c++0x pods and constructors

[dragan]

dragan wrote:
Can C++0x PODs/aggregates have user-defined constructors
and still be PODs/aggregates?
From [dcl.init.aggr]:
<< 1 An aggregate is an array or a class (Clause 9) with no
user-provided constructors (12.1), no private or
protected non-static data members (Clause 11), no base classes
(Clause 10), and no virtual functions (10.3). >>
So, no, once you add a user-defined c-tor, it's not an
aggregate.

I'm not sure what aggregates are good for.

Initialization syntax. The most important use is for arrays,
since you can let the compiler count the members when using
aggregate initialization, e.g.:

std::string const names[] = { "Abert", "Bertha", "Charles" ... } ;

Ah, OK. See, an example is so useful. When I see "aggregate", I will now
think "array" AND initialization.

It's also useful for structs, however, when you need static
initialization (especially for tables of structs), e.g.:

struct MapInit
{
char const* key;
int value;
};

MapInit const initTable[] =
{
{ "toto", 42 },
{ "titi", 0 },
// ...
};

I think "aggregate" gets thrown around so much, but it should probably be
reserved more for implementer-level discussion rather than user-level. It
seems so minor relative to the notion of POD (or layout compatibility!
Errgh, so frustrating!).

Layout compatibility is an awkward problem. C++ can mandate a
certain number of things with regards to how the C++ compiler
lays out data, but it can't mandate anything with regards to how
other languages layout data. In the end, it's a lot like the
``extern "C"'' fiasco: C++ requires a compiler to support it,
stating that functions declared ``extern "C"'' must be callable
from C (or something along those lines). Which is all nice and
fine, but what does it mean if the platform doesn't have a C
compiler? Or if it has several C compilers, with different
calling conventions?

"extern C" fiasco? News to me that it was a fiasco (you must be an
implementor/insider, I bet)! I thought it was more like "thank god we can
still use DLLs in C++".

In practice, of course, on most small and medium sized systems
today, the system ABI is defined in terms of C, which means that
there will be a C compiler, and all C compilers will use a
commun layout and common calling conventions,

That does seem to be "the least common denominator". Or at least "one of
them" things.

so the problem
doesn't come up; C is the lowest common denominator.

I really did write the above before I read you saying it (I feel proud of
myself! ).

If you
look at it closely, you'll see that the standard doesn't
actually give any guarantees with regards to standard layout and
other languages, for the reason stated above.

'Care to restate that reason?

There may be some
advanced programming techniques which depend on standard layout,
but for the everyday user, the only really significant
distinction is whether the class could be written in C or
not

You meant if a class could look like a struct to C, surely.

---if it could, it will be accessible from C, and probably
from most other languages as well
(on typical small and medium
general purpose computers

That covers a lot (the most of it?) of ground, and propably shouldn't be
flattened.

---I wouldn't count on it on
mainframes, nor for that matter on embedded systems); otherwise,
it won't be. (You may, of course, add non-virtual member
functions, including conversion operators, but not constructors,
destructors or assignment operators, without causing problems.)

I'm still hoping that Mr. Coffin was right about "convenience constructors"
being OK. (Else I'll have to research and find out why there cannot be such.
I'll blindly accept the concrete "special functions" as being special and
reserved, but I really would want to know why "convenience constructors" are
a no-go, if indeed so).

The other thing that is often important is whether the class is
an aggregate which supports static initialization. But in
practice, the two overlap; there are very few cases where you
can use static aggregate initialization but couldn't write the
class in C. (The presence of a pointer to member in the class
would be an example of one.)

I should have stopped responding to your post probably after I asked whether
you meant the current standard or the yet-to-be (draft) C++. All of a
sudden, I feel like I know NOTHING! Errrgh!! I think my house may explode if
I start up Visual C++ and _I_ use it! I saw no red sticker on the box.. I
got it for free as a download!

A conversion operator is just an ordinary member function.

A positive note for the post to end on! At least something is concrete!
(Until someone says a POD can't have member functions!)

(Aside: Are overloaded operators fast compared to functions? I don't know
why that sticks in my mind. Yeah I know, I could test it, but I think in the
past I have and have found that. I may be mistaken).

 

[dragan]

Only if it's empty itself.

OK. (Not really "OK", I just don't want to think about it now. Again, I
think a few more examples to explain C++'s rules would be great. I would
partition those into categories though: "you have to know this", "mostly of
concern for implementors", etc.)

The problem is that you can't express the requirements in terms
of can or cannot have.

Why is that? Is it so absolutely, or just because we are talking about C++?

At least partially, you have to express
them in terms of "must have": a POD must have a trivial default
constructor, for example (which is not the same thing as not
having a non-trivial default constructor---you have three
possibilities: no default constructor, trivial default
constructor and non-trivial default constructor).

But by "requirements", you mean "required by an implementation", yes?

 

[dragan]

Attention: the requirement isn't that the class not have a
non-trivial X; it is that it have a trivial X. Thus, for
example:

struct S { S( int ); int i; };

doesn't have a non-trivial default constructor, but it doesn't
have a trivial one either, and thus, is not a POD.

Is it still layout-compatible with C though? (I'm not sure of the
usefullness of "POD" anymore).

IIUC, in the
next version of the standard, you will be able to write:

struct S { S( int ); S() = default; int i; };

and it will have the required trivial default constructor.

And what will the compiler do with such a trivial default constructor? Why
does it need to enslave that thing?

But IIUC, only if you ensure that it also has a trivial default
constructor. Which it won't have be default if you've provided
any other constructor.

Unless you specify it as you did above, right? I think people are going to
forget it if the reasons a compiler needs to control "trivial default
constructor" are not known. But it seems like too much low-level things are
percolating up to the user level.

 

[Jerry Coffin]

[ ... ]

'Care to restate that reason?

The bottom line is pretty simple: something like 'extern "C"' is
really trying to give a guarantee about a C compiler, but the C++
standard doesn't really have any control over C compilers. You can
only accomplish something useful with 'extern "C"' by having a C
compiler that's written to cooperate with your C++ compiler -- but
the C++ standard can't really do anything about that, and for that
matter the C standard really can't either.

You meant if a class could look like a struct to C, surely.

I think what he's saying (in effect) is that the main time PODs
matter or get used, is when you have a header that you can compile as
either C or C++, and you want to have (at least some) assurance of
compatibility between the two.

[ ... ]

I'm still hoping that Mr. Coffin was right about "convenience
constructors" being OK. (Else I'll have to research and find out
why there cannot be such. I'll blindly accept the concrete "special
functions" as being special and reserved, but I really would want
to know why "convenience constructors" are a no-go, if indeed so).

Yes -- to be a POD, the default ctor, copy ctor and dtor must all be
trivial. A trivial ctor must be generated by the compiler -- any user
defined ctor is non-trivial. Even if it's carefully written to do
exactly what a compiler-generated ctor would have done, the fact that
you wrote it still means it's non-trivial.

In C++ 98/03, if you define any ctor, that prevents the compiler from
generating a default ctor. If you want a default ctor, you need to
define one, and by definition, when/if you do so, it won't be a
trivial ctor. Likewise with the dtor, copy ctor and copy assignment
operator.

In C++ 0x, however, there's an "=default" syntax that allows you to
tell the compiler to generate a default ctor, copy ctor, copy
assignment operator, and/or dtor, even if you _have_ defined some
other ctor. For example:

class X {
int data;
public:
X(int val) : data(val) {}
X() = default;
X(x const &) = default;
X &operator=(X const &other) = default;
~X() = default;
};

Now, we have a user-defined ctor that takes a value of type int that
is used to initialize the data member -- but we also have explicitly
defaulted the definitions of the other special member functions, so
they remain "trivial" and the type is still a POD.

One other strange wrinkle: it's possible to create a non-trivial
special member function using the '= default;' syntax. To be trivial,
the function must (among other things) be inline, and the
'=default;' must be on the _declaration_ of the function. If (for
whatever strange reason) you choose to, you can do something like:

class Y {
// ...
public:
Y();
};

Y::Y() = default;

You've now explicitly defaulted the default ctor, BUT the '=default'
isn't on the initial declaration, and your definition is _not_ an
inline function, so it's no longer trivial.

[ ... ]

(Aside: Are overloaded operators fast compared to functions? I
don't know why that sticks in my mind. Yeah I know, I could test
it, but I think in the past I have and have found that. I may be
mistaken).

An overloaded operator _is_ a function -- it just has a rather
different name than most other functions. Overloaded operators are
typically quite short functions, so there's a good chance that the
compiler can generate inline code. Nonetheless, the fact that it's an
overloaded operator does not (in itself, at least with any compiler
I've ever seen or heard of) make any difference in whether the code
will be generated inline. The code generation depends on the size and
complexity of the function, not the name it's given.

 

[James Kanze]

I think "aggregate" gets thrown around so much, but it should
probably be reserved more for implementer-level discussion
rather than user-level. It seems so minor relative to the
notion of POD (or layout compatibility! Errgh, so
frustrating!).

Yes and no. For the user, aggregate initialization can be an
advantage in certain cases, e.g. my examples of letting the
compiler calculate array size. For the user, there are really
three things which interesting:

1. Aggregate initialization: useful for letting the compiler
calculate the size of an array, and in general for static
data.

2. Static initialization, for avoiding order of initialization
issues. This requires trivial constructors and constant
expressions in the initializer; for compound types (struct's
and arrays), the type must be an aggregate, and aggregate
initialization must be used. (Otherwise, you must have a
constructor somewhere, and the type ceases to be trivial.)

3. PODs or layout compatibility: mainly important with regards
to communicating with other languages, especially C. (The
system API is generally defined in terms of C these days.)

"extern C" fiasco? News to me that it was a fiasco (you must
be an implementor/insider, I bet)! I thought it was more like
"thank god we can still use DLLs in C++".

Fiasco may be too strong a word, but formally, the standard
guarantees much less than it seems to. (I certainly doesn't
guarantee anything with regards to DLLs, for example.) All that
it says (with regards to C) is ``Every implementation shall
provide for linkage to functions written in the C programming
language, "C", and linkage to C + + functions, "C++".'' Except
that it doesn't specify how these "functions written in the C
programming language" are to be compiled or linked; an
implementation may require you to compile them using a special C
compiler, which generates code incompatible with the system API,
for example. And it doesn't address the issue of what to do if
there is no C compiler for the system.

At least some of the people involved with the standardization
proceedure are aware of this weakness, although the committee
doesn't seem to have wanted to address it. I think the general
intent is more along the lines that IF there is at least one C
implementation available on the platform, then the compiler must
support linkage with one of them, and document which one, and
what is necessary to make it work.

The probable reason the committee didn't address the problem is
that it isn't one in practice. Practically all platforms that
support C++ have a standard C ABI, which is adhered to by all C
compilers, and practically all C++ compilers come with a C
compiler, and require no special steps to link C and C++. So
the standard may not say exactly what was wanted, but in
practice, everything works as was desired.

Until the day someone decides that C is dead enough that they
don't need a C compiler for their platform.

[...]

'Care to restate that reason?

The C++ standard can't impose anything on other languages, only
on C++.

You meant if a class could look like a struct to C, surely.

More or less, yes. With the addition of non-virtual member
functions, but that's about it.

That covers a lot (the most of it?) of ground, and propably
shouldn't be flattened.

Typical small and medium general purpose computers represent but
a small percentage of all computers. I'd guess off hand that
something like 90% of all computers are embedded systems. At
the other end, and this is the place where I suspect most of the
problems come, are the mainframes. IBM System z doesn't define
everything in terms of C, at least under the traditional OS's.
(You can also get it with Linux.) Cobol is still the
predominant language there.

I'm still hoping that Mr. Coffin was right about "convenience
constructors" being OK. (Else I'll have to research and find
out why there cannot be such. I'll blindly accept the
concrete "special functions" as being special and reserved,
but I really would want to know why "convenience constructors"
are a no-go, if indeed so).

You can't add "convenience constructors" today, and I'm sure
that Jerry didn't say that. The next version of the standard
will allow it, I think, but only if you explicitly say you want
the "trivial" default constructor. And while initialization
syntax has been somewhat unified, I still don't think you can
get all of the features of aggregate initialization (especially
static initialization) if you provide any constructor; at most,
you'll be guaranteed a C compatible layout (but that's typically
the case today, with most compilers).

I should have stopped responding to your post probably after I
asked whether you meant the current standard or the yet-to-be
(draft) C++. All of a sudden, I feel like I know NOTHING!
Errrgh!! I think my house may explode if I start up Visual C++
and _I_ use it! I saw no red sticker on the box.. I got it for
free as a download!

Unless I explicitly state otherwise, I'm talking about the
current standard. There are no implementations of the next
standard (can't be, since it doesn't exist yet), and the way
things are going, I'm beginning to wonder if there will be any
in my lifetime.

With regards to Visual C++, it's a concrete implementation;
concrete implementations always give you a lot more guarantees
than the standard gives.

As long as you're not concerned with portability, you can use
all of the guarantees the implementation gives. Once you're
concerned about portability, especially about potential
portability to systems you don't know yet, it becomes trickier.

A positive note for the post to end on! At least something is
concrete! (Until someone says a POD can't have member
functions!)

(Aside: Are overloaded operators fast compared to functions? I
don't know why that sticks in my mind. Yeah I know, I could
test it, but I think in the past I have and have found that. I
may be mistaken).

An overloaded operator is just an ordinary function with a funny
name. The only time they might impact performance is when they
are used instead of a built-in operator.

 

[James Kanze]

[ ... ]

'Care to restate that reason?

The bottom line is pretty simple: something like 'extern "C"'
is really trying to give a guarantee about a C compiler, but
the C++ standard doesn't really have any control over C
compilers. You can only accomplish something useful with
'extern "C"' by having a C compiler that's written to
cooperate with your C++ compiler -- but the C++ standard can't
really do anything about that, and for that matter the C
standard really can't either.

It's not quite that bad; you don't need active cooperation with
the C compiler. You do need a C compiler, however, and you need
to know what it does and generates in order to make ``extern
"C"'' work in C++.

I think what he's saying (in effect) is that the main time
PODs matter or get used, is when you have a header that you
can compile as either C or C++, and you want to have (at least
some) assurance of compatibility between the two.

Exactly. Otherwise, you're not generally too concerned with
POD-ness in itself, only certain properties of it (allows
agglomerate initialization---but a lot of non-PODs do that---or
static initialization).

One interesting idea that I've never seen done would be to
conditionally add convenience functions in the header, something
like:

struct Toto
{
// ...
#ifdef __cplusplus
void someFunction() ;
#endif
};

In practice, I'm sure that this will work, but formally...

[ ... ]
In C++ 0x, however, there's an "=default" syntax that allows
you to tell the compiler to generate a default ctor, copy
ctor, copy assignment operator, and/or dtor, even if you
_have_ defined some other ctor. For example:

class X {
int data;
public:
X(int val) : data(val) {}
X() = default;
X(x const &) = default;
X &operator=(X const &other) = default;
~X() = default;
};

Now, we have a user-defined ctor that takes a value of type
int that is used to initialize the data member -- but we also
have explicitly defaulted the definitions of the other special
member functions, so they remain "trivial" and the type is
still a POD.

Just curious, but what effect does this have on static
initialization using the agglomerate syntax. Can I still write
something like:

static X const table[] = { 1, 2, 3 };

and be guaranteed that the initialization precedes all code that
I've written (including code in constructors of static objects)?

 

[James Kanze]

James Kanze wrote: [...]

The problem is that you can't express the requirements in
terms of can or cannot have.

Why is that? Is it so absolutely, or just because we are
talking about C++?

It's because the actual requirements contain several "must
have", not can have or cannot have.

But by "requirements", you mean "required by an
implementation", yes?

No. I mean the requirements on a class for it to be a POD.

 

[James Kanze]

[...]

Is it still layout-compatible with C though?

In practice, with most compilers, yes.

The current standard doesn't talk about "layout-compatible", and
no C++ standard can impose anything on the C compiler. I didn't
see anything in N2914 (the latest draft that I happen to have
handy) that guaranteed that a "standard-layout" class is in
anyway compatible with C.

On the other hand, the whole concept of "layout-compatible"
seems to have been lifted from the C standard, and from a QoI
point of view, I would expect that anything with
"standard-layout" be compatible with C if the system defines a
standard C ABI (as most do).

(I'm not sure of the usefullness of "POD" anymore).

Per se, they don't have much use. They do restrict the
implementation in some ways, however, that typically allows them
to be used in mixed language headers (headers which are used by
both C and C++).

And what will the compiler do with such a trivial default
constructor?

Nothing. Otherwise, the constructor wouldn't be trivial.

Why does it need to enslave that thing?

I'm not sure what you're trying to ask there, but if it is "why
do we need a special syntax for this?", it's because the current
standard says that if a class has any constructor, the compiler
won't provide a default constructor implicitly, and a lot of
code counts on this.

 

[Jerry Coffin]

(e-mail address removed)>, (e-mail address removed)
says...

[ ... ]

It's not quite that bad; you don't need active cooperation with
the C compiler. You do need a C compiler, however, and you need
to know what it does and generates in order to make ``extern
"C"'' work in C++.

I was thinking primarily of the fact that anytime the C compiler
changes anything about its ABI, the C++ compiler needs to know about
it and take steps to compensate. Fortunately, at least in most cases
the ABI is fairly stable, so this doesn't arise a lot in practice.

[ ... ]

One interesting idea that I've never seen done would be to
conditionally add convenience functions in the header, something
like:

struct Toto
{
// ...
#ifdef __cplusplus
void someFunction() ;
#endif
};

In practice, I'm sure that this will work, but formally...

In practice, it's likely to work at least as long as none of the
functions is virtual. A virtual function would stand a good chance of
causing breakage though...

[ ... ]

Just curious, but what effect does this have on static
initialization using the agglomerate syntax. Can I still write
something like:

static X const table[] = { 1, 2, 3 };

and be guaranteed that the initialization precedes all code that
I've written (including code in constructors of static objects)?

I haven't studied that in a lot of detail yet, but it looks like
there's a fair bit of new material to study. First of all, they've
added thread local storage, and associated initialization rules for
it. Then they've added constexpr's, which allow static initialization
in quite a few cases that it wouldn't work in C++ 98/03.

Those have caused enough changes to the text that I haven't yet
sorted out whether the underlying rules have changed a lot, or
whether it's mostly just different text describing mostly similar
rules with a couple of new possibilities thrown into the mix.

 

[Brian Wood]

I think "aggregate" gets thrown around so much, but it should
probably be reserved more for implementer-level discussion
rather than user-level. It seems so minor relative to the
notion of POD (or layout compatibility!  Errgh, so
frustrating!).

Yes and no.  For the user, aggregate initialization can be an
advantage in certain cases, e.g. my examples of letting the
compiler calculate array size.  For the user, there are really
three things which interesting:

 1. Aggregate initialization: useful for letting the compiler
    calculate the size of an array, and in general for static
    data.

 2. Static initialization, for avoiding order of initialization
    issues.  This requires trivial constructors and constant
    expressions in the initializer; for compound types (struct's
    and arrays), the type must be an aggregate, and aggregate
    initialization must be used.  (Otherwise, you must have a
    constructor somewhere, and the type ceases to be trivial.)

 3. PODs or layout compatibility: mainly important with regards
    to communicating with other languages, especially C.  (The
    system API is generally defined in terms of C these days.)

"extern C" fiasco? News to me that it was a fiasco (you must
be an implementor/insider, I bet)! I thought it was more like
"thank god we can still use DLLs in C++".

Fiasco may be too strong a word, but formally, the standard
guarantees much less than it seems to.  (I certainly doesn't
guarantee anything with regards to DLLs, for example.)  All that
it says (with regards to C) is ``Every implementation shall
provide for linkage to functions written in the C programming
language, "C", and linkage to C + + functions, "C++".''  Except
that it doesn't specify how these "functions written in the C
programming language" are to be compiled or linked; an
implementation may require you to compile them using a special C
compiler, which generates code incompatible with the system API,
for example.  And it doesn't address the issue of what to do if
there is no C compiler for the system.

At least some of the people involved with the standardization
proceedure are aware of this weakness, although the committee
doesn't seem to have wanted to address it.  I think the general
intent is more along the lines that IF there is at least one C
implementation available on the platform, then the compiler must
support linkage with one of them, and document which one, and
what is necessary to make it work.

The probable reason the committee didn't address the problem is
that it isn't one in practice.  Practically all platforms that
support C++ have a standard C ABI, which is adhered to by all C
compilers, and practically all C++ compilers come with a C
compiler, and require no special steps to link C and C++.  So
the standard may not say exactly what was wanted, but in
practice, everything works as was desired.

Until the day someone decides that C is dead enough that they
don't need a C compiler for their platform.

    [...]

'Care to restate that reason?

The C++ standard can't impose anything on other languages, only
on C++.

You meant if a class could look like a struct to C, surely.

More or less, yes.  With the addition of non-virtual member
functions, but that's about it.

That covers a lot (the most of it?) of ground, and propably
shouldn't be flattened.

Typical small and medium general purpose computers represent but
a small percentage of all computers.  I'd guess off hand that
something like 90% of all computers are embedded systems.  At
the other end, and this is the place where I suspect most of the
problems come, are the mainframes.  IBM System z doesn't define
everything in terms of C, at least under the traditional OS's.
(You can also get it with Linux.)  Cobol is still the
predominant language there.

I'm still hoping that Mr. Coffin was right about "convenience
constructors" being OK. (Else I'll have to research and find
out why there cannot be such.  I'll blindly accept the
concrete "special functions" as being special and reserved,
but I really would want to know why "convenience constructors"
are a no-go, if indeed so).

You can't add "convenience constructors" today, and I'm sure
that Jerry didn't say that.  The next version of the standard
will allow it, I think, but only if you explicitly say you want
the "trivial" default constructor.  And while initialization
syntax has been somewhat unified, I still don't think you can
get all of the features of aggregate initialization (especially
static initialization) if you provide any constructor; at most,
you'll be guaranteed a C compatible layout (but that's typically
the case today, with most compilers).

I should have stopped responding to your post probably after I
asked whether you meant the current standard or the yet-to-be
(draft) C++. All of a sudden, I feel like I know NOTHING!
Errrgh!! I think my house may explode if I start up Visual C++
and _I_ use it! I saw no red sticker on the box.. I got it for
free as a download!

Unless I explicitly state otherwise, I'm talking about the
current standard.  There are no implementations of the next
standard (can't be, since it doesn't exist yet), and the way
things are going, I'm beginning to wonder if there will be any
in my lifetime.

Perhaps what's needed are on line compilers that output
fully instantiated code which can be handled by
"existing" C++ compilers. By that I mean that some
changes to existing compilers would be needed, but the
changes would be relatively minor. That might help in
terms of getting the new version of the language out
and in use. Without doing something radically
different, I think the number of C++0x compilers is
likely to be less than the number of C++ compilers
today. I hope you live to see C++0x in general use,
but at the very least, I think things will be
interesting (curioser and curioser) from here on in.


Brian Wood
http://www.webEbenezer.net

 

[dragan]

[ ... ]

'Care to restate that reason?

The bottom line is pretty simple: something like 'extern "C"' is
really trying to give a guarantee about a C compiler, but the C++
standard doesn't really have any control over C compilers. You can
only accomplish something useful with 'extern "C"' by having a C
compiler that's written to cooperate with your C++ compiler -- but
the C++ standard can't really do anything about that, and for that
matter the C standard really can't either.

Whatever happened with projects to bring forth a standard C++ ABI. Itanium
only?

I think what he's saying (in effect) is that the main time PODs
matter or get used, is when you have a header that you can compile as
either C or C++, and you want to have (at least some) assurance of
compatibility between the two.

I would think that cross-platform interoperability is a larger concern. At
least with a C ("POD") struct, you stand a chance at it looking the same in
more than one machine. Not hardly so for classes (I mean with one using all
the C++ "fancy" things).

[ ... ]

I'm still hoping that Mr. Coffin was right about "convenience
constructors" being OK. (Else I'll have to research and find out
why there cannot be such. I'll blindly accept the concrete "special
functions" as being special and reserved, but I really would want
to know why "convenience constructors" are a no-go, if indeed so).

Yes -- to be a POD, the default ctor, copy ctor and dtor must all be
trivial. A trivial ctor must be generated by the compiler -- any user
defined ctor is non-trivial. Even if it's carefully written to do
exactly what a compiler-generated ctor would have done, the fact that
you wrote it still means it's non-trivial.

I understand all that and am accepting of it.

In C++ 98/03, if you define any ctor, that prevents the compiler from
generating a default ctor.
If you want a default ctor, you need to
define one, and by definition, when/if you do so, it won't be a
trivial ctor. Likewise with the dtor, copy ctor and copy assignment
operator.

I've probably encountered that, but I didn't "know" that. So are you
retracting your statement that "convenience constructors" are allowed?

In C++ 0x, however, there's an "=default" syntax that allows you to
tell the compiler to generate a default ctor, copy ctor, copy
assignment operator, and/or dtor, even if you _have_ defined some
other ctor. For example:

class X {
int data;
public:
X(int val) : data(val) {}
X() = default;
X(x const &) = default;
X &operator=(X const &other) = default;
~X() = default;
};

OK, things are looking rosie then for C++0x. A good enough reason to drop
the current lineup ASAP, IMO. Hopefully the C++0x goodies will start
trickling into compilers even before "ratification" or whatever they call
it.

Now, we have a user-defined ctor that takes a value of type int that
is used to initialize the data member -- but we also have explicitly
defaulted the definitions of the other special member functions, so
they remain "trivial" and the type is still a POD.

And that makes me a happy camper, cuz not allowing "convenience
constructors" is a pet peave of mine. I saw/see no reason why it has to be
without them, so now I know indeed there was no reason other than oversight.

One other strange wrinkle: it's possible to create a non-trivial
special member function using the '= default;' syntax. To be trivial,
the function must (among other things) be inline, and the
'=default;' must be on the _declaration_ of the function. If (for
whatever strange reason) you choose to, you can do something like:

class Y {
// ...
public:
Y();
};

Y::Y() = default;

You've now explicitly defaulted the default ctor, BUT the '=default'
isn't on the initial declaration, and your definition is _not_ an
inline function, so it's no longer trivial.

I would have assumed that the out-of-line Y::Y() would not allow "=default"!
The question becomes: Why does it?! Hehe, so Scott Meyers can right more
"Effective C++" books! (Great series by the way, but if it ever becomes
contrived... oh nevermind).

[ ... ]

(Aside: Are overloaded operators fast compared to functions? I
don't know why that sticks in my mind. Yeah I know, I could test
it, but I think in the past I have and have found that. I may be
mistaken).

An overloaded operator _is_ a function -- it just has a rather
different name than most other functions. Overloaded operators are
typically quite short functions, so there's a good chance that the
compiler can generate inline code. Nonetheless, the fact that it's an
overloaded operator does not (in itself, at least with any compiler
I've ever seen or heard of) make any difference in whether the code
will be generated inline. The code generation depends on the size and
complexity of the function, not the name it's given.

Alright. Maybe long ago when that stuck in my mind I had experienced the
inline vs. online thing. But I get most of my information from books (and
lately online) and "operators are FAAAST!" sticks in my mind, so I think I
may have read it somewhere. It makes perfect sense to me that operators are
no different than functions except for the obvious syntax. Oh well, could
have been a brainfart.

 

[dragan]

Yes and no. For the user, aggregate initialization can be an
advantage in certain cases, e.g. my examples of letting the
compiler calculate array size. For the user, there are really
three things which interesting:

1. Aggregate initialization: useful for letting the compiler
calculate the size of an array, and in general for static
data.

2. Static initialization, for avoiding order of initialization
issues. This requires trivial constructors and constant
expressions in the initializer; for compound types (struct's
and arrays), the type must be an aggregate, and aggregate
initialization must be used. (Otherwise, you must have a
constructor somewhere, and the type ceases to be trivial.)

3. PODs or layout compatibility: mainly important with regards
to communicating with other languages, especially C. (The
system API is generally defined in terms of C these days.)

What about cross-platform interoperability? Write a struct here, read it
there?

Fiasco may be too strong a word, but formally, the standard
guarantees much less than it seems to. (I certainly doesn't
guarantee anything with regards to DLLs, for example.)

I does indirectly. "C-linkage" (is that the right terminology?) is all that
can be exported from DLLs ("sorry" to be so Windows-ish). I guess you
probably could get all that name-mangled stuff to come out of a DLL on a
single machine. You won't be selling that app to other people though, that's
for sure.

All that
it says (with regards to C) is ``Every implementation shall
provide for linkage to functions written in the C programming
language, "C", and linkage to C + + functions, "C++".'' Except
that it doesn't specify how these "functions written in the C
programming language" are to be compiled or linked; an
implementation may require you to compile them using a special C
compiler, which generates code incompatible with the system API,
for example. And it doesn't address the issue of what to do if
there is no C compiler for the system.

Are you being too hypothetical? I don't give it a second thought: I KNOW
that I'll always be able to call "extern C" functions in a DLL (after
LoadLibrary(), blah, blah.)

At least some of the people involved with the standardization
proceedure are aware of this weakness, although the committee
doesn't seem to have wanted to address it. I think the general
intent is more along the lines that IF there is at least one C
implementation available on the platform, then the compiler must
support linkage with one of them, and document which one, and
what is necessary to make it work.

I seem to have missed "the weakness". I'm sure there is one if you say so.
Anyway, I feel like "extern C" and "layout" are separate. It's all good
information though.

The probable reason the committee didn't address the problem is
that it isn't one in practice.

YES! Thanks for letting me know I'm not "out of it" (a bit insecure maybe,
"out of it", no way!).

Practically all platforms that
support C++ have a standard C ABI,

I didn't know that. Well maybe I sorta did because I rely on it ("extern C",
"POD"). So there is a formality like the C++ ABI for Itanium for C?

which is adhered to by all C
compilers, and practically all C++ compilers come with a C
compiler, and require no special steps to link C and C++. So
the standard may not say exactly what was wanted, but in
practice, everything works as was desired.

Interesting. Big kudos for C. But it is much easier in C to make a std ABI
than in C++. C++ should have specified one from the beginning, but too late
for that now. Live and learn.

Until the day someone decides that C is dead enough that they
don't need a C compiler for their platform.

C/Std ABI: 1 point. C++/No Std ABI: -1 point. C++ is keeping C alive!

[...]

'Care to restate that reason?

The C++ standard can't impose anything on other languages, only
on C++.
k.

You meant if a class could look like a struct to C, surely.

More or less, yes. With the addition of non-virtual member
functions, but that's about it.
k.

That covers a lot (the most of it?) of ground, and propably
shouldn't be flattened.

Typical small and medium general purpose computers represent but
a small percentage of all computers.

And I know that too. I think in terms of the desktop/server space mostly as
those are the applications I am building. I wasn't thinking "what pie piece
of all computers", but rather "the desktop/server space is billions and
billions of dollars of space" (with apologies to Carl Sagan).

I'd guess off hand that
something like 90% of all computers are embedded systems. At
the other end, and this is the place where I suspect most of the
problems come, are the mainframes. IBM System z doesn't define
everything in terms of C, at least under the traditional OS's.
(You can also get it with Linux.) Cobol is still the
predominant language there.

I would wager that a poll (eww.. not one of THOSE!) would show that most
readers of this ng are in the client-server (includes all things web etc.
too) space. That is, working on software that interacts with humans (rather
than controlling anti-lock brakes, etc). (Or I am a fish out of water?).

You can't add "convenience constructors" today, and I'm sure
that Jerry didn't say that.

YOU said "today". The thread is all about C++0x. Indeed he did say that, but
not with your "today" qualifier. I'm 99.9% sure that he is correct too.

The next version of the standard
will allow it, I think, but only if you explicitly say you want
the "trivial" default constructor.

That's what he restated, but the thread was about C++0x, so you are sort of
"preaching to the choir".

And while initialization
syntax has been somewhat unified, I still don't think you can
get all of the features of aggregate initialization (especially
static initialization) if you provide any constructor; at most,
you'll be guaranteed a C compatible layout (but that's typically
the case today, with most compilers).

I assume that too today: that I can write "convenience constructors" and not
have a worry. I just don't do it because it's not official.

Unless I explicitly state otherwise, I'm talking about the
current standard.

Well, and again, the thread is specifically about C++0x. Did you miss that
key aspect?

There are no implementations of the next
standard (can't be, since it doesn't exist yet), and the way
things are going, I'm beginning to wonder if there will be any
in my lifetime.

Now THAT is an interesting tangent. I was thinking that C++0x features (at
least a few stocking-stuffers) would be in my favorite compiler for Xmas!

With regards to Visual C++, it's a concrete implementation;
concrete implementations always give you a lot more guarantees
than the standard gives.

What do you mean by "concrete implementation"? Did you just mean one
specific implementation? As opposed to worrying about all compilers or the
majority or average of them across all platforms and domains? I just have
some Windows apps to write.

As long as you're not concerned with portability, you can use
all of the guarantees the implementation gives. Once you're
concerned about portability, especially about potential
portability to systems you don't know yet, it becomes trickier.

I understand that. Hopefully that "portable everywhere" hype doesn't keep
"good people down". Which brings up another interesting (to me) tangent:
that C++ doesn't compete with other languages as much as it is in
competition with platform-specific/domain-specific solutions.

An overloaded operator is just an ordinary function with a funny
name. The only time they might impact performance is when they
are used instead of a built-in operator.

I believe it.

 

[dragan]

James Kanze wrote: [...]

I downloaded N2960. I like to think in terms of the things
a POD can or cannot have rather than the lingo like:
trivially copyable, trivial class, standard layout, and on
and on. In another post I wrote my understanding of what is
and isn't allowed.
The problem is that you can't express the requirements in
terms of can or cannot have.

Why is that? Is it so absolutely, or just because we are
talking about C++?

It's because the actual requirements contain several "must
have", not can have or cannot have.

While very abstract this subthread portion seems, you seem to be saying:
"yes, you're right, so you're wrong". Let me clarify: I said I can think in
simple terms but not or can't in C++ because C++ has to be like a
president/politician walking the fence trying to be all things to all people
and I live in Greenbough Alabama and just need a new ping pong paddle. In
short, I'd MUCH rather have a language tool that I could compare against my
checkoffs (can/cannot haves) rather than have to spend oodles of time
deciphering terms like "trivial constructor" and the complex interplay
between all the other 1250 pages of the standard. I was just explaining what
I said, but suddenly, I feel catharted!

No. I mean the requirements on a class for it to be a POD.

k. I'm not sure from where in your other post I quoted "requirements".

 

[dragan]

(e-mail address removed) says...

[...]

Attention: the requirement isn't that the class not have a
non-trivial X; it is that it have a trivial X. Thus, for
example:
struct S { S( int ); int i; };
doesn't have a non-trivial default constructor, but it
doesn't have a trivial one either, and thus, is not a POD.

Is it still layout-compatible with C though?

In practice, with most compilers, yes.

I've been way to meek about using the guarantees given by my chosen
compiler. More and more, I find myself at MSDN as my source for C++
information. What I find there is guaranteed to work because that's the
compiler I'm using. C++ isn't losing ground, it's lost the war? Has most
value in the Smithsonian?

The current standard doesn't talk about "layout-compatible",

The current standard is off-topic in this thread.

and
no C++ standard can impose anything on the C compiler. I didn't
see anything in N2914 (the latest draft that I happen to have
handy) that guaranteed that a "standard-layout" class is in
anyway compatible with C.

Hmm. I thought that was at least part of the point behind "standard layout".
But, from a purely terminological point of view, it would seem to be ripe
for bandaiding the lack of standard C++ ABI. A stepping stone for the
original oversight.

On the other hand, the whole concept of "layout-compatible"
seems to have been lifted from the C standard, and from a QoI
point of view, I would expect that anything with
"standard-layout" be compatible with C if the system defines a
standard C ABI (as most do).

Any guarantee is better than no guarantee. Even Master C++er Bjarne
Stroupstrup acknowledged this point-of-quality (though somewhat
condescendingly) in a presentation he gave (available on UTube).

Per se, they don't have much use.

I'd say (assertively) that they do, but I'm still struggling with the
definition of such a thing.

They do restrict the
implementation in some ways, however, that typically allows them
to be used in mixed language headers (headers which are used by
both C and C++).



Nothing. Otherwise, the constructor wouldn't be trivial.


I'm not sure what you're trying to ask there, but if it is "why
do we need a special syntax for this?", it's because the current
standard says that if a class has any constructor, the compiler
won't provide a default constructor implicitly, and a lot of
code counts on this.

I was just wondering why the compiler needed control of that special
function. What book did I not buy or decide not to buy because I don't think
that as an application programmer I need to worry about stuff that my
employer calls "get up to speed on your technical skills on your own time so
you can bring direct/real value to the company via purposeful software
applications"?

 

[James Kanze]

(e-mail address removed)>, (e-mail address removed)
says...

[ ... ]

It's not quite that bad; you don't need active cooperation
with the C compiler. You do need a C compiler, however, and
you need to know what it does and generates in order to make
``extern "C"'' work in C++.

I was thinking primarily of the fact that anytime the C
compiler changes anything about its ABI, the C++ compiler
needs to know about it and take steps to compensate.
Fortunately, at least in most cases the ABI is fairly stable,
so this doesn't arise a lot in practice.

Or that there are two different C compilers on the system, and
they use different ABI's. Here, too, the case should be fairly
rare, due to the fact that most OS's today define their
interface in terms of C.

[ ... ]

One interesting idea that I've never seen done would be to
conditionally add convenience functions in the header, something
like:
struct Toto
{
// ...
#ifdef __cplusplus
void someFunction() ;
#endif
};
In practice, I'm sure that this will work, but formally...

In practice, it's likely to work at least as long as none of
the functions is virtual. A virtual function would stand a
good chance of causing breakage though...

It would almost certainly break something.

[ ... ]

Just curious, but what effect does this have on static
initialization using the agglomerate syntax. Can I still
write something like:
static X const table[] = { 1, 2, 3 };
and be guaranteed that the initialization precedes all code
that I've written (including code in constructors of static
objects)?

I haven't studied that in a lot of detail yet, but it looks
like there's a fair bit of new material to study.

Yes. And for various reasons, I've not been able to study much
in the last year or so,

 

[James Kanze]

[...]

What about cross-platform interoperability? Write a struct
here, read it there?

That generally doesn't work, regardless. After all, it doesn't
work for int; how could it work for a struct which contains an
int?

[...]

I does indirectly. "C-linkage" (is that the right
terminology?) is all that can be exported from DLLs ("sorry"
to be so Windows-ish).

That's not true. I've used DLL's without any C linkage on at
least three platforms: Solaris, Linux and Windows.

I guess you probably could get all that name-mangled stuff to
come out of a DLL on a single machine. You won't be selling
that app to other people though, that's for sure.

The only time name mangling is relevant is when the user
explicitly loads a DLL and requests the function name (dlsym
under Unix). In such cases, the usual solution is to have a
single factory function as entry point, and declare that (and
only that) as extern "C".

Of course, you can only load DLL's which are binary compatible.
Which means a lot of different things, on different platforms.

Are you being too hypothetical? I don't give it a second
thought: I KNOW that I'll always be able to call "extern C"
functions in a DLL (after LoadLibrary(), blah, blah.)

If you're talking about the standard, you have to consider all
possible cases. I explicitly stated elsewhere that this isn't a
problem in practice.

Of course, if you're calling LoadLibrary, then you're 100%
Windows anyway, so you can count on the guarantees Microsoft
gives as well as those in the standard.

I seem to have missed "the weakness". I'm sure there is one if
you say so. Anyway, I feel like "extern C" and "layout" are
separate. It's all good information though.

The weakness is that for a C++ compiler to compile something
which can be called from C, it needs a C compiler and some
collaboration on the part of the C compiler. And the C++
standard has no influence over C compilers.

And extern "C" and layout are pretty orthogonal; depending on
what you're doing, you might need both extern "C" and some sort
of layout compatibility, however.

YES! Thanks for letting me know I'm not "out of it" (a bit
insecure maybe, "out of it", no way!).

I didn't know that. Well maybe I sorta did because I rely on
it ("extern C", "POD"). So there is a formality like the C++
ABI for Itanium for C?

It depends on the platform, but for at least Posix and Windows,
there has to be, since the system ABI is defined in terms of C.
In practice, given the simplicity of C, most of the layout
issues depend directly on the hardware---I don't need to read
the compiler specifications for an IBM mainframe, for example,
to know the layout, since a compiler isn't going to introduce
unnecessary padding (it can, but it won't), and the hardware
determines where the padding, etc., is necessary.

C++ is more complicated, since you have to consider things like
how vtables are layed out. And for both C and C++, the calling
conventions can vary. I don't know the situation on modern
mainframes, but a long time ago, I know that different C
compilers for the Siemens BS2000 used different calling
conventions.

Interesting. Big kudos for C. But it is much easier in C to
make a std ABI than in C++. C++ should have specified one from
the beginning, but too late for that now. Live and learn.

The language specification cannot specify a standard ABI. That
has to be hardware specific. Historically, C++ became
significant after most hardware architectures had been
specified, and there were several different C++ compilers around
before C++ became important enough. The Itanium is the rare
exception of an architecture which was defined after C++ was
important, but before any compilers existed for it.

C/Std ABI: 1 point. C++/No Std ABI: -1 point. C++ is keeping C
alive!

History is keeping C alive.

[...]
Unless I explicitly state otherwise, I'm talking about the
current standard.

Well, and again, the thread is specifically about C++0x. Did
you miss that key aspect?

Yes. Generally speaking, this newsgroup is about programming in
C++; there is a separate group, comp.std.c++ for discussions of
the standard. And while the separation isn't absolute, you
can't program in C++0x today, so most discussion of it probably
would be more appropriate in csc++.

Now THAT is an interesting tangent. I was thinking that C++0x
features (at least a few stocking-stuffers) would be in my
favorite compiler for Xmas!

That is, of course, completely impossible, since the standard
still hasn't been approved, and in fact, we still don't know
exactly what will be in it. Generally, it takes a few years
after the finalization of a standard for implementations to
catch up; for that matter, most compilers today don't implement
all of C++03.

What do you mean by "concrete implementation"?

That it's real. It exists. It's not just an abstract
specification.

Did you just mean one specific implementation?

Yes. (I suppose that technically, it's several implementations,
since you can control some of the aspects, like the signedness
of a plain char, from the command line.)

As opposed to worrying about all compilers or the majority or
average of them across all platforms and domains? I just have
some Windows apps to write.

If all you're targetting is Windows, then all you need to know
is Visual C++.

 

[James Kanze]

James Kanze wrote:

[...]

Attention: the requirement isn't that the class not have a
non-trivial X; it is that it have a trivial X. Thus, for
example:
struct S { S( int ); int i; };
doesn't have a non-trivial default constructor, but it
doesn't have a trivial one either, and thus, is not a POD.
Is it still layout-compatible with C though?

In practice, with most compilers, yes.

I've been way to meek about using the guarantees given by my
chosen compiler. More and more, I find myself at MSDN as my
source for C++ information. What I find there is guaranteed to
work because that's the compiler I'm using. C++ isn't losing
ground, it's lost the war? Has most value in the Smithsonian?

I don't follow you. One of the reasons (not the only one) C++
continues to be important is because it leaves significant
liberty to the implementation. It can be implemented on just
about every platform.

The current standard is off-topic in this thread.

Some abstract future standard is arguably off-topic in this
newsgroup. At least, there's a more appropriate news group for
it: comp.std.c++.

Hmm. I thought that was at least part of the point behind
"standard layout".

Sort of. I suspect that the intent behind "standard-layout"
does have to do with C compatibility. But the standard can't
impose that, because compatibility is a two way street, and the
C++ standard cannot impose anything on C. In practice, I think
the intent is clear, and if an implementation provides both a C
and a C++ compiler (the usual case when C++ is present), then I
would expect, from a QoI point of view, that a standard-layout
class be compatible with C, provided that it only uses elements
of C (i.e. no pointers to members).

But, from a purely terminological point of view, it would seem
to be ripe for bandaiding the lack of standard C++ ABI. A
stepping stone for the original oversight.

It would be nice if some architectures did define a standard C++
ABI. But it's up to the architectures; the C++ standard can't
do it.

Any guarantee is better than no guarantee. Even Master C++er
Bjarne Stroupstrup acknowledged this point-of-quality (though
somewhat condescendingly) in a presentation he gave (available
on UTube).

A standard can't guarantee a usable implementation. It's never
more than part of what you count on.

I'd say (assertively) that they do, but I'm still struggling
with the definition of such a thing.

Per se, limited strictly to the guaranteed in the standard, they
don't. What they do is...

So in practice, they end up having a utility beyond that
specified in the standard.

[...]

I was just wondering why the compiler needed control of that
special function.

I'm not too sure what you mean by "control of that special
function". In C, it's possible to write something like:

struct S { int a; int b; };
S s;

and it was desired that this be possible in C++ as well, for
reasons of C compatibility, if nothing else. This means that
the compiler must implicitly declare and define a default
constructor, if the user doesn't. It was also felt that if the
user did define a constructor, it should be impossible to bypass
it. This gives you the rules of the present standard.

 

[dragan]

James Kanze wrote:

[...]
Attention: the requirement isn't that the class not have a
non-trivial X; it is that it have a trivial X. Thus, for
example:
struct S { S( int ); int i; };
doesn't have a non-trivial default constructor, but it
doesn't have a trivial one either, and thus, is not a POD.
Is it still layout-compatible with C though?
In practice, with most compilers, yes.

I've been way to meek about using the guarantees given by my
chosen compiler. More and more, I find myself at MSDN as my
source for C++ information. What I find there is guaranteed to
work because that's the compiler I'm using. C++ isn't losing
ground, it's lost the war? Has most value in the Smithsonian?

I don't follow you. One of the reasons (not the only one) C++
continues to be important is because it leaves significant
liberty to the implementation. It can be implemented on just
about every platform.

That is not important.

Some abstract future standard is arguably off-topic in this
newsgroup. At least, there's a more appropriate news group for
it: comp.std.c++.

That's your OPINION. (Or your attempt at imposition).

Sort of. I suspect that the intent behind "standard-layout"
does have to do with C compatibility. But the standard can't
impose that, because compatibility is a two way street, and the
C++ standard cannot impose anything on C. In practice, I think
the intent is clear, and if an implementation provides both a C
and a C++ compiler (the usual case when C++ is present), then I
would expect, from a QoI point of view, that a standard-layout
class be compatible with C, provided that it only uses elements
of C (i.e. no pointers to members).

You said: C++ is dead.

It would be nice if some architectures did define a standard C++
ABI. But it's up to the architectures; the C++ standard can't
do it.

Hello. Bottleneck imposing itself .

A standard can't guarantee a usable implementation. It's never
more than part of what you count on.

"If you're stupid, you vote".

Per se, limited strictly to the guaranteed in the standard, they
don't. What they do is...

What do they do? Hmm?

So in practice, they end up having a utility beyond that
specified in the standard.

[...]

I was just wondering why the compiler needed control of that
special function.

I'm not too sure what you mean by "control of that special
function". In C, it's possible to write something like:

struct S { int a; int b; };
S s;

and it was desired that this be possible in C++ as well, for
reasons of C compatibility, if nothing else. This means that
the compiler must implicitly declare and define a default
constructor, if the user doesn't. It was also felt that if the
user did define a constructor, it should be impossible to bypass
it. This gives you the rules of the present standard.

I don't believe you. I live in Romania and know your weak presentations.

 

[James Kanze]

James Kanze wrote:
James Kanze wrote:

[...]

I don't follow you. One of the reasons (not the only one)
C++ continues to be important is because it leaves
significant liberty to the implementation. It can be
implemented on just about every platform.

That is not important.

The fact that C++ can be implemented on just about every
platform is important to the majority of the committee. Some
people don't care about anything but Windows, but they're far
from a majority.

That's your OPINION. (Or your attempt at imposition).

It's not an absolute, and some discussion of the future standard
is certainly acceptable (since the future standard also concerns
C++). But in general, people who want to discuss purely
standard issues (and anything to do with the future standard is
purely a standard issue, since it is pretty much irrelevant to
programming today) do so in the group which is specialized for
it: comp.std.c++.

You said: C++ is dead.

When did I ever say that? Just the contrary, C++ is a lot more
alive than many people want.

Hello. Bottleneck imposing itself .

In what sense? It's just common sense: a standard designed to
be implemented on many different platforms can't impose anything
which has to be specific to a given platform.

"If you're stupid, you vote".

You're not making sense.

What do they do? Hmm?

What I explained immediately following the "Per se, they don't
have much use." Which still immediately follows.

They do restrict the implementation in some ways, however,
that typically allows them to be used in mixed language
headers (headers which are used by both C and C++).

So in practice, they end up having a utility beyond that
specified in the standard.
[...]

I'm not sure what you're trying to ask there, but if it is
"why do we need a special syntax for this?", it's because
the current standard says that if a class has any
constructor, the compiler won't provide a default
constructor implicitly, and a lot of code counts on this.
I was just wondering why the compiler needed control of that
special function.

I'm not too sure what you mean by "control of that special
function". In C, it's possible to write something like:
struct S { int a; int b; };
S s;
and it was desired that this be possible in C++ as well, for
reasons of C compatibility, if nothing else. This means
that the compiler must implicitly declare and define a
default constructor, if the user doesn't. It was also felt
that if the user did define a constructor, it should be
impossible to bypass it. This gives you the rules of the
present standard.

I don't believe you.

So read Stroustrup's D&E. Those are exactly the reasons
(somewhat summarized) he gives.

I live in Romania and know your weak presentations.

What does living in Romania have to do with anything? Except
maybe that English isn't your first language, so you
misunderstand some of the finer points, or misexpress what you
are trying to say---but overall, your English seems quite good?
And I don't know what you mean by "weak presentations", either.
I'm just presenting the facts as I know them. There's no effort
on my part to make a particular "presentation".

 

[dragan]

James Kanze wrote:
James Kanze wrote:

[...]
I don't follow you. One of the reasons (not the only one)
C++ continues to be important is because it leaves
significant liberty to the implementation. It can be
implemented on just about every platform.

That is not important.

The fact that C++ can be implemented on just about every
platform is important to the majority of the committee. Some
people don't care about anything but Windows, but they're far
from a majority.

"Oopsies" then, I must be autistic: I thought the world revolved around
Windows! I mean, it's EVERYwhere: desktop, server, mobile, embedded. What
more is there? Is anything more ubiquitous than Windows? Why bother with
anything else? Learn one thing and be done with it, right? So what if every
stoplight has QNX controlling it? Soon it will be Windows Embedded.

It's not an absolute, and some discussion of the future standard
is certainly acceptable (since the future standard also concerns
C++). But in general, people who want to discuss purely
standard issues (and anything to do with the future standard is
purely a standard issue, since it is pretty much irrelevant to
programming today) do so in the group which is specialized for
it: comp.std.c++.

Easy solution for those who don't want to discuss such things: change the
channel (don't view the thread)! Right?

When did I ever say that? Just the contrary, C++ is a lot more
alive than many people want.

In looking to forget it's lineage but still purporting something like that.
You can't have it both ways. And when is this next standard going to be real
anyway? It needs to happen EVERY year or "yer (C++) outta here!". IMHO, of
course.

In what sense? It's just common sense: a standard designed to
be implemented on many different platforms can't impose anything
which has to be specific to a given platform.

"Bottlenecks" are those things that cue other things behind them because
they move too slowly: they impede progresss. Parallel paths are prudent and
C++ needs to be "forked" or something (like in open source development): the
backwards-compatible path (which will die off sooner than you think once the
other paths open) and the path going forward. I said that quite well,
actually, if I do say so myself. Kudos to MEEEE!

You're not making sense.

It was a spoof on "designed by comittee". I think "the comittee" needs
direction.

What I explained immediately following the "Per se, they don't
have much use." Which still immediately follows.

And that's why the comittee and standard are undergoing so much revision on
the issue? Maybe I just know what I read and it is not current about what's
going on with the holy standard, but I do know that I read similar
"sentiments" to mine that make your quip "they don't have much use", well,
hogwash.

They do restrict the implementation in some ways, however,
that typically allows them to be used in mixed language
headers (headers which are used by both C and C++).
So in practice, they end up having a utility beyond that
specified in the standard.
[...]
I'm not sure what you're trying to ask there, but if it is
"why do we need a special syntax for this?", it's because
the current standard says that if a class has any
constructor, the compiler won't provide a default
constructor implicitly, and a lot of code counts on this.
I was just wondering why the compiler needed control of that
special function.
I'm not too sure what you mean by "control of that special
function". In C, it's possible to write something like:
struct S { int a; int b; };
S s;
and it was desired that this be possible in C++ as well, for
reasons of C compatibility, if nothing else. This means
that the compiler must implicitly declare and define a
default constructor, if the user doesn't. It was also felt
that if the user did define a constructor, it should be
impossible to bypass it. This gives you the rules of the
present standard.

I don't believe you.

So read Stroustrup's D&E. Those are exactly the reasons
(somewhat summarized) he gives.

I was just maybe getting tired after your previous hogwash statement... "I
don't believe you" follows directly: try to fool me once, shame on you. (Try
to fool my twice, shame on me).

What does living in Romania have to do with anything?

I don't really live there. I'm from there.

Except
maybe that English isn't your first language,

Not hardly, but I think I do pretty well with it because I read other
people's posts and can't decipher them for go to the bank!

so you
misunderstand some of the finer points, or misexpress what you
are trying to say---but overall, your English seems quite good?

I speak better in English than what I learned as an elephant. I find English
quite less demanding: know "rolling rrrrr's" and such. Spelling is pretty
hard though, but I on purpose don't use a spell checker because I don't want
people to think I am someone who I am not.

And I don't know what you mean by "weak presentations", either.
I'm just presenting the facts as I know them. There's no effort
on my part to make a particular "presentation".

I have been following your posts that are in a topic that I find interesting
and you seem to have a pattern. That pattern makes me think that you are
"defender of C++" or "guaranteeing your job". ?