GH-46989: [CI][R] Use Ubuntu 20.04 instead of OpenSUSE for R 4.1

[kou]

Rationale for this change

OpenSUSE 15.5 ships old GCC (7.5) that doesn't have enough C++17 support.

What changes are included in this PR?

Use Ubuntu 20.04 that ships GCC 9.3 instead of OpenSUSE 15.5.

Ubuntu 20.04 reached EOL but we can use it for now.

We discussed why we need OpenSUSE 15.5 based job at #45718 (comment) . We have the job because https://arrow.apache.org/docs/developers/cpp/building.html said "gcc 7.1 and higher should be sufficient".

We need require GCC 9 or later with #46813.

Are these changes tested?

Yes.

Are there any user-facing changes?

No.

• GitHub Issue: [CI][C++] test-r-rstudio-r-base-4.1-opensuse155 is failing #46989

[kou]

@github-actions crossbow submit test-r-rstudio-r-base-*

[github-actions]

⚠️ GitHub issue #46989 has been automatically assigned in GitHub to PR creator.

[github-actions]

Revision: c2d4d46

Submitted crossbow builds: ursacomputing/crossbow @ actions-e1fe36eccc

Task	Status
test-r-rstudio-r-base-4.1-noble
test-r-rstudio-r-base-4.2-noble

[kou]

Hmm... They don't work...

https://dev.azure.com/ursacomputing/crossbow/_build/results?buildId=70218&view=logs&j=0da5d1d9-276d-5173-c4c4-9d4d4ed14fdb&t=6c939d89-0d1a-51f2-8b30-091a7a82e98c&l=24235

══ Failed tests ════════════════════════════════════════════════════════════════
── Error ('test-dplyr-filter.R:295:5'): Filtering on a column that doesn't exist errors correctly ──
<validation_error/rlang_error/error/condition>
Error in `not_a_col == 42`: object 'not_a_col' not found
  9. ├─testthat::expect_error(...)
 10. │ └─testthat:::expect_condition_matching(...)
 11. │   └─testthat:::quasi_capture(...)
 12. │     ├─testthat (local) .capture(...)
 13. │     │ └─base::withCallingHandlers(...)
 14. │     └─rlang::eval_bare(quo_get_expr(.quo), quo_get_env(.quo))
 15. ├─Table$create(tbl) %>% mutate(newvar = NOTAVAR + 2)
 16. ├─dplyr::mutate(., newvar = NOTAVAR + 2)
 17. └─arrow:::mutate.ArrowTabular(., newvar = NOTAVAR + 2)
 18.   ├─arrow:::try_arrow_dplyr(...)
 19.   │ ├─base::tryCatch(...)
 20.   │ │ └─base (local) tryCatchList(expr, classes, parentenv, handlers)
 21.   │ │   └─base (local) tryCatchOne(expr, names, parentenv, handlers[[1L]])
 22.   │ │     └─base (local) doTryCatch(return(expr), name, parentenv, handler)
 23.   │ └─base::eval(expr, parent)
 24.   └─arrow:::arrow_eval(exprs[[i]], mask)
 25.     └─base::tryCatch(...)
 26.       └─base (local) tryCatchList(expr, classes, parentenv, handlers)
 27.         └─base (local) tryCatchOne(expr, names, parentenv, handlers[[1L]])
 28.           └─value[[3L]](cond)
 29.             └─arrow:::validation_error(msg, call = expr)
 30.               └─rlang::abort(...)

[ FAIL 2 | WARN 2 | SKIP 30 | PASS 8283 ]

[kou]

@github-actions crossbow submit test-r-rstudio-r-base-*

[github-actions]

Revision: 7d9bf79

Submitted crossbow builds: ursacomputing/crossbow @ actions-c0bd9ba83d

Task	Status
test-r-rstudio-r-base-4.1-focal
test-r-rstudio-r-base-4.2-focal

[kou]

@Ziy1-Tan @pitrou Do you think that we can use GCC 7 compatible implementation for #46813? Or should we require GCC 9 (8?) or later like this PR?

[pitrou]

I'm strongly in favor of requiring GCC 9. We recently reached consensus in favor of switching to C++20, so we will need at least that version anyway.

[amoeba]

+1

This should be fine as far as CRAN goes. Left one comment about splitting this into two PRs for better visibility about the gcc 7.1 -> 9 change.

[kou]

I'll merge this for 21.0.0 release.

[conbench-apache-arrow]

After merging your PR, Conbench analyzed the 4 benchmarking runs that have been run so far on merge-commit 27cca0f.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 30 possible false positives for unstable benchmarks that are known to sometimes produce them.