提供: wiki
移動先: 案内検索

Blender Dev FAQ

What Language is Blender written in?

Short Answer C/C++/Python.

  • C - for Blender's [startup/main-loop/shutdown] and most internal functionality, tools, drawing, GUI and editors.
  • C++ - physics, audio, game-engine and Cycles render engine.
  • Python - The interface definitions, addons, most format import/export and some tools.

Why C/C++/Python?

Part of it is historic, part of it is convention and choice of the current developers.

Early on when Blender was developed, (1990's), C was very common for developing graphics applications (C++ compilers were expensive, where C compilers were free).

Python was selected because of its great community.

Nowadays Blenders core is still in C, some libraries are in C++, and OSX integration even uses Objective-C, Scripting is done in Python.

Currently this works fairly well and we don't have plans to rewrite large parts of Blender in some other languages.

What Editors/IDE's do blender developers use?

These IDE's are known to work well for blender development where CMake is used to create project files.

  • XCode on OSX
  • MSVC 2013/2015/2017 on Windows
  • Eclipse on any
  • QtCreator on any
  • Netbeans on any

Some developers use editors such as vim, emacs, geany, notepad++ and scite.

Note that its common to use a different editor/ide for C/C++ and Python code.

Where is the right place to request features in blender?

There are so many possible features to add to blender that its difficult to manage them usefully so we don't accept feature requests on our tracker as some projects do.

If the feature requests are small, like an option to an existing tool you could try to contact the developer of that tool, or at least the maintainer.
See Module Owners

If you're suggestion is a larger change, try make this a suggestion on bf-funboard and use this to get feedback,
See bf-funboard mailing list

Why did my Blender compile fail?

There may be many reason why it failed, but here are some common causes. Also remember to look for the first error that is printed, as this is usually the root cause, further errors may only be a result of this.

File not found or Unresolved Symbols

This usually means a library was not installed, not configured in the build system, or the wrong version of the library was used. Also make sure to download the lib/ folder on Window and Mac, and to install the needed development packages on windows. It's also possible to disable certain libraries when building Blender to avoid the issue. See the Building Blender documentation for details on dependencies.

Blenderplayer failed to compile due to missing functions

Many developers don't compile the blenderplayer, so don't notice this - if it happens grep the source, find the function prototype and copy and paste the function prototype in the stubs.c file with the appropriate NULL or 0 return. If a struct is return you should cast NULL to the appropriate struct type

DNA compile errors

Usually this means you added something to a struct in a header file and forgot to adjust the amount of padding in the struct. Blender requires that the structs be multiples of 8, see the SDNA Notes for more detail. So if you add a short, and there is a char pad[1]; you change it to char pad[7], to keep it a multiple of eight. Of course if you forget the rules just keep adding a char pad till it compiles without errors :) (which is what you will do if it compiles in your 32 bit machine and others say it is failing in their 64 bit ones... save some time and read the rules instead, alignment matters too, no matter how much padding you add).

RNA compile errors

This usually means the wrong DNA struct type or property name was specified.

Missing Interface Buttons and Menus

Usually this means you added something to the user interface python code and made a mistake. Check the console for errors, often this is because of mixing tabs and spaces. All python scripts in Blender uses spaces for indentation.

Menu or checkbox defaults to the wrong value

Make sure you added a default value in the function that creates the struct, and that you create a version patch in do_versions in readfile.c to initialize values on existing structs.

Why did my Blender compile with warnings?

While Blender can compile warning-free on some platforms/configurations, some warnings persist and can't simply be "fixed" or supressed.

If you see warnings while compiling, you can assume Blender developers are aware of them and they're almost certainly false-positves, or at least not causing real-world problems.

(This goes for static analysis tools too).

However this doesn't mean you should ignore warnings completely, For code you write, check warnings carefully and resolve them where appropriate (fix/supress).

Whats the situation with automated testing?

Testing in Blender is a work in progress, and while we aren't anywhere near full coverage, we have tests for:

  • Python: some basic tests exist already in: source/tests/.
  • GTest: tests for C/C++ code.
  • CTest (from CMake), can be used to call the tests above.

See the project-page: Automated Testing

How Does Blender Deal with Security?

Blender does not attempt to achieve the same level of security as many other applications (web browsers for example),

The ability to have blend files that execute bundled Python scripts does pose a security risk if you don't know who created the file.

The Trusted Source option has been added to the file so you can load a blend file without running scripts as a precaution to simple attacks.

However this is no protection against more advanced exploits such as hand crafting a blend file which uses buffer overflows to run malicious code.

For more discussion on this topic:

Source Code FAQ

How do I build Blender?

Follow the excellent instructions on Dev:Doc/Building_Blender.

What documentation is there?

Dev:Source/Architecture has a bunch of documentation on the concepts behind Blender's C code. It is not easy for a newcomer to understand, there is a lot to read, and it is not completely up to date. But you should keep coming back to it - the more you understand pieces of Blender, the more all of this will make sense and be helpful.

Where does blender start and what happens next?

The file source/creator/creator.c has the main() routine in it.

After parsing args, a bunch of initialization, and reading a .blend file, main() passes control to WM_main() in source/blender/windowmanager/intern/wm.c. That is an infinite loop that just processes events and notifications and draws updates on the screen. Events and notifications are things caused either by the user doing something (e.g., mouse click) or the internal blender systems signaling to other parts of blender that something has changed and that means something has to be recalculated, redrawn, or something else.

What is an overview of how the files and subsystems are organized?

Here is a great picture.

Most of the code you will care about is under the source/blender/ subdirectory, though some Python code you may want to look at is under release/scripts/.

Where is the UI code?

The UI layout is defined in Python but the drawing is done in C, and some of the UI is defined in C too.

The Python code is in

  • release/scripts/modules/bl_ui/. E.g., in that directory is the code for the 3D View editor type (space).

The C interface drawing code is in:

  • source/blender/editors/interface/

The glue code defining the API python uses to create interfaces

  • source/blender/makesrna/intern/rna_ui_api.c

The Python UI code mostly just lays out user-interaction widgets that adjust properties of operators. Operators are Blender tools. At some point you should read Dev:2.5/Source/Architecture/Operators.

How are Operators connected to C code?

If you hover over a button like Spin in the user interface, it shows you the Python name for the operator that will be invoked if you hit the button - bpy.ops.mesh.spin(). This is implemented in C by an operator named MESH_OT_spin (see the naming convention in Dev:2.5/Source/Architecture/Operators). If you look for that name, you'll find it as a function that fills in various function pointers and flags in a wmOperatorType struct. The important one is the invoke function, in this case spin_mesh_invoke(). If you put a breakpoint on that you'll find that it gets called soon after you push the Spin button.

The Python bpy_operator.c code knows how to go back and forth between module.func and MODULE_OT_func.

What is this DNA and RNA stuff?

The names come from the genetic nucleic acids and hint at their function: DNA, as in genetics, is the code that lets you reproduce something. In the case of Blender, this is essentially what is written to and read from disk as a .blend file - it has all of the user settings and mesh, object, scene, etc. data. RNA, as in genetics, is a helper to transmit the DNA code into the rest of the internal Blender system.

The .h files in source/blender/makesdna/ define the structs, enums, etc., for all the data that needs to be persisted to disk. These are usually used in memory in the running system too. For example, DNA_mesh_types.h defines the Mesh struct that has enough information to store a user's mesh on disk. While running, there may be a Mesh in memory, but there will likely be various derived mesh structures that are more efficient to use during editing, say.

Another aspect of DNA is that it is also a system for helping compatibility between different versions of Blender. The way this works is that a coded description of the layout of each struct is also written to and read from disk. This means that the layout does not have to stay fixed. The C files in the makesdna directory are for generating and interpreting the structure descriptions. See Dev:Source/Data_Structures/DNAStructs.

RNA is also something that is partially made by code generation, using code in source/blender/makesrna. It is supposed to be a somewhat more abstract and nicer interface to the DNA data. For example, where a DNA struct might use a particular bit in a short for some flag, RNA would expose that as a bool property. So far, this nicer Data API isn't used too much in the C source code. Where it is heavily used is as the interface to and from Python. Most of the Python access to Blender data is generated automatically to use the RNA system. See Dev:2.5/Source/Architecture/RNA for more information about RNA.

Where are the icons?

The icons are stored in a file called blender_icons.svg is one large SVG file with almost all of the button icons in a kind of matrix arrangement in it. There is a shell file to convert that file into 16x16 and 32x32 per icon versions of the SVG file as data files (not user editable), these are stored in release/datafiles/blender_icons16, release/datafiles/blender_icons32.

The brush icons are separate PNG files in the brushicons subdirectory. The datafiles directory also has the splash image and some TrueType fonts.

All of these data files are turned into C files that declare their data as char arrays at build time, so any modifications to these files will be visible after rebuilding.

The code for initializing the icons is in source/blender/editors/interface/interface_icons.c, in the init_internal_icons() function. The order of the icons in the image is, bottom to top, left to right, the order of the DEF_ICON declarations in source/blender/editors/include/UI_icons.h.