ARM linux CI jobs fail when using torch 2.9 rc wheels

[GregoryComer]

🐛 Describe the bug

ARM Linux CI jobs are failing on the release/1.0 branch when using prebuilt PyTorch + domain packages. This appears to be due to the PyTorch packages requiring glibcxx 3.4.30. The existing ARM Linux images we use have 3.4.29.

This is a known problem with Conda on Ubuntu 22.04 (see pytorch/pytorch#105248) and there is a workaround in install_conda.sh:

executorch/.ci/docker/common/install_conda.sh

Lines 59 to 62 in 181ed4d

	# WARNING: This is a HACK from PyTorch core to be able to build PyTorch on 22.04.
	# Specifically, ubuntu-20+ all comes lib libstdc++ newer than 3.30+, but anaconda
	# is stuck with 3.29. So, remove libstdc++6.so.3.29 as installed by
	# https://anaconda.org/anaconda/libstdcxx-ng/files?version=11.2.0

However, this workaround is not used on ARM Linux because it apparently breaks the CI jobs (introduced in #12068):

executorch/.ci/docker/common/install_conda.sh

Lines 74 to 78 in 181ed4d

	# Hack breaks the job on aarch64 but is still necessary everywhere
	# else.
	if [ "$(uname -m)" != "aarch64" ]; then
	fix_conda_ubuntu_libstdcxx
	fi

ImportError: /opt/conda/envs/py_3.10/bin/../lib/libstdc++.so.6: version `GLIBCXX_3.4.30' not found (required by /opt/conda/envs/py_3.10/lib/python3.10/site-packages/executorch/extension/pybindings/_portable_lib.cpython-310-aarch64-linux-gnu.so)

From https://github.com/pytorch/executorch/actions/runs/18111201043/job/51550387476

We need to figure out what the proper fix is, perhaps updating the libcxx version in the docker image build or somewhere else? As a short-term fix, I'm installing a newer version with Conda in #14677.

Versions

release/1.0 - commit 3854f4b

[larryliu0820]

Is this fixed? Maybe close?