ansible: install Clang on RHEL hosts

[richardlau]

Refs: #4091

---

Deployment status:

• test-digitalocean-rhel8-x64-1
• test-ibm-rhel8-x64-1
• test-ibm-rhel8-x64-2
• test-ibm-rhel8-x64-3
• test-digitalocean-rhel9-x64-1
• test-ibm-rhel9-x64-1
• test-ibm-rhel9-x64-2
• test-digitalocean-ubi81_container-x64-1
• test-digitalocean-ubi81_container-x64-2
• test-ibm-ubi81_container-x64-1
• test-equinix-rhel8_container-arm64-1
• test-equinix-rhel8_container-arm64-2
• test-osuosl-rhel8_container-arm64-1
• test-ibm-rhel8-s390x-1
• test-ibm-rhel8-s390x-2
• test-ibm-rhel8-s390x-3
• test-ibm-rhel8-s390x-4
• test-linuxonecc-rhel9-s390x-1
• test-linuxonecc-rhel9-s390x-2
• test-linuxonecc-rhel9-s390x-3
• test-linuxonecc-rhel9-s390x-4
• test-osuosl-rhel8-ppc64_le-1
• test-osuosl-rhel8-ppc64_le-2
• test-osuosl-rhel8-ppc64_le-3
• test-osuosl-rhel8-ppc64_le-4
• test-osuosl-rhel9-ppc64_le-1
• test-osuosl-rhel9-ppc64_le-2
• test-osuosl-rhel9-ppc64_le-3
• test-osuosl-rhel9-ppc64_le-4
• release-digitalocean-rhel8-x64-1
• release-ibm-rhel8-s390x-1
• release-ibm-rhel8-x64-1
• release-ibm-rhel8-x64-2
• release-osuosl-rhel8-arm64-1
• release-osuosl-rhel8-ppc64_le-1

[richardlau]

Test CIs

• https://ci.nodejs.org/job/richardlau-node-test-commit-arm/1/ ✔️
• https://ci.nodejs.org/job/richardlau-node-test-commit-linux-containered/90/nodes=ubi81_sharedlibs_openssl111fips_x64/ ✔️
• https://ci.nodejs.org/job/richardlau-node-test-commit-linux/25/ ✔️
• https://ci.nodejs.org/job/richardlau-node-test-commit-linuxone/33/ ❌
• https://ci.nodejs.org/job/richardlau-node-test-commit-plinux/10/ ❌

[richardlau]

Failures:

• https://ci.nodejs.org/job/richardlau-node-test-commit-plinux/10/nodes=rhel8-ppc64le/console

14:54:41 clang: error: unknown argument: '-mminimal-toc'

• https://ci.nodejs.org/job/richardlau-node-test-commit-linuxone/33/nodes=rhel9-s390x/console

15:19:17 <inline asm>:7:13: error: invalid register
15:19:17     7 |   stmg %r6, %sp, 48(%sp)                           
15:19:17       |             ^
15:19:17 <inline asm>:8:7: error: invalid register
15:19:17     8 |   lay %sp, -160(%sp)                               
15:19:17       |       ^
15:19:17 <inline asm>:10:16: error: invalid register
15:19:17    10 |   lay %r4, 208(%sp)                                
15:19:17       |                ^
15:19:17 <inline asm>:12:12: error: invalid register
15:19:17    12 |   lmg %r14,%sp, 272(%sp)                           
15:19:17       |            ^

[richardlau]

For Linux ppc64le Node.js is using a gcc-only flag which we'll need to make conditional. PR: nodejs/node#59484
New CI: https://ci.nodejs.org/job/richardlau-node-test-commit-plinux/13/ ❌

For Linux s390x we'll need to cherry-pick (list via @miladfarca ):

• https://chromium-review.googlesource.com/c/v8/v8/+/6603554
• https://chromium-review.googlesource.com/c/v8/v8/+/6597365

PR: nodejs/node#59485
New CI: https://ci.nodejs.org/job/richardlau-node-test-commit-linuxone/34/ ✔️

[richardlau]

For Linux ppc64le Node.js is using a gcc-only flag which we'll need to make conditional. PR: nodejs/node#59484 New CI: https://ci.nodejs.org/job/richardlau-node-test-commit-plinux/13/ ❌

Single executable application tests are crashing in https://ci.nodejs.org/job/richardlau-node-test-commit-plinux/13/nodes=rhel8-ppc64le/
I don't know if this is clang specific -- those tests have a history of crashing on Linux ppc64le with gcc and are currently skipped on

• v22.x [v22.x] test: skip broken sea on rhel8 node#58914
• v20.x [v20.x] test: skip broken sea on rhel8 node#58761

I think we should land nodejs/node#59484 anyway as a separate fix.

[richardlau]

I've opened nodejs/node#59561 to track the single executable application test failures and nodejs/node#59563 to skip the tests for now on Linux ppc64le on main (as we already do on v20.x/v22.x).

Once that lands we can take this out of draft/wip and land.

[richardlau]

FWIW I ran a build with clang Linux ppc64le build for nodejs/node#59560 and all tests (including the single executable application tests) passed: https://ci.nodejs.org/job/richardlau-node-test-commit-plinux/22/

I think my preferred plan is to:

1. Land the skips (test: skip sea tests on Linux ppc64le node#59563).
2. Land this PR to switch the builds to clang on RHEL 8/9.
3. Re-evaluate the skips (including possibly reverting them) later on if they no longer fail after sea: implement execArgvExtension node#59560.

[richardlau]

I ran a few more clang builds on main today and they both passed

• https://ci.nodejs.org/job/richardlau-node-test-commit-plinux/23/
• https://ci.nodejs.org/job/richardlau-node-test-commit-plinux/25/

they were failing on main a few days ago, so maybe this is related after all to whatever was causing the failures in the same tests for macOS but have since appeared to have cleared up?

In any case, I'm out of office tomorrow and next week, so it's probably safer to stick to the plan in my previous comment and disable the tests on Linux ppc64le for now and I can check the status again in September.