Add CI scripts and setup files to automate installation

[Bill94l]

Hi @Dolu1990,

• Added Makefile for build and test orchestration.
• Added multiple CI scripts for environment setup and dependency installation:
  • build_regress_simulator.sh
  • build_spike_rvls.sh
  • clone-submodules.sh
  • install-elfio.sh
  • install-libsdl.sh
  • install-openjdk.sh
  • install-riscv-gnu-toolchain.sh
  • install-rvls.sh
  • install-sbt.sh
  • install-spike-spinalhdl.sh
  • install-verilator.sh
• Added README.md in src/test for installation and test-related documentation.

Thanks

[Dolu1990]

Why removing ADDCFLAGS += -CFLAGS -lSDL2 ?
It seems it is the reason why the CI fail ?

[Bill94l]

to include this ADDCFLAGS += -LDFLAGS -L$(realpath ${SPIKE_LIB})
For me it works fine without ADDCFLAGS += -CFLAGS -lSDL2
I can put it back if that's the problem.

[Dolu1990]

I see in the CI logs :
main.cpp:(.text+0x406e): undefined reference to SDL_Init' /usr/bin/ld: main.cpp:(.text+0x4099): undefined reference to SDL_CreateWindow'
/usr/bin/ld: main.cpp:(.text+0x40aa): undefined reference to SDL_CreateRenderer' /usr/bin/ld: main.cpp:(.text+0x40c4): undefined reference to SDL_CreateTexture'

Seems it is related ^^

[Bill94l]

These modifications aim to eliminate the need for sudo by sourcing libsdl and elfio locally.

[Dolu1990]

Thanks :D

[Bill94l]

Hi @Dolu1990 ,
I don't have full visibility into how your current continuous integration (CI) pipeline is set up. However, with the changes I made:

1. Complete installation:
   The full installation can now be executed with the following commands:

# Switch to the appropriate branch
git checkout rvls-update  

# Perform the full installation
make install

This command installs the core, toolchain, simulators, and tests in one step.

2. Library inclusion:
   The libraries ELFIO and SDL are now included directly from ext/include and ext/lib. This avoids requiring sudo for installation.
3. Command differences:
   I’ve noticed a discrepancy between the command I use locally and the one executed in the CI pipeline:

• Local Command:

ccache g++ -I. -MMD -I/usr/local/share/verilator/include -I/usr/local/share/verilator/include/vltstd \
-DVM_COVERAGE=0 -DVM_SC=0 -DVM_TRACE=1 -DVM_TRACE_FST=1 -faligned-new -fcf-protection=none \
-Wno-bool-operation -Wno-sign-compare -Wno-uninitialized -Wno-unused-but-set-variable \
-Wno-unused-parameter -Wno-unused-variable -Wno-shadow -std=c++14 -Iobj_dir \
-INaxRiscv/ext/riscv-isa-sim/riscv -INaxRiscv/ext/riscv-isa-sim/fesvr -INaxRiscv/ext/riscv-isa-sim/softfloat \
-INaxRiscv/ext/riscv-isa-sim/build -INaxRiscv/ext/include -g -rdynamic -O3 -DTRACE -std=gnu++14 \
-c -o VNaxRiscv__Trace__7__Slow.o VNaxRiscv__Trace__7__Slow.cpp

• CI Pipeline Command:

g++  -I.  -MMD -I/home/runner/tools/share/verilator/include -I/home/runner/tools/share/verilator/include/vltstd - \
DVM_COVERAGE=0 -DVM_SC=0 -DVM_TRACE=1 -DVM_TRACE_FST=1 -faligned-new -fcf-protection=none -Wno-bool- \
operation -Wno-sign-compare -Wno-uninitialized -Wno-unused-but-set-variable -Wno-unused-parameter -Wno-unused-\
variable -Wno-shadow     -std=c++14 -Iobj_dir -I/home/runner/work/NaxRiscv/NaxRiscv/ext/riscv-isa-sim/riscv -I/home/\
runner/work/NaxRiscv/NaxRiscv/ext/riscv-isa-sim/fesvr -I/home/runner/work/NaxRiscv/NaxRiscv/ext/riscv-isa-sim/\
softfloat -I/home/runner/work/NaxRiscv/NaxRiscv/ext/riscv-isa-sim/build -I -g -rdynamic -O3 -DTRACE  -std=gnu++14 \
 -c -o VNaxRiscv__Trace__7__Slow.o VNaxRiscv__Trace__7__Slow.cpp
echo "" > VNaxRiscv__ALL.verilator_deplist.tmp

4. Key differences:

• In my local command, I explicitly include the path to the ext/include directory, which contains SDL2 and ELFIO (-INaxRiscv/ext/include).
• The CI command does not specify this path, which might cause issues with locating these libraries during the build process.

Could you confirm if the ext/include path is missing in the CI configuration, or if there is a difference in how dependencies are installed? Aligning this detail should resolve potential discrepancies.

[Dolu1990]

Hi,

Could you confirm if the ext/include path is missing in the CI configuration

Which configuration ?

I don't have full visibility into how your current continuous integration (CI) pipeline is set up. However, with the changes I made:

There is no specific setup. All comes from this repo. This is vanilla github actions : https://github.com/SpinalHDL/NaxRiscv/blob/main/.github/workflows/scala.yml#L15
I do not host any runner.

[Bill94l]

Hi,

Which configuration ?

The issue was that the ext/include path, which contains ELFIO and SDL, was missing between my local command and the pipeline command. To address this, I modified the scala.yml workflow to run make install from my Makefile. I acknowledge that reinstalling the riscv-gnu-toolchain locally during each pipeline execution takes considerable time, but if this approach works, we can focus on optimizing it later.

[Dolu1990]

Weird, i don't know what is happening with the github action :

2025-01-30T17:35:10.4285432Z Requested labels: ubuntu-22.04
2025-01-30T17:35:10.4285715Z Job defined at: SpinalHDL/NaxRiscv/.github/workflows/scala.yml@refs/pull/130/merge
2025-01-30T17:35:10.4285843Z Waiting for a runner to pick up this job...
2025-01-30T17:35:11.1849365Z Job is waiting for a hosted runner to come online.
2025-01-30T17:35:14.5882591Z Job is about to start running on the hosted runner: GitHub Actions 6 (hosted)

Did you enabled some sort of user hosted runner ?

[Bill94l]

No, I actually added the stage that runs regression tests with RVLS to the pipeline. However, the pipeline crashed due to insufficient memory, caused by the large Spike.log file generated by RVLS. My initial plan was to disable logging and restart the pipeline, but I haven't yet found a way to disable Spike.log without crashing the system. This is the message I received when the pipeline failed.

[Bill94l]

Hi ^^

We can confirm that the version enabling regression tests through SoCSim and RVLS is fully functional !

In the GitHub workflow pipeline, I first run regression tests via NaxRiscvRegression, followed by regression tests via NaxRiscvRvls , specifically for the RV64IMAFDC configuration with a Linux boot. The other configurations are also functional in NaxRiscvRvls , but due to time constraints—the total execution time for all configurations exceeding the 6-hour (360-minute) limit imposed by the GitHub workflow—they are commented out for the GitHub workflow run.

[Dolu1990]

Nice :D
Thanks !