[llvm-commits] [PATCH] Common symbols aren't aligned in RuntimeDyld

[Jim Grosbach]

The getObjFormat() is breaking the intended abstraction. If there's file format specific logic required, that should be in the platform specific bits. RuntimeDyldELF in this case.

Note that this is a good example of why I still don't much like the use of the libObject ObjectBuffer stuff here (totally not your fault, just a general observation). Eric, if you're still keen to clean some of this stuff up, here's one for you. ;)

-Jim
On Oct 12, 2012, at 7:05 AM, Amara Emerson <amara.emerson at arm.com> wrote:

> A new patch to use the OffsetToAlignment function in MathExtras.h, recommended by Tim during review of the lli alignment patch, instead of manual alignment offset calculation.
> 
> Amara
> 
> -----Original Message-----
> From: llvm-commits-bounces at cs.uiuc.edu [mailto:llvm-commits-bounces at cs.uiuc.edu] On Behalf Of Amara Emerson
> Sent: 12 October 2012 13:45
> To: llvm-commits at cs.uiuc.edu
> Subject: [llvm-commits] [PATCH] Common symbols aren't aligned in RuntimeDyld
> 
> Hi,
> 
> I noticed that common symbols in the RuntimeDyld do not have their alignment
> requirements honoured at all. For ELF relocatable objects, the value field
> of a common symbol in the symbol table is used to encode the alignment
> requirement. However, there seems to be no way to retrieve this information
> currently and so the symbols cannot be aligned.
> 
> The patch fixes this by adding a hook into ELFObjectFile to return the value
> of a symbol, and ensures proper alignment. The solution currently applies to
> ELF only, the relevant hooks in MachO and COFF have been stubbed for now. A
> test case is included as well.
> 
> Please review the attached patch.
> 
> Amara<common-symbol-alignment-revised1.patch>_______________________________________________
> llvm-commits mailing list
> llvm-commits at cs.uiuc.edu
> http://lists.cs.uiuc.edu/mailman/listinfo/llvm-commits