[llvm-profdata] Reintroduce use of InitLLVM

[tcreech-intel]

Before llvm-profdata participated in llvm-driver it directly called InitLLVM, which takes care of wildcard argument expansion for tools on Windows. When llvm-driver support was added to llvm-profdata this InitLLVM call was effectively moved into the common llvm-driver wrapper mechanism.

More recently, in #162191, llvm-driver support was temporarily backed out of llvm-profdata due to an issue with cl::opt handling. This change reintroduces the direct call to InitLLVM in order to restore wildcard expansion and also adds a test for the wildcard expansion on Windows.

[llvmbot]

@llvm/pr-subscribers-pgo

Author: Tim Creech (tcreech-intel)

Changes

Now that llvm-profdata does not participate in llvm-driver it makes sense to directly call InitLLVM again in order to expand wildcard arguments on Windows.

This change reintroduces the use of InitLLVM (as it was before llvm-profdata was moved to llvm-driver) and also adds a test for the wildcard expansion on Windows.

---

Full diff: https://github.com/llvm/llvm-project/pull/164736.diff

2 Files Affected:

• (added) llvm/test/tools/llvm-profdata/input-wildcard.test (+15)
• (modified) llvm/tools/llvm-profdata/llvm-profdata.cpp (+2-1)

diff --git a/llvm/test/tools/llvm-profdata/input-wildcard.test b/llvm/test/tools/llvm-profdata/input-wildcard.test
new file mode 100644
index 0000000000000..f2c46c962a817
--- /dev/null
+++ b/llvm/test/tools/llvm-profdata/input-wildcard.test
@@ -0,0 +1,15 @@
+# This test verifies that llvm-profdata will do wildcard expansion on its
+# arguments. The expansion is done by Windows-specific support in InitLLVM, so
+# we only expect this to work on Windows hosts.
+# REQUIRES: system-windows
+
+# Create two files to glob.
+RUN: echo '# empty profile 1' >  %t.prof1.proftxt
+RUN: echo '# empty profile 2' >> %t.prof2.proftxt
+
+# Prevent LIT itself from globbing by quoting the wildcard argument.
+RUN: llvm-profdata merge "%t.*.proftxt" -dump-input-file-list -o /dev/null | FileCheck %s
+
+# Verify that llvm-profdata expanded the wildcard argument.
+CHECK: 1,{{.*}}.prof1.proftxt
+CHECK-NEXT: 1,{{.*}}.prof2.proftxt
diff --git a/llvm/tools/llvm-profdata/llvm-profdata.cpp b/llvm/tools/llvm-profdata/llvm-profdata.cpp
index 15ddb05f953ee..a356bcd0773e0 100644
--- a/llvm/tools/llvm-profdata/llvm-profdata.cpp
+++ b/llvm/tools/llvm-profdata/llvm-profdata.cpp
@@ -34,7 +34,7 @@
 #include "llvm/Support/FileSystem.h"
 #include "llvm/Support/Format.h"
 #include "llvm/Support/FormattedStream.h"
-#include "llvm/Support/LLVMDriver.h"
+#include "llvm/Support/InitLLVM.h"
 #include "llvm/Support/MD5.h"
 #include "llvm/Support/MemoryBuffer.h"
 #include "llvm/Support/Path.h"
@@ -3465,6 +3465,7 @@ static int order_main() {
 }
 
 int main(int argc, const char *argv[]) {
+  InitLLVM X(argc, argv);
   StringRef ProgName(sys::path::filename(argv[0]));
 
   if (argc < 2) {

[petrhosek]

FWIW we want to support llvm-driver in llvm-profdata, but we first need to replace the use of cl::opt with OptTable.

[Prabhuk]

@tcreech-intel -- As @petrhosek mentioned, I am working on porting llvm-profdata to OptTable so that it can be added back to the llvm driver though its a bit low on my priorities right now. Will it be alright with you if I remove this InitLLVM call when I make llvm-profdata part of the llvm-driver again?

cc: @mingmingl-llvm

[tcreech-intel]

Hi @Prabhuk: yes, that's completely fine by me.

When you move back to llvm-driver an InitLLVM in llvm-profdata.cpp would be redundant with this one in llvm-driver-template.cpp:

llvm-project/llvm/cmake/modules/llvm-driver-template.cpp.in

Lines 15 to 18 in 83f751a

	int main(int argc, char **argv) {
	llvm::InitLLVM X(argc, argv);
	return @TOOL_NAME@_main(argc, argv, {argv[0], nullptr, false});
	}

For my purposes I'd just like to ensure that a call to InitLLVM is preserved somewhere to do the expansion. The new test should help.

[Prabhuk]

Thank you @tcreech-intel

[tcreech-intel]

FWIW we want to support llvm-driver in llvm-profdata, but we first need to replace the use of cl::opt with OptTable.

Thanks, @petrhosek. I'll clarify the commit message a little to reflect that the current situation is likely temporary.