[PATCH] D142822: [clang] ASTImporter: Fix importing of va_list types and declarations

Vince Bridgers via Phabricator via cfe-commits cfe-commits at lists.llvm.org
Thu Feb 16 16:47:36 PST 2023


vabridgers added a comment.

@donat.nagy , regarding the namespace leaking, there was a change -> https://reviews.llvm.org/D116774 that modified the behavior for aarch64 and arm. While not incorrect, it may offer some historical view or examples of how to address the current cases.

@whisperity , I'll try your advice and post the results.



================
Comment at: clang/test/AST/ast-dump-overloaded-operators.cpp:27
 // CHECK-NEXT: | `-ParmVarDecl {{.*}} <col:18> col:19{{( imported)?}} 'E'
-// CHECK-NEXT: `-FunctionDecl {{.*}} <line:14:1, line:18:1> line:14:6{{( imported)?}} test 'void ()'
+// CHECK-NEXT:  -FunctionDecl {{.*}} <line:14:1, line:18:1> line:14:6{{( imported)?}} test 'void ()'
 // CHECK-NEXT:   `-CompoundStmt {{.*}} <col:13, line:18:1>
----------------
donat.nagy wrote:
> Why did a backtick disappear here and in many other locations? Is this somehow related to va_list handling?
this backspace appeared in the original patch submitted by @mizvekov in review https://reviews.llvm.org/D136886, discussed for the test case clang/test/AST/ast-dump-overloaded-operators.cpp between @aaron.ballman and @mizvekov. 

Comments from that review from @mizvekov

<q>
What is happening here (and in all the other such instances) is that on the import case, this declaration stops being the last one on the TU. So the beginning of the line would match on |- instead of `-, but the non-import case this remains the last one.

So I simply relaxed the match.
</q>


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D142822/new/

https://reviews.llvm.org/D142822



More information about the cfe-commits mailing list