Allow tuning max_dupes for build dependencies (#48948)

Up to now, Spack was allowing all build-tools that
may appear in the DAG to have 2 max_dupes.

This is not needed in practice for most of them,
and adding them out of caution just increases
grounding and concretization time.

This PR makes the value of max_dupes configurable
per package, and sets only a few known packages to
2 max_dupes by default.

In case user needs different values, they can
tune the configuration for their use case.

etc/spack/defaults/concretizer.yaml

@@ -43,6 +43,22 @@ concretizer:
     # (e.g. py-setuptools, cmake etc.)
     # "full" (experimental): allows separation of the entire build-tool stack (e.g. the entire "cmake" subDAG)
     strategy: minimal
+    # Maximum number of duplicates in a DAG, when using a strategy that allows duplicates. "default" is the
+    # number used if there isn't a more specific alternative
+    max_dupes:
+      default: 1
+      # Virtuals
+      c: 2
+      cxx: 2
+      fortran: 1
+      # Regular packages
+      cmake: 2
+      gmake: 2
+      py-cython: 2
+      py-flit-core: 2
+      py-setuptools: 2
+      gcc: 2
+      llvm: 2
   # Option to specify compatibility between operating systems for reuse of compilers and packages
   # Specified as a key: [list] where the key is the os that is being targeted, and the list contains the OS's
   # it can reuse. Note this is a directional compatibility so mutual compatibility between two OS's

lib/spack/spack/schema/concretizer.py

@@ -84,7 +84,11 @@
             "duplicates": {
                 "type": "object",
                 "properties": {
-                    "strategy": {"type": "string", "enum": ["none", "minimal", "full"]}
+                    "strategy": {"type": "string", "enum": ["none", "minimal", "full"]},
+                    "max_dupes": {
+                        "type": "object",
+                        "additional_properties": {"type": "integer", "minimum": 1},
+                    },
                 },
             },
             "static_analysis": {"type": "boolean"},

lib/spack/spack/solver/input_analysis.py

@@ -468,16 +468,29 @@ def possible_packages_facts(self, gen, fn):
         gen.newline()
 
         gen.h2("Packages with at multiple possible nodes (build-tools)")
+        default = spack.config.CONFIG.get("concretizer:duplicates:max_dupes:default", 2)
         for package_name in sorted(self.possible_dependencies() & build_tools):
-            gen.fact(fn.max_dupes(package_name, 2))
+            max_dupes = spack.config.CONFIG.get(
-            gen.fact(fn.multiple_unification_sets(package_name))
+                f"concretizer:duplicates:max_dupes:{package_name}", default
+            )
+            gen.fact(fn.max_dupes(package_name, max_dupes))
+            if max_dupes > 1:
+                gen.fact(fn.multiple_unification_sets(package_name))
         gen.newline()
 
-        gen.h2("Maximum number of nodes (virtual packages)")
+        gen.h2("Maximum number of nodes (link-run virtuals)")
-        for package_name in sorted(self.possible_virtuals()):
+        for package_name in sorted(self._link_run_virtuals):
             gen.fact(fn.max_dupes(package_name, 1))
         gen.newline()
 
+        gen.h2("Maximum number of nodes (other virtuals)")
+        for package_name in sorted(self.possible_virtuals() - self._link_run_virtuals):
+            max_dupes = spack.config.CONFIG.get(
+                f"concretizer:duplicates:max_dupes:{package_name}", default
+            )
+            gen.fact(fn.max_dupes(package_name, max_dupes))
+        gen.newline()
+
         gen.h2("Possible package in link-run subDAG")
         for name in sorted(self._link_run):
             gen.fact(fn.possible_in_link_run(name))