llvm dev - Jan 2013 - [LLVMdev] ELFObjectFile::getSymbolFileOffset

While working on some other changes I came across a problem where
ELFObjectFile::getSymbolFileOffset was returning a different value than I
expected in the case where the symbol in question was a section.  Looking at the
code, it seems obviously wrong, but  I didn't want to just commit my change
without at least asking if anyone knew of a good reason for the current
behavior.  I know there have been some discussions in the past about the
address/offset ambiguity in this interface but I couldn't find anything
directly addressing this question.

Looking at the revision history, it seems that the current implementation has
been in place since getSymbolOffset was created (in
lib/Object/ELFObjectFile.cpp, r139683) based on an earlier getSymbolAddress
implementation and wasn't updated when other aspects of the function were
corrected and the function was renamed getSymbolFileOffset(in
lib/Object/ELFObjectFile.cpp, r145408) .

Here's the change I'm proposing (also attached as a patch file):

Index: include/llvm/Object/ELF.h
==================================================================---
include/llvm/Object/ELF.h     (revision 172846)
+++ include/llvm/Object/ELF.h  (working copy)
@@ -1000,7 +1000,7 @@
   switch (symb->getType()) {
   case ELF::STT_SECTION:
-    Result = Section ? Section->sh_addr : UnknownAddressOrSize;
+    Result = Section ? Section->sh_offset : UnknownAddressOrSize;
     return object_error::success;
   case ELF::STT_FUNC:
   case ELF::STT_OBJECT:

-Andy
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20130118/506f6f0a/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: section-offset.patch
Type: application/octet-stream
Size: 494 bytes
Desc: section-offset.patch
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20130118/506f6f0a/attachment.obj>

On Fri, Jan 18, 2013 at 3:13 PM, Kaylor, Andrew <andrew.kaylor at
intel.com> wrote:
> While working on some other changes I came across a problem where
> ELFObjectFile::getSymbolFileOffset was returning a different value than I
> expected in the case where the symbol in question was a section.  Looking
at
> the code, it seems obviously wrong, but  I didn’t want to just commit my
> change without at least asking if anyone knew of a good reason for the
> current behavior.  I know there have been some discussions in the past
about
> the address/offset ambiguity in this interface but I couldn’t find anything
> directly addressing this question.
>
>
>
> Looking at the revision history, it seems that the current implementation
> has been in place since getSymbolOffset was created (in
> lib/Object/ELFObjectFile.cpp, r139683) based on an earlier getSymbolAddress
> implementation and wasn’t updated when other aspects of the function were
> corrected and the function was renamed getSymbolFileOffset(in
> lib/Object/ELFObjectFile.cpp, r145408) .
>
>
>
> Here’s the change I’m proposing (also attached as a patch file):
>
>
>
> Index: include/llvm/Object/ELF.h
>
> ==================================================================>
> --- include/llvm/Object/ELF.h     (revision 172846)
>
> +++ include/llvm/Object/ELF.h  (working copy)
>
> @@ -1000,7 +1000,7 @@
>
>    switch (symb->getType()) {
>
>    case ELF::STT_SECTION:
>
> -    Result = Section ? Section->sh_addr : UnknownAddressOrSize;
>
> +    Result = Section ? Section->sh_offset : UnknownAddressOrSize;
>
>      return object_error::success;
>
>    case ELF::STT_FUNC:
>
>    case ELF::STT_OBJECT:
>
>
>
> -Andy
>
>
> _______________________________________________
> LLVM Developers mailing list
> LLVMdev at cs.uiuc.edu         http://llvm.cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>
LGTM.

- Michael Spencer