[PATCH] D20586: IRLinker: fix double scheduling of mapping a global value because of an alias
Rafael Espíndola via llvm-commits
llvm-commits at lists.llvm.org
Tue May 24 14:08:55 PDT 2016
On 24 May 2016 at 14:56, Mehdi AMINI <mehdi.amini at apple.com> wrote:
> joker.eph created this revision.
> joker.eph added reviewers: rafael, dexonsmith.
> joker.eph added a subscriber: llvm-commits.
> This test was hitting an assertion in the value mapper because
> the IRLinker was trying to map two times @A while materializing
> the initializer for @C.
> Fix http://llvm.org/PR27850
> Index: test/Linker/alias-3.ll
> --- /dev/null
> +++ test/Linker/alias-3.ll
> @@ -0,0 +1,13 @@
> +; RUN: llvm-link %s -S -o - | FileCheck %s
> +target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
> +target triple = "x86_64-apple-macosx10.9"
> +; CHECK-DAG: @A = internal constant i8 1
> +; CHECK-DAG: @B = alias i8, i8* @A
> +; CHECK-DAG: @C = global [2 x i8*] [i8* @A, i8* @B]
> + at A = internal constant i8 1
> + at B = alias i8, i8* @A
> + at C = global [2 x i8*] [i8* @A, i8* @B]
> Index: lib/Linker/IRMover.cpp
> --- lib/Linker/IRMover.cpp
> +++ lib/Linker/IRMover.cpp
> @@ -557,6 +557,17 @@
> + // When linking a global for an alias, it will always be linked. However we
> + // need to check if it was not already scheduled to satify a reference from a
> + // regular global value initializer. We know if it has been schedule if the
> + // "New" GlobalValue that is mapped here for the alias is the same as the one
> + // already mapped. If there is an entry in the ValueMap but the value is
> + // different, it means that the value already had a definition in the
> + // destination module (linkonce for instance), but we need a new definition
> + // for the alias ("New" will be different.
> + if (ForAlias && ValueMap.lookup(Old) == New)
> + return;
> if (ForAlias || shouldLink(New, *Old))
> linkGlobalValueBody(*New, *Old);
More information about the llvm-commits