Managed Python sysconfig returns incorrect path to CC on Amazon Linux 2023 #12901
Description
Summary
This simplest way to reproduce this is to build the following simple Docker image:
FROM amazonlinux:2023
RUN dnf install -y gcc g++ tar
RUN curl -LO https://astral.sh/uv/install.sh \
&& sh install.sh \
&& rm install.sh \
&& source $HOME/.local/bin/env \
&& uv version \
&& uv venv --python 3.10 \
&& source ./.venv/bin/activate \
&& python3 -c 'import sysconfig; print(sysconfig.get_config_var("CC"))'That ultimately logs /usr/bin/aarch64-linux-gnu-gcc, but on this system, the correct path is /usr/bin/aarch64-amazon-linux-gcc.
If we add the following command, we can see that CC indeed points at the wrong path:
cat /root/.local/share/uv/python/cpython-3.10.17-linux-aarch64-gnu/lib/python3.10/_sysconfigdata__linux_aarch64-linux-gnu.py# system configuration generated and used by the sysconfig module
build_time_vars = {
# omitted for brevity...
"CC": "/usr/bin/aarch64-linux-gnu-gcc",
# omitted for brevity...
}To reproduce an actual error that stems from this, build the following Docker image, which installs pygraphviz, which in turn needs to be built with gcc:
FROM amazonlinux:2023
RUN dnf install -y gcc g++ tar
RUN curl -LO https://astral.sh/uv/install.sh \
&& sh install.sh \
&& rm install.sh \
&& source $HOME/.local/bin/env \
&& uv venv --python 3.10 \
&& uv pip install pygraphviz==1.14This ultimately fails with the following:
error: command '/usr/bin/aarch64-linux-gnu-gcc' failed: No such file or directory
Platform
Amazon Linux 2023 aarch64 (Linux 6.10.14-linuxkit aarch64 GNU/Linux)
Version
uv 0.6.14
Python version
3.10.17