explicitly export LC_ALL for sphinx builds

[valeriupredoi]

Description

Investigations done in
#2706
our builds suffering from this issue sphinx-doc/sphinx#11739 - no idea why it's reared its head only now, and why it's not manifesting itself in ESMValTool

Docs: https://esmvaltool--2708.org.readthedocs.build/projects/ESMValCore/en/2708/

Closes #2705 #2706

---

Before you get started

• ☝ Create an issue to discuss what you are going to do

Checklist

It is the responsibility of the author to make sure the pull request is ready to review. The icons indicate whether the item will be subject to the 🛠 Technical or 🧪 Scientific review.

• 🛠 This pull request has a descriptive title and labels
• 🛠 All checks below this pull request were successful

---

To help with the number pull requests:

• 🙏 We kindly ask you to review two other open pull requests in this repository

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 95.10%. Comparing base (e2f4d60) to head (494119f).
⚠️ Report is 61 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #2708      +/-   ##
==========================================
- Coverage   95.11%   95.10%   -0.01%     
==========================================
  Files         258      258              
  Lines       15028    15028              
==========================================
- Hits        14294    14293       -1     
- Misses        734      735       +1     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[valeriupredoi]

this is unfortunately not the solution

setting it to LC_ALL=C looks to be doing the part

[bouweandela]

Yes, I would have been surprised if python -m locale actually changed the locale settings. It's just supposed to list them.

[valeriupredoi]

well, here goes me then - I reset all the bits like it used to be before I left for EGU (when it was passing many times in a row), and now it fails - it's like the locale gets stuck in a state where once stuck it's hard to dislodge afterwards. Am out of bullets, but I really don't want us to pin to exclude the latest Python 3.13.3 unless we know for sure that version messes up the locale

[valeriupredoi]

@bouweandela @schlunma I think this finally does it for RTD: 3 green runs in a row! Make that 4 greens in a row - smells like an actual fix 🥳 BTW Bouwe, I've seen issues reported for Python 3.13 and locale - they deprecated funcs like getdefaultlocale() and those things are breaking some stuff

[valeriupredoi]

whoa 6 greens in a row! Dis solution, kolega 🥳

[bouweandela]

But you said two weeks ago that it worked with just setting language=en without the locale setting. Are the new changes even needed? #2708 (comment)

[bouweandela]

I would be really keen to understand what is going on here, so we can actually fix the issue in cf_units (if it turns out that that is the problem)

[valeriupredoi]

I would be really keen to understand what is going on here, so we can actually fix the issue in cf_units (if it turns out that that is the problem)

absolutely! I propose we merge this to have an ad-hoc fix so our tests pass, then I will start looking into the real cause; - I'll start by isolating the issue to cf_units (which I'm not even 100% is the issue) 🍺

[valeriupredoi]

have kicked off (more) investigations in #2706

[valeriupredoi]

@bouweandela I think I finally cracked it! And by the looks of it, it's neither iris nor cf-units, nor any combination of - see #2706 (comment) - it's just Python: 3.13.3 changes the locale to ('en_US', 'UTF-8') which is not liked at all by sphinx, whereas 3.12.2 has that to ('C', 'UTF-8') which is always liked by sphinx. So this is why the fix in this PR will always work since it reverts the Python locale to something that 3.13.2 had and had no issues with it. Why 3.13.3 changes that to that and why that is not liked by sphinx is completely beyond me, but I can 100% sure say this is the fix we need, or pin to Python (but am not keen on that)

[bouweandela]

Thanks!