Docker: clean up after setup

[abeimler]

Hello,
I'm using setup-cpp in my Dockerfile and saw the generated Layer is big (when installing a lot of tools).

Problem

16.3 GB

(it's hard to upload something this big to docker hub)

Dockerfile

## base image
FROM ubuntu:latest as base

ARG extra_libraries

ENV DEBIAN_FRONTEND noninteractive
RUN apt-get update -y && apt-get install --no-install-recommends -y gnupg software-properties-common
# Install packages available from standard repos
RUN apt-get update -y && \
    apt-get install --no-install-recommends -yq \
        wget curl pkg-config zip unzip tar git && \
    apt-get clean && rm -rf /var/lib/apt/lists/*

# add setup_cpp
RUN wget --no-verbose "https://github.com/aminya/setup-cpp/releases/download/v0.13.1/setup_cpp_linux"
RUN chmod +x ./setup_cpp_linux
RUN ./setup_cpp_linux --compiler llvm --cmake true --ninja true --ccache true --clangformat true --clangtidy true

I add rm -rf /tmp/* && apt-get clean && rm -rf /var/lib/apt/lists/* for clean up and got: 14.3 GB

RUN ./setup_cpp_linux --compiler llvm \
    --cmake true --ninja true --ccache true \
    --cppcheck true --clangtidy true --clangformat true \
    --gcovr true --conan true --doxygen true && \
    rm -rf /tmp/* && apt-get clean && rm -rf /var/lib/apt/lists/*

(cleaning up the /tmp-folder seem to a bit help ?)

different setups

Readme example

RUN ./setup_cpp_linux --compiler llvm --cmake true --ninja true --ccache true --vcpkg true

gcc + conan

RUN ./setup_cpp_linux --compiler gcc --cmake true --ninja true --ccache true --conan true

Layer Size: 303.4 MB

with clang tools

idk why but the clang-tools seems huge after setup.

RUN ./setup_cpp_linux --compiler llvm --cmake true --ninja true --ccache true --conan true --cppcheck true --clangtidy true --clangformat true

Layer Size: 16 GB

with clangtidy

RUN ./setup_cpp_linux --compiler llvm --cmake true --ninja true --ccache true --clangtidy true

Layer Size: 11 GB

with clangformat

RUN ./setup_cpp_linux --compiler llvm --cmake true --ninja true --ccache true --clangformat true

Layer Size: 11 GB

My old setup

apt-get install --no-install-recommends -yq 
          gcc wget curl pkg-config zip unzip tar git build-essential &&     
apt-get install --no-install-recommends -y
          clang binutils bison
          python3 python3-pip doxygen graphviz         
          cmake make ninja-build ccache cppcheck cmake-extras   
          clang-tidy clang-format    
          valgrind gcovr linux-tools-common linux-tools-generic google-perftools         
          neovim emacs nano &&     
apt-get clean && rm -rf /var/lib/apt/lists/*

My old version with only apt-get setup was 1.1 GB (Layer Size)

Directory sizes in (docker) image

RUN du -sh /root/cmake
RUN du -sh /root/ninja
RUN du -sh /root/clangtidy
RUN du -sh /root/clangformat
RUN du -sh /root/llvm

Output

141M    /root/cmake
224K    /root/ninja
4.4G    /root/clangtidy
4.4G    /root/clangformat
4.4G    /root/llvm

Seems like llvm, clangtidy and clangformat are the same.
when installing llvm, and clang-tools, clang+llvm-14.0.0-x86_64-linux-gnu-ubuntu-18.04.tar.xz gets downloaded multiple times. (and extract into /root/llvm, /root/clang... ?)

[aminya]

• We should use the same path for clang-tools. It seems that the caching system is not working for llvm inside Docker that it re-installs it. fix: use the same path for llvm, clangformat, clangtidy #62

However, keep in mind that this image is a dev/build image. You should only use it for building once in your CI system.
Once the building is done, to run your app, use a distroless image.
https://github.com/GoogleContainerTools/distroless

FROM ubuntu:devel as builder
# ... setup-cpp

# then build your app (e.g. using CMake)

FROM gcr.io/distroless/cc
COPY --from=builder /build/your_binary /
COPY --from=builder /build/your_dll_files /
CMD ["./your_binary"]

• I can add an example docker file that uses distroless images. docker: add building and running environments to the dockers #63

[aminya]

In the meantime, remove the extra llvm flags like --clangtidy, --clangformat. Inside GitHub Actions, these are re-used but in a local installation, the caching doesn't work. #60

[abeimler]

We should use the same path for clang-tools.

Yes I agree, --clangtindy and --clangformat are basically aliases for llvm, clang-tidy is already included in llvm.

[aminya]

0.13.2 is released that fixes this issue