[llvm] [Windows] Adjust exported symbols in static builds with plugin support (PR #165946)

[Victor Mustya via llvm-commits]

https://github.com/vmustya created https://github.com/llvm/llvm-project/pull/165946

When buildint LLVM and Clang on Windows with plugin support enabled,
some symbols are redundantly exported due to template instantiations and
lambda functions. These symbols are not needed in the importing
translation units and can be safely removed.

In the meantime, the global variables and static data members are needed
for correct linking and runtime behavior, so they are added to the
export list.

Also, the `llvm::<Class>::dump()` and `clang::<Class>::dump()` methods
are not needed for linking in importing translation units, because they
are only available in debug builds and should be only used for debugging
purposes. Therefore, these methods are removed from the export list.

>From a25b5bf51429b81c09fef230105bbe17fd9688a4 Mon Sep 17 00:00:00 2001
From: Victor Mustya <victor.mustya at intel.com>
Date: Fri, 31 Oct 2025 17:50:12 -0700
Subject: [PATCH] [Windows] Adjust exported symbols in static builds with
 plugin support

When buildint LLVM and Clang on Windows with plugin support enabled,
some symbols are redundantly exported due to template instantiations and
lambda functions. These symbols are not needed in the importing
translation units and can be safely removed.

In the meantime, the global variables and static data members are needed
for correct linking and runtime behavior, so they are added to the
export list.

Also, the `llvm::<Class>::dump()` and `clang::<Class>::dump()` methods
are not needed for linking in importing translation units, because they
are only available in debug builds and should be only used for debugging
purposes. Therefore, these methods are removed from the export list.
---
 llvm/utils/extract_symbols.py | 29 ++++++++++++++++++++++++++++-
 1 file changed, 28 insertions(+), 1 deletion(-)

diff --git a/llvm/utils/extract_symbols.py b/llvm/utils/extract_symbols.py
index 0cbfd2e2910e1..5254d16e410d5 100755
--- a/llvm/utils/extract_symbols.py
+++ b/llvm/utils/extract_symbols.py
@@ -97,6 +97,16 @@ def should_keep_microsoft_symbol(symbol, calling_convention_decoration):
     # don't
     elif symbol.startswith("??_G") or symbol.startswith("??_E"):
         return None
+    # Delete template instantiations. These start with ?$ and can be discarded
+    # because they will be instantiated in the importing translation unit if
+    # needed.
+    elif symbol.startswith("??$"):
+        return None
+    # Delete lambda object constructors and operator() functions. These start
+    # with ??R<lambda_ or ??0<lambda_ and can be discarded because lambdas are
+    # usually local to a function.
+    elif symbol.startswith("??R<lambda_") or symbol.startswith("??0<lambda_"):
+        return None
     # An anonymous namespace is mangled as ?A(maybe hex number)@. Any symbol
     # that mentions an anonymous namespace can be discarded, as the anonymous
     # namespace doesn't exist outside of that translation unit.
@@ -131,7 +141,24 @@ def should_keep_microsoft_symbol(symbol, calling_convention_decoration):
     #                 ::= .+@ (list of types)
     #                 ::= .*Z (list of types, varargs)
     # <throw-spec> ::= exceptions are not allowed
-    elif re.search(r"(llvm|clang)@@[A-Z][A-Z0-9_]*[A-JQ].+(X|.+@|.*Z)$", symbol):
+    elif re.search(r"@(llvm|clang)@@[A-Z][A-Z0-9_]*[A-JQ].+(X|.+@|.*Z)$", symbol):
+        # Remove llvm::<Class>::dump and clang::<Class>::dump methods because
+        # they are used for debugging only.
+        if symbol.startswith("?dump@"):
+            return None
+        return symbol
+    # Keep mangled global variables and static class members in llvm:: namespace.
+    # These have a type mangling that looks like (this is derived from
+    # clang/lib/AST/MicrosoftMangle.cpp):
+    # <type-encoding> ::= <storage-class> <variable-type>
+    # <storage-class> ::= 0  # private static member
+    #                 ::= 1  # protected static member
+    #                 ::= 2  # public static member
+    #                 ::= 3  # global
+    #                 ::= 4  # static local
+    # <variable-type> ::= <type> <cvr-qualifiers>
+    #                 ::= <type> <pointee-cvr-qualifiers> # pointers, references
+    elif re.search(r"@llvm@@[0-3].*$", symbol):
         return symbol
     return None