SFML 2.0 RC

[Contadotempo]

Don't know, might have done something wrong cause it just didn't want to generate the needed files. Gave some error I don't remember but that didn't make sense.

Are you using the express edition by any chance? It seems there was a bug with Visual Studio express 2012 and cmake 2.8.9
Take a look at this, it was fixed on cmake 2.8.10:
http://public.kitware.com/Bug/view.php?id=13348

And this for an alternative fix:
http://stackoverflow.com/questions/12434696/visual-studio-express-2012-and-cmake

[eXpl0it3r]

Are you using the express edition by any chance?

Well this could likely be, since I'm using Visual Studio 12 Premium (thanks university!).

[cekota]

Hello,

in "include/SFML/Graphics/Transform.h" we have :

sfTransform sfTransform_Identity;

I have by error modified it. So after this modification it is not an identity matrix anymore.

"const" makes it unmodificable, but still if I provide it by error to a function like sfTransform_rotate() the function will apply silently while actually the operation has not been done without any error reported to the programmer.

So I would suggest to remove "sfTransform_Identity", and provide instead "sfTransform_getIdentity()" :

sfTransform sfTransform_getIdentity(void);

Cheers

[Laurent]

"const" makes it unmodificable, but still if I provide it by error to a function like sfTransform_rotate() the function will apply silently while actually the operation has not been done without any error reported to the programmer.

Hum? Why would it apply silently? It will trigger a compiler error, that's the point of using the const keyword.

[cekota]

Hum? Why would it apply silently? It will trigger a compiler error, that's the point of using the const keyword.

Just remember what CSFML is made for

Is it made mainly to write pure plain C ?
Or is its main purpose for something else ?

[cekota]

and this is also why a function would be better.

[Laurent]

I still don't get it.

[cire]

I believe Cekota was suggesting sfTransform_Identity be accessed through a function that supplies a copy of it.

Currently it's not constant and may be changed by a user, and subsequent uses of the identity will (silently) fail to act as one would expect.  The function would make similar scenarios less likely.

[FRex]

Yeye it's always the fault of the software...

Or authors.
Your mileage may vary but software tends to work weird on some setups. On my laptop vc++2008's intelli sense works alright(generously speaking, I'm having problem with sfg shared ptrs). But on my pc, opening or closing a block with sharp bracket, using scope operator on class that just got a new function added ect. doesn't update itself and I must reopen the solution. There is no apparent reason to this(same windows, same installation, same settings, same files even).

[Laurent]

I believe Cekota was suggesting sfTransform_Identity be accessed through a function that supplies a copy of it.

Currently it's not constant and may be changed by a user, and subsequent uses of the identity will (silently) fail to act as one would expect.  The function would make similar scenarios less likely.

But the const keyword solves the problem. That's what I don't understand (he says it doesn't).

[cire]

I don't know exactly what he meant there, either, unless he's paranoid enough to want protection against people casting away const, in which case -- good luck. 

[cekota]

sorry, I thought it was so evident that CSFML is often used for building bindings to higher level languages, but it seems it is not.

indeed, if the variable is wrapped, then the const qualification is discarded

[Laurent]

Ok, but... the binding will also make it const. CSFML is not supposed to solve potential mistakes in bindings that use it.

If I use a function and return a new copy every time it is called, it can even be worse: a binding that would not make the corresponding object const/readonly/whatever, wouldn't see it and it would "work" although the design has a major flaw.

This is purely a design issue, technical choices won't help. The only good solution is to make sure that the objects are const/readonly/... in all the bindings. So thank you for pointing out that they were not in CSFML

[cekota]

Ok, but... the binding will also make it const. CSFML is not supposed to solve potential mistakes in bindings that use it.

If I use a function and return a new copy every time it is called, it can even be worse: a binding that would not make the corresponding object const/readonly/whatever, wouldn't see it and it would "work" although the design has a major flaw.

This is purely a design issue, technical choices won't help. The only good solution is to make sure that the objects are const/readonly/... in all the bindings. So thank you for pointing out that they were not in CSFML

Well, I don't want to make you change your mind, but I would like to understand, because your are an experienced programmer, and I'm a beginner, so I would really want to understand in order to improve my skills.

The problem is that I really don't see what I can do with a const sfTransform_Identity. If I want to use it I will copy it and then use some rotate and translate functions on the copy. So in this case I was thinking that sfTransform_getIdentity() would makes sens.

So please explain me what I'm supposed to do with this const sfTransform_Identity ?

[cekota]

CSFML: include/SFML/Graphics/Texture.h:
CSFML_GRAPHICS_API void sfTexture_bind(const sfTexture* texture);

SFML: include/SFML/Graphics/Texture.hpp:
void bind(CoordinateType coordinateType = Normalized) const;

The parameter "coordinateType" is missing.