inequation / dpx Goto Github PK

What steps will reproduce the problem?
1. Take a dpx in 4:2:2 CbCr 10bits filledmethodA with width 2048
2. Try to read the data or to convert to tiff

What is the expected output? What do you see instead?
There is one "Y" that is truncated at the end of the first line. => second line 
things Cb is Y.

What version of the product are you using? On what operating system?
Windows. "0.5.0". Operating system not related item.

Please provide any additional information below.
offset calculation is wrong in Read10bitFilled function.

When reading a full image using Read10bitFilled function:
For an image of 2048 bytes, offset of line 1 is calculated as 
"1365 * 4 bytes". It should be "1366 * 4 bytes"

Indeed, you need 1 4bytes word for 3 components (Y or Cb or Cr).
=> offset in number of words on line 1 should be 
number of components = (2048*2)
number of words = (2048*2)/3 = 1365,3333

If you use 1365, the last "Y" of the line will be truncated and lost. Correct 
value is 1366.

What steps will reproduce the problem?
1. Trying to write dpx file using the data taken from IplImage (OpenCv)/ This 
means the data is unsigned char with 3 channels for RGB. Using 10-bits format. 
The application crushed due to accessing to unallocated memory. Could correct 
this issue by changing reinterpret_cast<IB*> (line 274 WriterInternal.h) to 
creating new array of type IB and filling it with values from imageBuf.
2. So the application could finish successfully, but still the resulting 
written image is fully black.

What is the expected output? What do you see instead?
RGB image of dpx format with proper scene on it, instead i get black image.


What version of the product are you using? On what operating system?
I'm running r3 on Windows, Visual Studio 2008.


Please provide any additional information below.
If i correct any issue in the project if you don't mind i could update version 
here.

When compiling on Debian Linux, the compilation fails because GCC can't find a 
"DPXstream.h". This is an obvious case sensitivity issue, which probably 
results from the library being developed on win32 or another case-insensitive 
platform.

Also, there's a warning in DPXHeader.cpp where a value of 0xffffffff is 
assigned to a 16-bit unsigned integer.

Attached is a patch that fixes the two issues.

When compiling with -Wall -Wextra, there are a couple more warnings that 
appear. Attached is a patch that fixes them.

Single-channel (i.e. grayscale/luma) 10- and 12-bit filled DPX images are 
displayed incorrectly - column triplets are mirrored horizontally.

Packed 10- and 12-bit images are displayed incorrectly because the wrong 
variable is being used to calculate the bit shift.

A patch is attached that fixes these problems.