Fix self-hosted runner labels for CUDA CI job

[aaronj0]

The prepare-dell job needs the 'spotter' label to target the always-on runner, otherwise the wake-onlan packet never reaches the ip. The build matrix entry also needs [self-hosted, cuda, heavy] to route to the correct machine once it's awake. adapted from the working clad CI configuration

[vgvassilev]

LGTM!

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 79.75%. Comparing base (5f757f7) to head (393289a).
⚠️ Report is 3 commits behind head on main.

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #864   +/-   ##
=======================================
  Coverage   79.75%   79.75%           
=======================================
  Files          11       11           
  Lines        4031     4031           
=======================================
  Hits         3215     3215           
  Misses        816      816           

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[mcbarton]

  sudo apt-get update (don't need)
  sudo apt-get install valgrind ninja-build (dont need to run test under Valgrind on self hosted runners)
  sudo apt-get install git g++ debhelper devscripts gnupg python3 doxygen graphviz python3-sphinx (most documentation stuff related. probably have python already installed)
  sudo apt-get install -y libc6-dbg (probably already installed on runner)
  sudo apt-get install valgrind (don't need valgrind. also this is repeated and should be removed for github hosted runners too) 
  sudo apt autoremove (not needed on self hosted runner)
  sudo apt clean (no needed on self hosted runner)
  # Install libraries used by the cppyy test suite
  sudo apt install libeigen3-dev (depends on whether self hosted runners have boost or eigen installed already. also don't need to test these on self hosted runners, since they already tested on github hosted runners)
  sudo apt install libboost-all-dev

@aaronj0 @vgvassilev This is what it tries to install when it fails. I have commented next to each one. I'd just skip the step altogether based on what i see.

[aaronj0]

@vgvassilev Can we install ninja on the machine?

Updating files: 100% (160123/160123), done.
CMake Error: CMake was unable to find a build program corresponding to "Ninja".  CMAKE_MAKE_PROGRAM is not set.  You probably need to select a different build tool.
CMake Error: CMAKE_C_COMPILER not set, after EnableLanguage
`` 

[vgvassilev]

We can install ninja if that's required...

[mcbarton]

We can install ninja if that's required...

Its not strictly needed for the self hosted runner. Its just useful. Its used in the building of llvm. We could have the default generator set to ninja, and switch it to makefiles for the self hosted runner. It doesn't build any quicker on the runner (at least on the Github hosted runners) using Ninja. Using Ninja just results in a smaller llvm cache.

[aaronj0]

The current error is due to the [[maybe_unused]] attribute on the member variable:

CppInterOp/lib/CppInterOp/Compatibility.h

Line 427 in 613a40b

[[maybe_unused]] Interpreter* m_Interpreter;

which is not supported on gcc11. @vgvassilev Does the self host machine have gcc12?

[vgvassilev]

The current error is due to the [[maybe_unused]] attribute on the member variable:

CppInterOp/lib/CppInterOp/Compatibility.h

Line 427 in 613a40b

[[maybe_unused]] Interpreter* m_Interpreter;

which is not supported on gcc11. @vgvassilev Does the self host machine have gcc12?

No. You can try installing it via apt...

[mcbarton]

@aaronj0 this PR only needs to be one commit. can you squash all the commits into one before any merging?

[aaronj0]

@aaronj0 this PR only needs to be one commit. can you squash all the commits into one before any merging?

I've condensed it into 1) updating the runner labels and 2) llvm/gcc dependency installation and handling. Merging as the emscripten jobs are unrelated.