[flang] Don't duplicate hermetic module file dependencies

[klausler]

When emitting the modules on which a module depends under the -fhermetic-module-files options, eliminate duplicates by name rather than by symbol addresses. This way, when a dependent module is in the symbol table more than once due to the use of a nested hermetic module, it doesn't get emitted multiple times to the new module file.

[llvmbot]

@llvm/pr-subscribers-flang-semantics

Author: Peter Klausler (klausler)

Changes

When emitting the modules on which a module depends under the -fhermetic-module-files options, eliminate duplicates by name rather than by symbol addresses. This way, when a dependent module is in the symbol table more than once due to the use of a nested hermetic module, it doesn't get emitted multiple times to the new module file.

---

Full diff: https://github.com/llvm/llvm-project/pull/143605.diff

3 Files Affected:

• (modified) flang/lib/Semantics/mod-file.cpp (+11-7)
• (added) flang/test/Semantics/modfile76.F90 (+33)
• (added) flang/test/Semantics/modfile77.F90 (+37)

diff --git a/flang/lib/Semantics/mod-file.cpp b/flang/lib/Semantics/mod-file.cpp
index a72641866aa15..9f9e9f5840456 100644
--- a/flang/lib/Semantics/mod-file.cpp
+++ b/flang/lib/Semantics/mod-file.cpp
@@ -143,18 +143,22 @@ void ModFileWriter::Write(const Symbol &symbol) {
   std::string path{context_.moduleDirectory() + '/' +
       ModFileName(symbol.name(), ancestorName, context_.moduleFileSuffix())};
 
-  UnorderedSymbolSet hermeticModules;
-  hermeticModules.insert(symbol);
+  std::set<std::string> hermeticModuleNames;
+  hermeticModuleNames.insert(symbol.name().ToString());
   UnorderedSymbolSet additionalModules;
   PutSymbols(DEREF(symbol.scope()),
       hermeticModuleFileOutput_ ? &additionalModules : nullptr);
   auto asStr{GetAsString(symbol)};
   while (!additionalModules.empty()) {
-    for (auto ref : UnorderedSymbolSet{std::move(additionalModules)}) {
-      if (hermeticModules.insert(*ref).second &&
-          !ref->owner().IsIntrinsicModules()) {
-        PutSymbols(DEREF(ref->scope()), &additionalModules);
-        asStr += GetAsString(*ref);
+    UnorderedSymbolSet nextPass{std::move(additionalModules)};
+    additionalModules.clear();
+    for (const Symbol &modSym : nextPass) {
+      if (!modSym.owner().IsIntrinsicModules() &&
+          hermeticModuleNames.find(modSym.name().ToString()) ==
+              hermeticModuleNames.end()) {
+        hermeticModuleNames.insert(modSym.name().ToString());
+        PutSymbols(DEREF(modSym.scope()), &additionalModules);
+        asStr += GetAsString(modSym);
       }
     }
   }
diff --git a/flang/test/Semantics/modfile76.F90 b/flang/test/Semantics/modfile76.F90
new file mode 100644
index 0000000000000..5b783e217e14d
--- /dev/null
+++ b/flang/test/Semantics/modfile76.F90
@@ -0,0 +1,33 @@
+!RUN: %flang -c -fhermetic-module-files -DWHICH=1 %s && %flang -c -fhermetic-module-files -DWHICH=2 %s && %flang -c -fhermetic-module-files %s && cat modfile76c.mod | FileCheck %s
+
+#if WHICH == 1
+module modfile76a
+  integer :: global_variable = 0
+end
+#elif WHICH == 2
+module modfile76b
+  use modfile76a
+ contains
+  subroutine test
+  end
+end
+#else
+module modfile76c
+  use modfile76a
+  use modfile76b
+end
+#endif
+
+!CHECK: module modfile76c
+!CHECK: use modfile76a,only:global_variable
+!CHECK: use modfile76b,only:test
+!CHECK: end
+!CHECK: module modfile76a
+!CHECK: integer(4)::global_variable
+!CHECK: end
+!CHECK: module modfile76b
+!CHECK: use modfile76a,only:global_variable
+!CHECK: contains
+!CHECK: subroutine test()
+!CHECK: end
+!CHECK: end
diff --git a/flang/test/Semantics/modfile77.F90 b/flang/test/Semantics/modfile77.F90
new file mode 100644
index 0000000000000..a82904ebbcc22
--- /dev/null
+++ b/flang/test/Semantics/modfile77.F90
@@ -0,0 +1,37 @@
+!RUN: %flang -c -fhermetic-module-files -DWHICH=1 %s && %flang -c -fhermetic-module-files -DWHICH=2 %s && %flang -c -fhermetic-module-files %s && cat modfile77c.mod | FileCheck %s
+
+#if WHICH == 1
+module modfile77a
+  interface gen
+    procedure proc
+  end interface
+ contains
+  subroutine proc
+    print *, 'ok'
+  end
+end
+#elif WHICH == 2
+module modfile77b
+  use modfile77a
+end
+#else
+module modfile77c
+  use modfile77a
+  use modfile77b
+end
+#endif
+
+!CHECK: module modfile77c
+!CHECK: use modfile77a,only:proc
+!CHECK: use modfile77a,only:gen
+!CHECK: interface gen
+!CHECK: end interface
+!CHECK: end
+!CHECK: module modfile77a
+!CHECK: interface gen
+!CHECK: procedure::proc
+!CHECK: end interface
+!CHECK: contains
+!CHECK: subroutine proc()
+!CHECK: end
+!CHECK: end

[akuhlens]

This looks good to me.