[LLD] [COFF] Don't export symbols that have corresponding __imp_ symbols

[mstorsjo]

Summary

GNU ld has got an exception for such symbols, and mingw-w64 occasionally uses that exception to avoid exporting symbols in cases where they otherwise aren't caught by the other exclusion mechanisms.

[ruiu]

COFF/MinGW.cpp

103–109

nit: add a blank line before each comment.

105–108

Don't you have to handle the i386 name mangling which is ___imp_ (triple underscores)?

[mstorsjo]

COFF/MinGW.cpp

103–109

Ok, will do.

105–108

No, the __imp_ prefix is added before the actual symbol name. The function foo which for cdecl gets mangled into _foo has the corresponding __imp__foo (while it'd be __imp_foo for x86_64), so here it works out automatically.

[ruiu]

LGTM

COFF/MinGW.cpp

 //===- MinGW.cpp ----------------------------------------------------------===//
 //
 //                             The LLVM Linker
 //
 // This file is distributed under the University of Illinois Open Source
 // License. See LICENSE.TXT for details.
 //
 //===----------------------------------------------------------------------===//
 
 #include "MinGW.h"
+#include "SymbolTable.h"
 #include "lld/Common/ErrorHandler.h"
 #include "llvm/Object/COFF.h"
 #include "llvm/Support/Path.h"
 #include "llvm/Support/raw_ostream.h"
 
 using namespace lld;
 using namespace lld::coff;
 using namespace llvm;
if (!Sym || !Sym->isLive() || !Sym->getChunk())
     return false;
 
   // Only allow the symbol kinds that make sense to export; in particular,
   // disallow import symbols.
   if (!isa<DefinedRegular>(Sym) && !isa<DefinedCommon>(Sym))
     return false;
   if (ExcludeSymbols.count(Sym->getName()))
     return false;
+  // Don't export anything that looks like an import symbol (which also can be
+  // a manually defined data symbol with such a name).
+  if (Sym->getName().startswith("__imp_"))
+    return false;
+  // If a corresponding __imp_ symbol exists and is defined, don't export it.
+  if (Symtab->find(("__imp_" + Sym->getName()).str()))
+    return false;
 
   // Check that file is non-null before dereferencing it, symbols not
   // originating in regular object files probably shouldn't be exported.
   if (!Sym->getFile())
     return false;
 
   StringRef LibName = sys::path::filename(Sym->getFile()->ParentName);
 

test/COFF/export-all.s

 # REQEUIRES: x86
 
 # RUN: llvm-mc -triple=i686-windows-gnu %s -filetype=obj -o %t.obj
 
 # RUN: lld-link -lldmingw -dll -out:%t.dll -entry:DllMainCRTStartup@12 %t.obj -implib:%t.lib
 # RUN: llvm-readobj -coff-exports %t.dll | FileCheck %s
 # RUN: llvm-readobj %t.lib | FileCheck -check-prefix=IMPLIB %s
 
 # CHECK-NOT: Name: DllMainCRTStartup
+# CHECK-NOT: Name: _imp__unexported
 # CHECK: Name: dataSym
 # CHECK: Name: foobar
+# CHECK-NOT: Name: unexported
 
 # IMPLIB: Symbol: __imp__dataSym
 # IMPLIB-NOT: Symbol: _dataSym
 # IMPLIB: Symbol: __imp__foobar
 # IMPLIB: Symbol: _foobar
 
 .global _foobar
 .global _DllMainCRTStartup@12
 .global _dataSym
+.global _unexported
+.global __imp__unexported
 .text
 _DllMainCRTStartup@12:
   ret
 _foobar:
   ret
+_unexported:
+  ret
 .data
 _dataSym:
   .int 4
+__imp__unexported:
+  .int _unexported
 
 # Test specifying -export-all-symbols, on an object file that contains
 # dllexport directive for some of the symbols.
 
 # RUN: yaml2obj < %p/Inputs/export.yaml > %t.obj
 #
 # RUN: lld-link -out:%t.dll -dll %t.obj -lldmingw -export-all-symbols -output-def:%t.def
 # RUN: llvm-readobj -coff-exports %t.dll | FileCheck -check-prefix=CHECK2 %s