SCTK returns "C" rather than "C++" for the Primary Language of whisper.cpp

[DennisClark]

A recent scan of whisper.cpp-1.6.2.tar.gz from https://github.com/ggerganov/whisper.cpp/archive/refs/tags/v1.6.2.tar.gz returned a Primary Language of "C". (Scan results attached.) Given the name of the project, whisper.cpp, that seems very odd to me. Also please note that GitHub https://github.com/ggerganov/whisper.cpp/tree/v1.6.2 shows the following Language breakdown:
C++ 74.1%
C 16.2%
Cuda 4.4%
Metal 1.6%
Objective-C 1.5%
CMake 0.6%
Other 1.6%

This example alone is probably not a serious bug, but may be worth investigating.

whisper.cpp-1.6.2.tar.gz_scan.zip

[RajX-dev]

Hi @DennisClark, I've analyzed this locally.
I ran a fresh scan on whisper.cpp and verified that src/whisper.cpp is correctly identified as C++ (MIME type text/x-c++).
Since the file-level detection is correct, the issue likely lies in how the 'Primary Language' is aggregated. I suspect ScanCode is determining the primary language based on file count (where ggml's many small .c files might be winning) rather than byte count (which is how GitHub calculates it, giving C++ the 74% majority).
I would like to work on this. My plan is to investigate the aggregation logic and see if weighing by file size yields the correct 'C++' result

[DennisClark]

@RajX-dev thanks for your insightful comments. Please feel free to work on this, understanding that your work will need to be reviewed by at least one of my colleagues.

[RajX-dev]

Hi @DennisClark, just following up to let you know I've submitted a PR for this: #4733

It implements the logic change to weigh language tallies by byte size rather than file count, which correctly identifies the project as C++. Ready for review whenever you have a moment!

[RajX-dev]

Hi @DennisClark, just a quick follow-up on this.

I understand it needs a review from one of your colleagues—whenever someone has a bit of bandwidth to take a look, I'm happy to address any feedback or make adjustments as needed. Thanks again!