[User] Failed to build with "make" on Windows 11

[Valikzant]

Prerequisites

Please answer the following questions for yourself before submitting an issue.

• [V] I am running the latest code. Development is very rapid so there are no tagged versions as of now.
• [V] I carefully followed the README.md.
• [V] I searched using keywords relevant to my issue to make sure that I am creating a new issue that is not already open (or closed).
• [V] I reviewed the Discussions, and have a new bug or useful enhancement to share.

Expected Behavior

I followed this step after cloning git:
On Windows:

Download the latest fortran version of [w64devkit](https://github.com/skeeto/w64devkit/releases).
Extract w64devkit on your pc.
Run w64devkit.exe.
Use the cd command to reach the llama.cpp folder.
From here you can run:
make

and expected a successful build of files i installed

Current Behavior

Instead of successful build i got an error while using 'make'. I provided full log in lower "Failure Logs" section

Environment and Context

Environment info:

E:/TextAIModels/LLama/llama.cpp $ git log | head -1
commit dc271c52ed65e7c8dfcbaaf84dabb1f788e4f3d0

CPU: АМD Ryzen 5 5600G with Radeon Graphics (12 CPUs) ~3.9GHz

E:/TextAIModels/LLama/llama.cpp $ python -V
Python 3.8.0

E:/TextAIModels/LLama/llama.cpp $ pip list | egrep "torch|numpy|sentencepiece"
numpy                1.24.2
sentencepiece        0.1.97
torch                1.13.1
torchaudio           0.13.1
torchvision          0.14.1

E:/TextAIModels/LLama/llama.cpp $ make --version | head -1
GNU Make 4.4

• Physical (or virtual) hardware you are using

I am using Windows 11

Failure Information (for bugs)

llama.cpp did not build by using 'make'

Steps to Reproduce

1. I used git clone https://github.com/ggerganov/llama.cpp in Windows powershell in a directory where i wanted llama.cpp to be installed
2. I installed the latest fortran version of w64devkit and lauched it
3. i reached directory where llama.cpp is located
4. i typed "make" and hit enter

Failure Logs

E:/TextAIModels/LLama/llama.cpp $ make
I llama.cpp build info:
I UNAME_S:  Windows_NT
I UNAME_P:  unknown
I UNAME_M:  i686
I CFLAGS:   -I.              -O3 -std=c11   -fPIC -DNDEBUG -Wall -Wextra -Wpedantic -Wcast-qual -Wdouble-promotion -Wshadow -Wstrict-prototypes -Wpointer-arith -march=native -mtune=native
I CXXFLAGS: -I. -I./examples -O3 -std=c++11 -fPIC -DNDEBUG -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function -Wno-multichar -march=native -mtune=native
I LDFLAGS:
I CC:       cc (GCC) 13.1.0
I CXX:      g++ (GCC) 13.1.0

cc  -I.              -O3 -std=c11   -fPIC -DNDEBUG -Wall -Wextra -Wpedantic -Wcast-qual -Wdouble-promotion -Wshadow -Wstrict-prototypes -Wpointer-arith -march=native -mtune=native   -c ggml.c -o ggml.o
g++ -I. -I./examples -O3 -std=c++11 -fPIC -DNDEBUG -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function -Wno-multichar -march=native -mtune=native -c llama.cpp -o llama.o
In file included from llama.cpp:8:
llama-util.h: In constructor 'llama_mmap::llama_mmap(llama_file*, bool)':
llama-util.h:234:94: note: '#pragma message: warning: You are building for pre-Windows 8; prefetch not supported'
  234 |         #pragma message("warning: You are building for pre-Windows 8; prefetch not supported")
      |                                                                                              ^
llama-util.h:202:47: warning: unused parameter 'prefetch' [-Wunused-parameter]
  202 |     llama_mmap(struct llama_file * file, bool prefetch = true) {
      |                                          ~~~~~^~~~~~~~~~~~~~~
llama.cpp: In function 'size_t llama_set_state_data(llama_context*, const uint8_t*)':
llama.cpp:2685:27: warning: cast from type 'const uint8_t*' {aka 'const unsigned char*'} to type 'void*' casts away qualifiers [-Wcast-qual]
 2685 |             kin3d->data = (void *) inp;
      |                           ^~~~~~~~~~~~
llama.cpp:2689:27: warning: cast from type 'const uint8_t*' {aka 'const unsigned char*'} to type 'void*' casts away qualifiers [-Wcast-qual]
 2689 |             vin3d->data = (void *) inp;
      |                           ^~~~~~~~~~~~
g++ -I. -I./examples -O3 -std=c++11 -fPIC -DNDEBUG -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function -Wno-multichar -march=native -mtune=native -c examples/common.cpp -o common.o
examples/common.cpp:20: warning: "NOMINMAX" redefined
   20 | #define NOMINMAX
      |
In file included from E:/Progs/Fortran/w64devkit/lib/gcc/i686-w64-mingw32/13.1.0/include/c++/i686-w64-mingw32/bits/c++config.h:679,
                 from E:/Progs/Fortran/w64devkit/lib/gcc/i686-w64-mingw32/13.1.0/include/c++/bits/requires_hosted.h:31,
                 from E:/Progs/Fortran/w64devkit/lib/gcc/i686-w64-mingw32/13.1.0/include/c++/string:38,
                 from examples/common.h:7,
                 from examples/common.cpp:1:
E:/Progs/Fortran/w64devkit/lib/gcc/i686-w64-mingw32/13.1.0/include/c++/i686-w64-mingw32/bits/os_defines.h:45: note: this is the location of the previous definition
   45 | #define NOMINMAX 1
      |
examples/common.cpp: In function 'int estimateWidth(char32_t)':
examples/common.cpp:622:28: warning: unused parameter 'codepoint' [-Wunused-parameter]
  622 | int estimateWidth(char32_t codepoint) {
      |                   ~~~~~~~~~^~~~~~~~~
g++ -I. -I./examples -O3 -std=c++11 -fPIC -DNDEBUG -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function -Wno-multichar -march=native -mtune=native examples/main/main.cpp ggml.o llama.o common.o -o main
examples/main/main.cpp:26: warning: "NOMINMAX" redefined
   26 | #define NOMINMAX
      |
In file included from E:/Progs/Fortran/w64devkit/lib/gcc/i686-w64-mingw32/13.1.0/include/c++/i686-w64-mingw32/bits/c++config.h:679,
                 from E:/Progs/Fortran/w64devkit/lib/gcc/i686-w64-mingw32/13.1.0/include/c++/bits/requires_hosted.h:31,
                 from E:/Progs/Fortran/w64devkit/lib/gcc/i686-w64-mingw32/13.1.0/include/c++/string:38,
                 from ./examples/common.h:7,
                 from examples/main/main.cpp:6:
E:/Progs/Fortran/w64devkit/lib/gcc/i686-w64-mingw32/13.1.0/include/c++/i686-w64-mingw32/bits/os_defines.h:45: note: this is the location of the previous definition
   45 | #define NOMINMAX 1
      |
examples/main/main.cpp: In function 'int main(int, char**)':
examples/main/main.cpp:247:31: error: invalid 'static_cast' from type 'main(int, char**)::<lambda(DWORD)>' to type 'PHANDLER_ROUTINE' {aka 'int (__attribute__((stdcall)) *)(long unsigned int)'}
  247 |         SetConsoleCtrlHandler(static_cast<PHANDLER_ROUTINE>(console_ctrl_handler), true);
      |                               ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
make: *** [Makefile:205: main] Error 1

[DannyDaemonic]

Others have used w64devkit without compilation issues. I do wonder why the fortran version causes this issue. Unless it's not the fortran version that's the problem but the build version. Are you using an older release?

The obvious fix is just to pull the lambda function out. I can make a PR but you'll need to test it.

[Valikzant]

Others have used w64devkit without compilation issues. I do wonder why the fortran version causes this issue. Unless it's not the fortran version that's the problem but the build version. Are you using an older release?

The obvious fix is just to pull the lambda function out. I can make a PR but you'll need to test it.

Hi @DannyDaemonic. I used "w64devkit-i686-fortran-1.19.0.zip" to try build it. It is a latest version i found. If i did something wrong or i need to use another version, please correct me ':D

[DannyDaemonic]

I don't think it's you. It's just a mingw thing. Can you try out the patch in #1513 and reply in that pull request if it fixes this or not? If you don't know how to check it out you can download it from here and apply it like this: patch -p1 < 1513.patch

[Valikzant]

@DannyDaemonic. I took the time to test and tried to build the original llama.cpp, choosing a different version of fortran. It looks like I wasn't too careful when choosing the version, looking for the first 'fortran' word I could find and installing -686, which I looked on the website is the version for older computers. Using the version without -686 I encountered the following result at the end of the command line:
==== Run ./main -h for help. ====
which I understood to be a successful installation, without errors.
I apologize for my inattention, because the installation guide did not tell me which fortan to install and I chose the first one I saw. ':D
My problem solved by using w64devkit-fortran-1.19.0.zip
but w64devkit-i686-fortran-1.19.0.zip does not work for me with patch and without patch, so it is was my mistake by choosing this verison.

[Green-Sky]

i686

Oh yea, that's 32bit x86

[DannyDaemonic]

I guess i686 probably just isn't maintained as well. If you're going to use mingw you will need #1462 (patch) that is verified to fix an input issue.