use versioned symbols in ncurses built with system toolchain (by adding --with-versioned-syms configure option)

[Micket]

(created using eb --new-pr)
fixes #16014
but i still have a bunch of questions; since i'm not familiar with what versioned syms actually does.

1. do i need to rebuild all the stuff that depends on ncurses-6.x.eb after this change?
2. should we also change the ncurses we build for each toolchain as well?

[Micket]

Test report by @Micket
SUCCESS
Build succeeded for 4 out of 4 (4 easyconfigs in total)
alvis-c1 - Linux Rocky Linux 8.5, x86_64, Intel Xeon Processor (Skylake), Python 3.6.8
See https://gist.github.com/81e1a4baca6e510f0c8ff204faaa2bbe for a full test report.

[boegel]

(created using eb --new-pr) fixes #16014 but i still have a bunch of questions; since i'm not familiar with what versioned syms actually does.

1. do i need to rebuild all the stuff that depends on ncurses-6.x.eb after this change?

I think strictly speaking the answer is "yes" here (if you did a forced re-installation of ncurses that is configured with --with-versioned-syms), since it results in no longer exposing (internal) symbols (according to this), but in practice it probably won't matter (since stuff building on top of ncurses should not be using the internal symbols).

2. should we also change the ncurses we build for each toolchain as well?

Based on the problems reported in #16014, we should be using --with-versioned-syms in all ncurses easyconfigs...
Let's start with these 4 though, and try and get a test report for these easyconfigs + stuff that depends on them (like libreadline, gettext, CMake, Lua, htop, tmux)

[boegel]

@boegelbot please test @ generoso

[boegelbot]

@boegel: Request for testing this PR well received on login1

PR test command 'EB_PR=16064 EB_ARGS= /opt/software/slurm/bin/sbatch --job-name test_PR_16064 --ntasks=4 ~/boegelbot/eb_from_pr_upload_generoso.sh' executed!

• exit code: 0
• output:

Submitted batch job 9048

Test results coming soon (I hope)...

- notification for comment with ID 1224269093 processed

Message to humans: this is just bookkeeping information for me,
it is of no use to you (unless you think I have a bug, which I don't).

[boegelbot]

Test report by @boegelbot
SUCCESS
Build succeeded for 4 out of 4 (4 easyconfigs in total)
cns1 - Linux Rocky Linux 8.5, x86_64, Intel(R) Xeon(R) CPU E5-2667 v3 @ 3.20GHz (haswell), Python 3.6.8
See https://gist.github.com/9d8d0f4ee89efad86d2dc166faada12a for a full test report.

[Micket]

@boegel merge?

Also, we should probably start doing this to at least all the new ncurses easyconfigs we add?

[verdurin]

Test report by @verdurin
SUCCESS
Build succeeded for 4 out of 4 (4 easyconfigs in total)
easybuild-c7.novalocal - Linux CentOS Linux 7.9.2009, x86_64, Intel Xeon Processor (Skylake, IBRS), Python 3.6.8
See https://gist.github.com/a8e167f2a9f455c1013baf32ecda496e for a full test report.

[verdurin]

Looks fine.

[verdurin]

Going in, thanks @Micket!

[boegel]

I did some checking whether this would cause trouble for stuff built on top of ncurses, but as far as I can tell it's fine (for the stuff we build with system toolchain and that has a direct dependency declared on ncurses), no matter whether that software is rebuilt or not after ncurses was rebuilt with the --with-versioned-syms included.

I meant to upload test reports to confirm that, but didn't get around to it yet.

Also, we should probably start doing this to at least all the new ncurses easyconfigs we add?

Yes, and probably also to recent existing ncurses easyconfigs... I'll make a separate PR for that.