Revision History

No longer updated. Kept to preserve ancient history. For more recent history see the repo log at https://github.com/KhronosGroup/KTX-Software. See also the Release Notes in the repo.

Version 4.0

Added:

• Support for KTX Version 2.
• Support for encoding and transcoding Basis Universal images in KTX Version 2 files.
• Function to print info about a KTX file.

Version 3.0.1

Fixed:

• GitHub issue #159: compile failure with recent Vulkan SDKs.
• Incorrect mapping of GL DXT3 and DXT5 formats to Vulkan equivalents.
• Incorrect BC4 blocksize.
• Missing mapping of PVRTC formats from GL to Vulkan.
• Incorrect block width and height calculations for sizes that are not a multiple of the block size.
• Incorrect KTXorientation key in test images.

Version 3.0

Added:

• new ktxTexture object based API for reading KTX files without an OpenGL context.
• Vulkan loader. #include <ktxvulkan.h> to use it.

Changed:

• ktx.h to not depend on KHR/khrplatform.h and GL{,ES*}/gl{corearb,}.h. Applications using OpenGL must now include these files themselves.
• ktxLoadTexture[FMN], removing the hack of loading 1D textures as 2D textures when the OpenGL context does not support 1D textures. KTX_UNSUPPORTED_TEXTURE_TYPE is now returned.

Version 2.0.2

Added:

• Support for cubemap arrays.

Changed:

• New build system

Fixed:

• GitHub issue #40: failure to byte-swap key-value lengths.
• GitHub issue #33: returning incorrect target when loading cubemaps.
• GitHub PR #42: loading of texture arrays.
• GitHub PR #41: compilation error when KTX_OPENGL_ES2=1 defined.
• GitHub issue #39: stack-buffer-overflow in toktx
• Don't use GL_EXTENSIONS on recent OpenGL versions.

Version 2.0.1

Added:

• CMake build files. Thanks to Pavel Rotjberg for the initial version.

Changed:

• ktxWriteKTXF to check the validity of the type & format combinations passed to it.

Fixed:

• Public Bugzilla 999: 16-bit luminance texture cannot be written.
• compile warnings from compilers stricter than MS Visual C++. Thanks to Pavel Rotjberg.

Version 2.0

Added:

• support for decoding ETC2 and EAC formats in the absence of a hardware decoder.
• support for converting textures with legacy LUMINANCE, LUMINANCE_ALPHA, etc. formats to the equivalent R, RG, etc. format with an appropriate swizzle, when loading in OpenGL Core Profile contexts.
• ktxErrorString function to return a string corresponding to an error code.
• tests for ktxLoadTexture[FN] that run under OpenGL ES 3.0 and OpenGL 3.3. The latter includes an EGL on WGL wrapper that makes porting apps between OpenGL ES and OpenGL easier on Windows.
• more texture formats to ktxLoadTexture[FN] and toktx tests.

Changed:

• ktxLoadTexture[FMN] to discover the capabilities of the GL context at run time and load textures, or not, according to those capabilities.

Fixed:

• failure of ktxWriteKTXF to pad image rows to 4 bytes as required by the KTX format.
• ktxWriteKTXF exiting with KTX_FILE_WRITE_ERROR when attempting to write more than 1 byte of face-LOD padding.

Although there is only a very minor API change, the addition of ktxErrorString, the functional changes are large enough to justify bumping the major revision number.

Version 1.0.1

Implemented ktxLoadTextureM. Fixed the following:

• Public Bugzilla 571: crash when null passed for pIsMipmapped.
• Public Bugzilla 572: memory leak when unpacking ETC textures.
• Public Bugzilla 573: potential crash when unpacking ETC textures with unused padding pixels.
• Public Bugzilla 576: various small fixes.

Thanks to Krystian Bigaj for the ktxLoadTextureM implementation and these fixes.

Version 1.0

Initial release.