MS compiler bug?

[Martijn]

Hi,

The following code-fragment won't compile in the .Net-studio but it does
compile in Borland. So is anyone familiar with this problem, or someone
knows how to solve it?

void MidpointSolver::Solve(float* next_state, const float* current_state,
float time, float h, const unsigned int dim)
{
float derivative_state[dim];
...
}

The reported error is: C2057 - 'expected constant expression'. So the
compiler doesnt see 'dim' as an constant or something like that?!?

Anyone?

[John Harrison]

"Martijn" <r.m.maatman_@_student.utwente.nl> wrote in message
news:c6**********@netlx020.civ.utwente.nl...

Hi,

The following code-fragment won't compile in the .Net-studio but it does
compile in Borland. So is anyone familiar with this problem, or someone
knows how to solve it?

void MidpointSolver::Solve(float* next_state, const float* current_state,
float time, float h, const unsigned int dim)
{
float derivative_state[dim];
...
}

The reported error is: C2057 - 'expected constant expression'. So the
compiler doesnt see 'dim' as an constant or something like that?!?

Anyone?

No .NET is right. In this case constant expression means compile time
constant, which obviously dim isn't even though you've declared it as const.

The best way to do this is to use a vector.

#include <vector>

void MidpointSolver::Solve(float* next_state, const float* current_state,
float time, float h, const unsigned int dim)
{
std::vector<float> derivative_state(dim);
...

Now I can't see the rest of your code, but I wouldn't be surprised if it
compiled unchanged, even though you've switched from an array to a vector.

john

[Ioannis Vranos]

"Martijn" <r.m.maatman_@_student.utwente.nl> wrote in message
news:c6**********@netlx020.civ.utwente.nl...

Hi,

The following code-fragment won't compile in the .Net-studio but it does
compile in Borland. So is anyone familiar with this problem, or someone
knows how to solve it?

void MidpointSolver::Solve(float* next_state, const float* current_state,
float time, float h, const unsigned int dim)
{
float derivative_state[dim];
...
}

The reported error is: C2057 - 'expected constant expression'. So the
compiler doesnt see 'dim' as an constant or something like that?!?

It is not a constant expression, since it is not assigned a value at the
place of definition. It is as if you define a variabled-sized built in
array.

If you define dim inside the function scope then you will have no problem.
E.g.:

void MidpointSolver::Solve(float* next_state, const float* current_state,
float time, float h, const unsigned int dim)
{
const unsigned SIZE=7;

float derivative_state[SIZE];
...
}


Ioannis Vranos

[Chris Theis]

"Martijn" <r.m.maatman_@_student.utwente.nl> wrote in message
news:c6**********@netlx020.civ.utwente.nl...

Hi,

The following code-fragment won't compile in the .Net-studio but it does
compile in Borland. So is anyone familiar with this problem, or someone
knows how to solve it?

void MidpointSolver::Solve(float* next_state, const float* current_state,
float time, float h, const unsigned int dim)
{
float derivative_state[dim];
...
}

The reported error is: C2057 - 'expected constant expression'. So the
compiler doesnt see 'dim' as an constant or something like that?!?

Anyone?

The reported error is correct because dim is not a constant as the compiler
cannot see its actual value (which should be a defined constant). However,
it´s hard to believe that Borland will compile this without at least a
warning. Try to increase your warning levels and recompile under Borland
(BTW which version?).

HTH
Chris

[Andre Kostur]

"Chris Theis" <Ch*************@nospam.cern.ch> wrote in
news:c6**********@sunnews.cern.ch:

"Martijn" <r.m.maatman_@_student.utwente.nl> wrote in message
news:c6**********@netlx020.civ.utwente.nl...

Hi,

The following code-fragment won't compile in the .Net-studio but it
does compile in Borland. So is anyone familiar with this problem, or
someone knows how to solve it?

void MidpointSolver::Solve(float* next_state, const float*
current_state, float time, float h, const unsigned int dim)
{
float derivative_state[dim];
...
}

The reported error is: C2057 - 'expected constant expression'. So the
compiler doesnt see 'dim' as an constant or something like that?!?

Anyone?

The reported error is correct because dim is not a constant as the
compiler cannot see its actual value (which should be a defined
constant). However, it´s hard to believe that Borland will compile
this without at least a warning. Try to increase your warning levels
and recompile under Borland (BTW which version?).

It is possible that Borland might have implemented variable length arrays
as a compiler extension? But from a C++ standpoint, dim isn't a compile-
time constant and cannot be used as an array size specifier (from a
Standard point of view....)

[Chris Theis]

"Andre Kostur" <nn******@kostur.net> wrote in message
news:Xn*******************************@207.35.177. 135...
[SNIP]> > The reported error is correct because dim is not a constant as the

compiler cannot see its actual value (which should be a defined
constant). However, it´s hard to believe that Borland will compile
this without at least a warning. Try to increase your warning levels
and recompile under Borland (BTW which version?).

It is possible that Borland might have implemented variable length arrays
as a compiler extension?

[SNIP]
I´ve also had that idea but I doubt that they do not at least give a hint
via a warning. Unfortunately I do not have a Borland compiler available to
check this. Hence, probably somebody else could shed some more light onto
this matter.

Chris

[Ioannis Vranos]

"Chris Theis" <Ch*************@nospam.cern.ch> wrote in message
news:c6**********@sunnews.cern.ch...

"Martijn" <r.m.maatman_@_student.utwente.nl> wrote in message
news:c6**********@netlx020.civ.utwente.nl...

Hi,

The following code-fragment won't compile in the .Net-studio but it does
compile in Borland. So is anyone familiar with this problem, or someone
knows how to solve it?

void MidpointSolver::Solve(float* next_state, const float* current_state, float time, float h, const unsigned int dim)
{
float derivative_state[dim];
...
}

The reported error is: C2057 - 'expected constant expression'. So the
compiler doesnt see 'dim' as an constant or something like that?!?

Anyone?

The reported error is correct because dim is not a constant as the

compiler cannot see its actual value (which should be a defined constant). However,
it´s hard to believe that Borland will compile this without at least a
warning. Try to increase your warning levels and recompile under Borland
(BTW which version?).

Many compilers support VLAs as a system extension, the same happens with
long long. All of course give an error when in strict ISO mode.


Ioannis Vranos

[Richard Herring]

In message <c6**********@sunnews.cern.ch>, Chris Theis
<Ch*************@nospam.cern.ch> writes

"Andre Kostur" <nn******@kostur.net> wrote in message
news:Xn*******************************@207.35.177 .135...
[SNIP]> > The reported error is correct because dim is not a constant as the

> compiler cannot see its actual value (which should be a defined
> constant). However, it´s hard to believe that Borland will compile
> this without at least a warning. Try to increase your warning levels
> and recompile under Borland (BTW which version?).

It is possible that Borland might have implemented variable length arrays
as a compiler extension?

[SNIP]
I´ve also had that idea but I doubt that they do not at least give a hint
via a warning. Unfortunately I do not have a Borland compiler available to
check this. Hence, probably somebody else could shed some more light onto
this matter.

Not BCB5, at least:

void Solve(float* next_state, const float* current_state,
float time, float h, const unsigned int dim)
{
float derivative_state[dim];

}

[C++ Error] Unit1.cpp(6): E2313 Constant expression required

--
Richard Herring