[LLD][COFF] Use primary symbol table machine in Writer::writeHeader (NFC)

[cjacek]

Instead of duplicating the logic from LinkerDriver::setMachine.

[llvmbot]

@llvm/pr-subscribers-lld
@llvm/pr-subscribers-platform-windows

@llvm/pr-subscribers-lld-coff

Author: Jacek Caban (cjacek)

Changes

Instead of duplicating the logic from LinkerDriver::setMachine.

---

Full diff: https://github.com/llvm/llvm-project/pull/128442.diff

1 Files Affected:

• (modified) lld/COFF/Writer.cpp (+1-10)

diff --git a/lld/COFF/Writer.cpp b/lld/COFF/Writer.cpp
index 58727c1615769..3efbcdb452ebe 100644
--- a/lld/COFF/Writer.cpp
+++ b/lld/COFF/Writer.cpp
@@ -1758,16 +1758,7 @@ template <typename PEHeaderTy> void Writer::writeHeader() {
   assert(coffHeaderOffset == buf - buffer->getBufferStart());
   auto *coff = reinterpret_cast<coff_file_header *>(buf);
   buf += sizeof(*coff);
-  switch (config->machine) {
-  case ARM64EC:
-    coff->Machine = AMD64;
-    break;
-  case ARM64X:
-    coff->Machine = ARM64;
-    break;
-  default:
-    coff->Machine = config->machine;
-  }
+  coff->Machine = ctx.symtab.isEC() ? AMD64 : ctx.symtab.machine;
   coff->NumberOfSections = ctx.outputSections.size();
   coff->Characteristics = IMAGE_FILE_EXECUTABLE_IMAGE;
   if (config->largeAddressAware)

[mstorsjo]

LGTM. The logic for which symtab has which machine set in which mode is still hazy to me, but the change looks straightforward, and I presume we have relevant test coverage to be sure that this really is NFC.