Compile bug: `error: ‘svr’ was not declared in this scope` after #17216

[ddh0]

Git commit

0de8878

Operating systems

Linux

GGML backends

CUDA, CPU

Problem description & steps to reproduce

Since #17216 was merged, compilation of llama-server fails.

First Bad Commit

0de8878

Compile command

rm -rf build
time cmake -B build -DGGML_CUDA=1 -DCMAKE_CUDA_ARCHITECTURES="89" -DLLAMA_OPENSSL=1 -DGGML_CUDA_FA_ALL_QUANTS=1 # -DGGML_CCACHE=0 -DCMAKE_BUILD_TYPE=Debug
time cmake --build build --config Release -j 16

Relevant log output

/home/dylan/llama.cpp/tools/server/server-http.cpp: In member function ‘bool server_http_context::init(const common_params&)’:
/home/dylan/llama.cpp/tools/server/server-http.cpp:52:9: error: ‘svr’ was not declared in this scope
   52 |         svr.reset(
      |         ^~~
/home/dylan/llama.cpp/tools/server/server-http.cpp:57:9: error: ‘svr’ was not declared in this scope
   57 |         svr.reset(new httplib::Server());
      |         ^~~
gmake[2]: *** [tools/server/CMakeFiles/llama-server.dir/build.make:98: tools/server/CMakeFiles/llama-server.dir/server-http.cpp.o] Error 1
gmake[2]: *** Waiting for unfinished jobs....
[100%] Linking CXX executable ../bin/test-backend-ops
[100%] Built target test-backend-ops
gmake[1]: *** [CMakeFiles/Makefile2:4019: tools/server/CMakeFiles/llama-server.dir/all] Error 2
gmake: *** [Makefile:146: all] Error 2

[o7si]

You can temporarily disable -DLLAMA_OPENSSL=0 to ensure compilation passes.

I'm trying to fix it #17348.