2...

...

From: "Douglas Gregor"

...
safe_bool operator!() const { return (!this->empty())? 0 : &dummy::nonnull<int>; }

This is not related to the problem you are discussing, but operator! can (and probably should) return an ordinary bool.

So bool operator!() const { return !this->empty(); } should be OK? ... two minutes later ... Arghalskfvbafg1q24584%&§/(%§&$%&?!?! CW now gives me the same error in operator safe_bool() ... -- Free Crypto Token - FCT www.FreeCryptoToken.org

Peter Dimov

4:28 p.m.

New subject: [Boost-Users] Re: Re: Re: Re: boost.function with CW 8.2 on W2k

From: "Bertolt Mildner"

...

...
...
"Peter Dimov" schrieb im Newsbeitrag news:004301c258e2$cc559850$1d00a8c0@pdimov2... From: "Douglas Gregor"

safe_bool operator!() const { return (!this->empty())? 0 : &dummy::nonnull<int>; }

This is not related to the problem you are discussing, but operator! can (and probably should) return an ordinary bool.

So bool operator!() const { return !this->empty(); } should be OK? ... two minutes later ... Arghalskfvbafg1q24584%&§/(%§&$%&?!?! CW now gives me the same error in operator safe_bool() ...<<< I'm curious... Do you get an error when you put boost/shared_ptr.hpp in a precompiled header?

Bertolt Mildner

4:42 p.m.

"Peter Dimov" schrieb im Newsbeitrag news:001801c258e7$27b15d20$1d00a8c0@pdimov2...

...

From: "Bertolt Mildner"

...
...
...
"Peter Dimov" schrieb im Newsbeitrag news:004301c258e2$cc559850$1d00a8c0@pdimov2... From: "Douglas Gregor"

[...]

...

...
CW now gives me the same error in operator safe_bool() ...<<<

I'm curious... Do you get an error when you put boost/shared_ptr.hpp in a precompiled header?

No, I had no problems with Boost in precompiled headers so far except with Boost.Function. I'm using shared_ptr, BOOST_STATIC_ASSERT, lexical_cast, Boost.Threads, pool allocator, iterator adapters, any, ... Bertolt -- Free Crypto Token - FCT www.FreeCryptoToken.org

Peter Dimov

4:45 p.m.

New subject: [Boost-Users] Re: Re: Re: Re: Re: boost.function with CW 8.2 on W2k

...

"Peter Dimov" schrieb im Newsbeitrag news:001801c258e7$27b15d20$1d00a8c0@pdimov2...

...
From: "Bertolt Mildner"

...
...
...
"Peter Dimov" schrieb im Newsbeitrag news:004301c258e2$cc559850$1d00a8c0@pdimov2... From: "Douglas Gregor"

[...]

...
...
CW now gives me the same error in operator safe_bool() ...<<<

I'm curious... Do you get an error when you put boost/shared_ptr.hpp in a precompiled header?

No, I had no problems with Boost in precompiled headers so far except with Boost.Function.

I'm using shared_ptr, BOOST_STATIC_ASSERT, lexical_cast, Boost.Threads,

From: "Bertolt Mildner" pool allocator, iterator adapters, any, ...< shared_ptr reuses shared_ptr::get as the nonnull value, without a dummy nested class: // implicit conversion to "bool" typedef T * (this_type::*unspecified_bool_type)() const; operator unspecified_bool_type() const // never throws { return px == 0? 0: &this_type::get; } Perhaps something along these lines will work for boost::function.

Bertolt Mildner

5:03 p.m.

...

shared_ptr reuses shared_ptr::get as the nonnull value, without a dummy nested class:

// implicit conversion to "bool"

typedef T * (this_type::*unspecified_bool_type)() const;

operator unspecified_bool_type() const // never throws { return px == 0? 0: &this_type::get; }

Perhaps something along these lines will work for boost::function.

Ok, here is my next try: diff -r1.33 function_base.hpp 271c271 < function_base() : manager(0), functor(static_cast(0)) {} ---

...

inline function_base() : manager(0), functor(static_cast(0)) {}

274c274 < bool empty() const { return !manager; } ---

...

inline bool empty() const { return !manager; }

284c284 < operator bool () const { return !this->empty(); } ---

...

inline operator bool () const { return !this->empty(); }

287,291c287 < struct dummy { < void nonnull() {}; < }; < < typedef void (dummy::*safe_bool)(); ---

...

typedef bool (function_base::*safe_bool)() const;

294,295c290,294 < operator safe_bool () const < { return (this->empty())? 0 : &dummy::nonnull; } ---

...

inline operator safe_bool () const { return (this->empty())? 0 : &function_base::empty; }

inline bool operator!() const { return !this->empty(); }

297,298d295 < safe_bool operator!() const < { return (this->empty())? &dummy::nonnull : 0; } which again results in this error: Error : illegal use of precompiled header (included from: function.hpp:22 KeyStorage.h:12 HostMemoryKeyStorage.h:9 DesKey.h:10 MWPrefixWin32.pch++:15) function_base.hpp line 274 inline bool empty() const { return !manager; } Bertolt -- Free Crypto Token - FCT www.FreeCryptoToken.org

Bertolt Mildner

12 Sep 12 Sep

8:37 a.m.

...

...
[...]

...
...
CW now gives me the same error in operator safe_bool() ...<<<

I'm curious... Do you get an error when you put boost/shared_ptr.hpp in a precompiled header?

No, I had no problems with Boost in precompiled headers so far except with Boost.Function.

I'm using shared_ptr, BOOST_STATIC_ASSERT, lexical_cast, Boost.Threads, pool allocator, iterator adapters, any, ...<

shared_ptr reuses shared_ptr::get as the nonnull value, without a dummy nested class:

// implicit conversion to "bool"

typedef T * (this_type::*unspecified_bool_type)() const;

operator unspecified_bool_type() const // never throws { return px == 0? 0: &this_type::get; }

Perhaps something along these lines will work for boost::function.

OK, I think the problem is that function_base is not a template. This should fix it: diff -r1.33 function_base.hpp 268c268,269 < class function_base ---

...

template <typename do_not_use = int> class function_base_impl 271c272 < function_base() : manager(0), functor(static_cast(0)) {}

...

function_base_impl() : manager(0), functor(static_cast(0)) {}

287,291c288 < struct dummy { < void nonnull() {}; < }; < < typedef void (dummy::*safe_bool)(); ---

...

typedef bool (function_base_impl<>::*safe_bool)() const;

295c292 < { return (this->empty())? 0 : &dummy::nonnull; } ---

...

{ return (this->empty())? 0 : &function_base_impl<>::empty; }

297,298c294,295 < safe_bool operator!() const < { return (this->empty())? &dummy::nonnull : 0; } ---

...

bool operator!() const { return this->empty(); }

300a298,299

...

typedef function_base_impl<> function_base;

Bertolt -- Free Crypto Token - FCT www.FreeCryptoToken.org

Douglas Gregor

10 Sep 10 Sep

4:33 p.m.

New subject: [Boost-Users] Re: Re: Re: Re: boost.function with CW 8.2 on W2k

On Tuesday 10 September 2002 12:20 pm, Bertolt Mildner wrote:

...

Arghalskfvbafg1q24584%&§/(%§&$%&?!?!

CW now gives me the same error in operator safe_bool() ...

\me ponders running out to buy MW 8.2 just to get this working... Yet one more idea (arghhhhh): template<typename T> struct dummy { void nonnull(); }; typedef void (dummy<int>::*safe_bool)(); operator safe_bool() const { return this->empty? 0 : &dummy<int>::nonnull; } bool operator!() const { return this->empty(); } Doug

Douglas Gregor

4:29 p.m.

New subject: [Boost-Users] Re: Re: Re: boost.function with CW 8.2 on W2k

On Tuesday 10 September 2002 11:57 am, Peter Dimov wrote:

...

From: "Douglas Gregor"

...
safe_bool operator!() const { return (!this->empty())? 0 : &dummy::nonnull<int>; }

This is not related to the problem you are discussing, but operator! can (and probably should) return an ordinary bool.

Yes, it should return a bool. Thanks, I'll change this. Doug

8112

Age (days ago)

8115

Last active (days ago)

List overview

Download

18 comments

3 participants

participants (3)

Bertolt Mildner
Douglas Gregor
Peter Dimov