edgar-mtz-e / slimdx Goto Github PK

While I'm nitpicking about this class, is its name valid? It's basically a
generic memory stream (but more flexible than the .NET MemoryStream) for
locked/mapped/user-created resource data. If any other APIs (DInput,
DSound, XAct) require that kind of interface to buffer data, we might want
to consider renaming the class to be more generic (i.e., get rid of the
"graphics" bit). 

Presumably the investigation of whether or not this makes sense should be
completed before our first official versioned release so that we don't have
to break the API later.

Initial investigation suggests this might be worthwhile, as the DSound API
has a buffer class that can be locked similar to who D3D buffers are locked.

<Tape_Worm> Got a SlimDX issue for you.  I tried to post it to google code,
but the damn thing keeps giving me a server error.
<Tape_Worm> The issue is that you cannot create an additional swap chain
like you can with IDirect3DDevice9::CreateAdditionalSwapChain().   The only
constructor for the SwapChain object is SwapChain(IDirect3DSwapChain9 *),
which obviously really doesn't fit the bill.

Offhand I noticed light support was missing, and materials probably have 
holes. No hurry. Do it on a slow Sunday instead of cutting yourself or 
something.

This is silly because there is one in the base level that should be usable,
so as long the HRESULT codes don't overlap. Look into this and craft an
appropriate solution.

Should be straightforward.

Should be straightforward.

If anybody can ever stomach doing the work, this is something that needs to
be done. I'm not particularly concerned about getting it done soon though.

I get the following error when trying to compile revision 103:

1>Linking...
1>Texture.obj : error LNK2020: unresolved token (06000170)
SlimDX.Direct3D9.BaseTexture::GenerateMipSublevels
1>Sprite.obj : error LNK2020: unresolved token (060001C7)
SlimDX.Direct3D9.BaseTexture::GenerateMipSublevels
1>Effect.obj : error LNK2020: unresolved token (060001A9)
SlimDX.Direct3D9.BaseTexture::GenerateMipSublevels
1>Device.obj : error LNK2020: unresolved token (060001CF)
SlimDX.Direct3D9.BaseTexture::GenerateMipSublevels
1>D:\Tavaraa\SlimDX\build\Debug\SlimDX.dll : fatal error LNK1120: 4
unresolved externals
1>Build log was saved at "file://d:\Tavaraa\SlimDX\build\Debug\BuildLog.htm"
1>SlimDX - 5 error(s), 0 warning(s)

I use the exact same setup as with previous revisions.

//chinc

This class is basically empty right now.

Should be trivial addition.

They're ok for now, but it is starting to get a little unwieldy.

The following functions need to be written:
GetBufferPointer,
GetBufferSize,
GetConstantDesc,
GetConstantElement,
GetDesc,
SetDefaults,
SetBoolArray

Should be easy.

Currently there is no XACT support, or sound support of any kind. I'd like
to change that.

Perhaps we can try to take advantage of .NET features to make using SlimDX
easier. For example, instead of having the Get / Set RenderState function,
we  could have a RenderState property that points to a RenderState class
with all render state possibilities.

I think this is one of the things that XNA got absolutely right. Some of
the changes they made to the Direct3D interface make perfect sense, and it
would be nice if SlimDX had that elegance as well.

What steps will reproduce the problem?
1. The D3DXMesh routines are very useful for game tools but only a minimum
set of them are present for now.
2.
3.

What is the expected output? What do you see instead?

At least it should allow to create a mesh and access the VB and IB inside
of the mesh. So that ComputeTangent() can be called for computation
generated mesh. Other methods like Intersect() and access to attribute
buffer are also appreciated.

What version of the product are you using? On what operating system?


Please provide any additional information below.

Effect does not currently support defines, includes, skip constants, or 
pools.

When you have the time, can you change the project file 'Output Directory' 
value from:
$(SolutionDir)$(ConfigurationName)

to

$(ConfigurationName)

The current setting is messing up my directory structure (places a Debug 
directory in the root level of my solution directory tree).

Some of the math objects (Matrix and Vector) contain duplicate methods -
they have static and non-static versions of operations (such as
Matrix::Transpose()). I think we should eliminate one of the two, probably
the non-static methods.

    EffectScalarVariable^ EffectVariable::AsScalar()
    {
        ID3D10EffectScalarVariable* variable = m_Pointer->AsScalar
();
        if( variable == NULL || !variable->IsValid() );

I'd imagine that ';' at the end of the if statement is a typo? :)

[deleted issue]

GraphicsStream can currently accept a size of 0, which means "I don't know
my size." This is mostly a compatibility hack to allow locking of resources
we don't know the size of yet; however, it's a problem because it disables
some of the stream functionality. 

Should make a pass through anything that interacts with GraphicsStream and
try to get the size of the locked region, so that the "0 for unknown" hack
can be removed from the stream.

XACT is a higher priority than DSound, but both should be put in.

This file originally existed to force a lot of header-only things to
compile. Now that the codebase has been significantly patched up, it really
exists just to force the contents of Utils.h to compile. It should be
renamed to reflect that.

I've managed to finally get SlimDX running on one of my project files.  I 
did run into one issue regarding IDirect3DDevice9.SetTexture().  When I 
passed NULL for the texture object, it came back with an object reference 
not initialized error.  This was due to the following code:

hr = m_Pointer->SetTexture( sampler, texture->BaseTexturePointer );

Obviously, since 'texture' is nullptr, it cannot retrieve the instance of 
BaseTexturePointer.  For now I've just wrapped it in a nullptr check

hr = m_Pointer->SetTexture( sampler, texture != nullptr ? texture-
>BaseTexturePointer : NULL );

There may be more issues like this, and I'll update if I find them.

This class is basically empty right now.

There are Matrix, Scalar and Resource variable types, but nothing for 
vectors.

I tried treating a vector as an array of scalar variables, but that 
doesn't play nice.

[deleted issue]

See what is necessary to add R2VB support, and what to name it if we 
decide to add it.

Intellisense lists all methods/properties an object has, except the Dispose
() methods for objects that have them.

Try:

VertexBuffer buffer;
buffer.Dispo|

Then hit CTRL+Space to request intellisense popup (if it is not there 
already). Notice it is not listing the Dispose() method, yet the object 
DOES have a Dispose() method.

I've noticed that the binaries are not signed with a strong name.  This 
causes an issue when signed libraries depend on the assembly.  Not to 
mention it can't be put in the GAC should one choose to do that.

It'd be a good idea to sign the binaries, it's also helpful to confirm who 
created the dll should that ever be an issue.

Accidentally didn't mark these static. Need to change in Surface.h. Should
be trivial change.

From chinc in the GDNet thread:

"The geometric functions in SlimDX seem to be left-handed, and I decided to
make my new engine use right handed coordinate systems. Can you make right
handed versions of the functions? Of course I can do this myself, but it
would be convenient when upgrading if the functions were right in the box."

The implementation of GetDeviceCaps ignores the D3D return value. For an 
example an adapter that does not support hw rendering but was queried for 
it will lead to a invalid returned caps structure. Apart from that the 
user has no chance to know what went wrong like D3D not available, out of 
video memory or the like.

The Device has mostly Set functions and lacks Get functions. Gets for
surfaces should be considered priority. GetRenderTarget, etc.

The "Usage" enumeration doesn't contain a flag for software processing 
(SoftwareProcessing).  This may seem like an unnecessary flag, but a lot 
of on-board chipsets (e.g. intel) don't actually have T&L or shaders, so 
this could be necessary.

The TextureOperation has a broken item:
BumpEnvMapLuminanceD3DTOP_BUMPENVMAPLUMINANCE,

Should be:
BumpEnvMapLuminance = D3DTOP_BUMPENVMAPLUMINANCE,

It'd be nice to have a migration guide that is up-to-date for people who
are switching to SlimDX. In particular, we want to call out intention
omissions and differences, so we don't get bug reports on them, as well as
new or improved features.

Also, I wanted to play with the issue system and see if stimulates SVN
revision number updates like editing the wiki does. Because that's silly.

XAudio2 was introduced with the August 2007 SDK. It is a fairly small API 
and should not pose any major difficulties.

See: http://www.mdxinfo.com/tutorials/profiling.php

Vector2 lacks a couple of useful functions like Normalize or ToString(). 
The latter would be very useful for debugging Vector2 and Vector3 in 
Visual Studio.

Buffer constructor taking a native pointer to an ID3D10Buffer is public.

I can find the mouse state object in the DirectInput classes, but where is
keyboard state?

//chinc

I wanted to put in another vote that D3D10 is important and a good idea.

I know it's planned, but also please know there are those would start using
it immediately when ready.

Thanks to the small team for this project - It truly a great thing and so
much better than XNA.

You guys rock.

If all graphics exceptions are disabled, Query will not be usable. This
needs to be fixed.

Would it be possible to switch over to the Aug 2007 version of the SDK?  

It's not terribly vital, but it's just a little annoying whenever I do an 
update to have it mess up upon recompile because it's expecting the June 
2007 SDL.

[deleted issue]

I'm not 100% certain this is due to SlimDX or not.  But, seeing as the 
sample water demo exhibits the same behaviour, as does my code, I thought 
I might post it here.  And this problem does NOT exist in MDX, nor in the 
C++ samples.  I've also tested this in windowed mode for SlimDX and it 
works perfectly there.  So it's just fullscreen with SlimDX apps by the 
looks of it.

The problem:  When setting up for full screen mode (I quadruple checked my 
PresentParameters, so I'm certain those are not the issue) it goes into 
full screen mode just fine, then it enters the render loop.  In the render 
loop it draws the scene, flips (I've visually verified this as it flashes 
on my screen before dying), then processes messages (Application.DoEvents
()).  As soon as it processes messages it switches back to windowed mode 
(the window becomes minimized by the looks of it) and any attempt to 
restore it just crashes.

I've tried processing the messages before the new Device() command to 
clear out the queue, but no go.  As a further test I also placed the 
Application.DoEvents() BEFORE any drawing code and it has the problem, but 
this time nothing gets dumped to my screen before the mode switches.

I hope this is enough information.  If not, find me on IRC and we'll 
discuss it further.
Again, this behaviour is present on the water sample as well.  And I've 
verified that this does not happen with MDX nor the C++ API.  I've looked 
throught the slim dx code, and honestly I cannot see anything that would 
cause this behaviour.

Need to drop auto-release, since it doesn't work.

The new leak tracking features to be added:
* Assign every object an AllocID, like native DX does, and call out the ID
when reporting an undisposed object.
* Provide the option to break on AllocID in debug mode. Again, same as
native DX.
* DisposeHandler function for all objects. This is merely a pass through to
the Dispose function, but its signature makes it usable as an event handler.

D3D10 doesn't have the same ownership semantics (it calls
AddRef()/Release() differently than D3D9). Need to make sure the way D3D10
objects are implemented is correct - and figure out the correct way to
handle the various get* methods that D3D10 does not yet support, but will
soon need.

All of the Get* functions need to be written in; only GetValueString is 
currently in.