[flang][preprocessor] Expand some keyword macros in quoted character …

[klausler]

…literals

To help port codes from compilers using pre-ANSI C preprocessors, which didn't care much about context when replacing macros, support the replacement of keyword macros in quoted character literals when (and only when) the name of the keyword macro constitutes the entire significant portion of a free form continuation line. See the new test case for a motivating example.

Fixes #96781.

[llvmbot]

@llvm/pr-subscribers-flang-parser

Author: Peter Klausler (klausler)

Changes

…literals

To help port codes from compilers using pre-ANSI C preprocessors, which didn't care much about context when replacing macros, support the replacement of keyword macros in quoted character literals when (and only when) the name of the keyword macro constitutes the entire significant portion of a free form continuation line. See the new test case for a motivating example.

Fixes #96781.

---

Full diff: https://github.com/llvm/llvm-project/pull/96987.diff

2 Files Affected:

• (modified) flang/lib/Parser/prescan.cpp (+35)
• (added) flang/test/Preprocessing/kw-in-char.F90 (+14)

diff --git a/flang/lib/Parser/prescan.cpp b/flang/lib/Parser/prescan.cpp
index 42aa829e0ed5b..668429a882291 100644
--- a/flang/lib/Parser/prescan.cpp
+++ b/flang/lib/Parser/prescan.cpp
@@ -820,6 +820,41 @@ void Prescanner::QuotedCharacterLiteral(
       }
       break;
     }
+    // Here's a weird edge case.  When there's a two or more following
+    // continuation lines at this point, and the entire significant part of
+    // the next continuation line is the name of a keyword macro, replace
+    // it in the character literal with its definition.  Example:
+    //   #define FOO foo
+    //   subroutine subr() bind(c, name="my_&
+    //     &FOO&
+    //     &_bar") ...
+    // produces a binding name of "my_foo_bar".
+    while (at_[1] == '&' && nextLine_ < limit_ && !InFixedFormSource()) {
+      const char *idStart{nextLine_};
+      if (const char *amper{SkipWhiteSpace(nextLine_)}; *amper == '&') {
+        idStart = amper + 1;
+      }
+      if (IsLegalIdentifierStart(*idStart)) {
+        std::size_t idLen{1};
+        for (; IsLegalInIdentifier(idStart[idLen]); ++idLen) {
+        }
+        if (idStart[idLen] == '&') {
+          CharBlock id{idStart, idLen};
+          if (preprocessor_.IsNameDefined(id)) {
+            TokenSequence ppTokens;
+            ppTokens.Put(id, GetProvenance(idStart));
+            if (auto replaced{
+                    preprocessor_.MacroReplacement(ppTokens, *this)}) {
+              tokens.Put(*replaced);
+              at_ = &idStart[idLen - 1];
+              NextLine();
+              continue; // try again on the next line
+            }
+          }
+        }
+      }
+      break;
+    }
     end = at_ + 1;
     NextChar();
     if (*at_ == quote && !isEscaped) {
diff --git a/flang/test/Preprocessing/kw-in-char.F90 b/flang/test/Preprocessing/kw-in-char.F90
new file mode 100644
index 0000000000000..e8f3a33b8a749
--- /dev/null
+++ b/flang/test/Preprocessing/kw-in-char.F90
@@ -0,0 +1,14 @@
+! RUN: %flang -E %s 2>&1 | FileCheck %s
+! CHECK: subroutine test_b_wrapper_c() bind(C, name="test_b_c_f")
+#define TEMP_LETTER b
+#define VAL c
+subroutine test_&
+TEMP_LETTER&
+_wrapper_&
+VAL&
+ () bind(C, name="test_&
+    &TEMP_LETTER&
+    &_&
+    &VAL&
+    &_f")
+end

[psteinfeld]

All builds and tests correctly and looks good.