[Windows] Linker error "undefined symbol: quantize_i2_s" when building with -q i2_s or -q tl2 on Windows 11

When attempting to build the project on Windows at 16da220ae2b510caff437d403288882687f44ae5 using python setup_env.py -md models/BitNet-b1.58-2B-4T -q tl2(or -q i2_s), the build fails during the CMake linking phase.

The linker (lld-link) reports undefined symbol errors for quantize_i2_s and dequantize_row_i2_s inside ggml-base.vcxproj.

It appears that recent updates in the 3rdparty/llama.cpp submodule (specifically ggml.c) reference the i2_s quantization symbols, but the source files implementing these i2_s functions are not included in the compilation target, causing a linker failure.

Terminal Output:

(.venv) PS D:\Projects\BitNet> python setup_env.py -md models/BitNet-b1.58-2B-4T -q tl2
INFO:root:Compiling the code using CMake.
ERROR:root:Error occurred while running command: Command '['cmake', '--build', 'build', '--config', 'Release']' returned non-zero exit status 1., check details in logs\compile.log

Relevant Section from logs/compile.log:

lld-link : error : undefined symbol: quantize_i2_s [D:\Projects\BitNet\build\3rdparty\llama.cpp\ggml\src\ggml-base.vcxproj]
  >>> referenced by ggml-base.dir\Release\ggml.c.obj:(ggml_quantize_chunk)
  
lld-link : error : undefined symbol: dequantize_row_i2_s [D:\Projects\BitNet\build\3rdparty\llama.cpp\ggml\src\ggml-base.vcxproj]
  >>> referenced by ggml-base.dir\Release\ggml.c.obj:(type_traits)
  Building Custom Rule D:/Projects/BitNet/3rdparty/llama.cpp/common/CMakeLists.txt
  llama-common-base.vcxproj -> D:\Projects\BitNet\build\3rdparty\llama.cpp\common\Release\llama-common-base.lib
  Building Custom Rule D:/Projects/BitNet/3rdparty/llama.cpp/tools/mtmd/CMakeLists.txt
  llama-gemma3-cli.vcxproj -> D:\Projects\BitNet\build\bin\Release\llama-gemma3-cli.exe
  Building Custom Rule D:/Projects/BitNet/3rdparty/llama.cpp/tools/mtmd/CMakeLists.txt
  llama-llava-cli.vcxproj -> D:\Projects\BitNet\build\bin\Release\llama-llava-cli.exe
  Building Custom Rule D:/Projects/BitNet/3rdparty/llama.cpp/tools/mtmd/CMakeLists.txt
  llama-minicpmv-cli.vcxproj -> D:\Projects\BitNet\build\bin\Release\llama-minicpmv-cli.exe
  Building Custom Rule D:/Projects/BitNet/3rdparty/llama.cpp/tools/mtmd/CMakeLists.txt
  llama-qwen2vl-cli.vcxproj -> D:\Projects\BitNet\build\bin\Release\llama-qwen2vl-cli.exe
  Building Custom Rule D:/Projects/BitNet/3rdparty/llama.cpp/tools/ui/CMakeLists.txt
  llama-ui.vcxproj -> D:\Projects\BitNet\build\3rdparty\llama.cpp\tools\ui\Release\llama-ui.lib

Environment:

• OS: Windows 11 Home (Build 10.0.26200)
• Compiler / Toolchain: Visual Studio 2026 + Clang 22.1.3 (x86_64-pc-windows-msvc)
• CMake Version: 4.3.1-msvc1
• Python Version: 3.10.11

Start with 3rdparty/llama.cpp/ggml.c and the CMake targets that produce ggml-base.vcxproj, then reproduce the failure with setup_env.py using -q tl2 or -q i2_s. Trace where quantize_i2_s and dequantize_row_i2_s are defined and whether their source files enter the Windows target. Done means both commands complete the CMake build without the undefined-symbol linker errors.