Add support for RISC-V repository

.github/workflows/docs.yml

@@ -60,6 +60,13 @@ jobs:
           # Merge all these results together
           python scripts/merge_data_files.py out.yaml eessi*.yaml
           mv out.yaml docs/data/eessi_software_metadata.yaml
+          # Also do RISC-V repository
+          rm eessi*.yaml
+          export EESSI_ARCHDETECT_OPTIONS_OVERRIDE="riscv64/generic"
+          export EESSI_ACCELERATOR_TARGET_OVERRIDE="doesnotexist"
+          ( module load EESSI/2025.06 && module load EasyBuild/5 && python generate_data_files.py --eessi-version 2025.06-001 --repository "dev.eessi.io/riscv" )
+          python scripts/merge_data_files.py out_riscv.yaml eessi*.yaml
+          mv out_riscv.yaml docs/data/eessi_software_metadata_riscv.yaml
           # Generate json data files and markdown index/description for them
           cd docs/data
           python ../../scripts/process_eessi_software_metadata.py eessi_software_metadata.yaml eessi_api_metadata

scripts/generate_data_files.py

@@ -18,7 +18,16 @@
 from easybuild.tools.include import include_easyblocks
 from contextlib import contextmanager
 
-VALID_EESSI_VERSIONS = ["2025.06", "2023.06"]
+SUPPORTED_REPOSITORIES = {
+        'software.eessi.io': ["2025.06", "2023.06"],
+        'dev.eessi.io/riscv': ["2025.06-001"],
+}
+
+VALID_EESSI_VERSIONS = [
+    version
+    for versions in SUPPORTED_REPOSITORIES.values()
+    for version in versions
+]
 
 # Give order to my toolchains so I can easily figure out what "latest" means
 EESSI_SUPPORTED_TOP_LEVEL_TOOLCHAINS = OrderedDict(
@@ -205,20 +214,31 @@ def merge_dicts(d1, d2):
         "-e",
         required=True,
         choices=VALID_EESSI_VERSIONS,
-        help=f"Allowed versions: {', '.join(VALID_EESSI_VERSIONS)}",
+        help=f"Allowed versions (also dependent on repository): {', '.join(VALID_EESSI_VERSIONS)}",
+    )
+    parser.add_argument(
+        "--repository",
+        "-r",
+        default="software.eessi.io",
+        choices=SUPPORTED_REPOSITORIES.keys(),
+        help=f"Repository to scan: {', '.join(SUPPORTED_REPOSITORIES)} (default: %(default)s)",
     )
 
     args = parser.parse_args()
     eessi_version = args.eessi_version
+    repository = args.repository
+
+    if eessi_version not in SUPPORTED_REPOSITORIES[repository]:
+        raise ValueError(f"You must choose an EESSI version supported by the repository: {SUPPORTED_REPOSITORIES[repository]}")
 
     print(f"Using EESSI version: {eessi_version}")
 
     # We use a single architecture path to gather information about the software versions
     eessi_reference_architecture = os.getenv("EESSI_ARCHDETECT_OPTIONS_OVERRIDE", False)
     if not eessi_reference_architecture:
-        print("You must have selected a CPU architecture via EESSI_ARCHDETECT_OPTIONS_OVERRIDE")
+        print("You must have selected a CPU architecture via EESSI_ARCHDETECT_OPTIONS_OVERRIDE environment variable")
         exit()
-    base_path = f"/cvmfs/software.eessi.io/versions/{eessi_version}/software/linux/{eessi_reference_architecture}"
+    base_path = f"/cvmfs/{repository}/versions/{eessi_version}/software/linux/{eessi_reference_architecture}"
     cpu_easyconfig_files_dict = collect_eb_files(os.path.join(base_path, "software"))
     # We also gather all the acclerator installations for NVIDIA-enabled packages
     # We're not typically running this script on a node with a GPU so an override must have been set
@@ -243,7 +263,7 @@ def merge_dicts(d1, d2):
 
     # Store the toolchain hierarchies supported by the EESSI version
     eessi_software["eessi_version"][eessi_version]["toolchain_hierarchy"] = {}
-    for top_level_toolchain in EESSI_SUPPORTED_TOP_LEVEL_TOOLCHAINS[eessi_version]:
+    for top_level_toolchain in EESSI_SUPPORTED_TOP_LEVEL_TOOLCHAINS[eessi_version.split("-")[0]]:
         # versions are typically 2024a/2024b etc. for top level toolchains
         # so let's use that to make sorting easy
         toolchain_family = f"{top_level_toolchain['version']}_{top_level_toolchain['name']}"

scripts/process_eessi_software_metadata.py

@@ -24,6 +24,10 @@
     "x86_64/intel/cascadelake",
 ]
 
+RISCV_ARCHITECTURES = [
+    "riscv64/generic",
+]
+
 NVIDIA_ARCHITECTURES = [
     "accel/nvidia/cc70",
     "accel/nvidia/cc80",
@@ -63,7 +67,11 @@ def get_software_information_by_filename(file_metadata, original_path=None, tool
     # 1) Detect the architecture substring inside the path
     base_version_dict["cpu_arch"] = []
     detected_arch = None
-    for arch in ARCHITECTURES:
+    if '/riscv64/' in original_path:
+        architecture_group = RISCV_ARCHITECTURES
+    else:
+        architecture_group = ARCHITECTURES
+    for arch in architecture_group:
         if f"/{arch}/" in original_path:
             detected_arch = arch
             break
@@ -89,7 +97,7 @@ def get_software_information_by_filename(file_metadata, original_path=None, tool
     spider_cache = before_arch + detected_arch + "/.lmod/cache/spiderT.lua"
 
     # 3) Substitute each architecture and test module file existence in spider cache
-    for arch in ARCHITECTURES:
+    for arch in architecture_group:
         substituted_modulefile = modulefile.replace(detected_arch, arch)
         substituted_spider_cache = spider_cache.replace(detected_arch, arch)
         # os.path.exists is very expensive for CVMFS so we just look for the file in the spider cache
@@ -368,7 +376,7 @@ def main():
     base_json_metadata["architectures_map"] = {}
     for eessi_version in eessi_versions:
         base_json_metadata["architectures_map"][eessi_version] = {}
-        for architecture in ARCHITECTURES:
+        for architecture in ARCHITECTURES + RISCV_ARCHITECTURES:
             base_json_metadata["architectures_map"][eessi_version][architecture] = architecture
     base_json_metadata["gpu_architectures_map"] = {}
     base_json_metadata["category_details"] = {}