Workflows: switch to a conda install of clang-format

[bartgol]

Allows us to pick the version without relying on what is available on the runner.

[BFB]

---

[mahf708]

not the recommended way to deal with conda tools... see "install python deps" here for a better example:

E3SM/.github/workflows/e3sm-gh-pages.yml

Lines 54 to 66 in 379e000

	- name: Install Python deps
	uses: mamba-org/setup-micromamba@v2
	with:
	init-shell: bash
	environment-name: docs
	create-args: >-
	python=3.12
	mkdocs-material
	pymdown-extensions
	mkdocs-monorepo-plugin
	mdutils
	mkdocs-bibtex==2.18.0
	marp-cli

note how in that workflow you neve have to activate the env ... (it's all done for you)

[mahf708]

also pls add workflow path itself to this section

on:
  pull_request:
    branches: ["master"]
    paths:
      - 'components/eamxx/**/*.cpp'
      - 'components/eamxx/**/*.hpp'
+     - .github/workflows/eamxx-gh-clang-format.yml

[bartgol]

The micromamba version worked great, with no extra verbosity in the shell. See this run on my fork.

[mjschmidt271]

Looks great--nice work!

[mahf708]

you may need something like to make it work better

  Build-and-Deploy-docs:
    if: ${{ github.repository == 'E3SM-Project/E3SM' }}
    defaults:
      run:
        shell: bash -leo pipefail {0}

and also, you can drop the "conda activate" lines later

[bartgol]

you may need something like to make it work better

  Build-and-Deploy-docs:
    if: ${{ github.repository == 'E3SM-Project/E3SM' }}
    defaults:
      run:
        shell: bash -leo pipefail {0}

and also, you can drop the "conda activate" lines later

Good call! The workflow was actually still running clang-format-18 (from the OS!). I am testing this on my fork now.

[bartgol]

Ok, seems to run as expected (see here).

[mahf708]

I defer to @mjschmidt271 on the version desired

[bartgol]

FYI, I picked 20.1.8 because...it ran! The default version we have in master (18, from the runner OS) was getting a segfault when using the new settings in #7666. Since we don't have any real reason to pick one version or another (except that it must be >=17, so that those new options will work), I just used whatever my 1st attempt with conda install gave me, which happened to be 20.1.8 and ran correctly.

If you guys have any reason to prefer other versions, I can change (or we can update in the future).

[mjschmidt271]

v20.1.8 is just fine with me. I don't really have any great attachment to a particular version as long as it works and remains consistent for the long-medium term