[llvm-commits] MCStreamer::EmitCOFFSymbolDef

[Nathan Jeffords]

On Sat, May 8, 2010 at 11:36 AM, Chris Lattner <clattner at apple.com> wrote:

>
> On May 7, 2010, at 4:58 PM, Nathan Jeffords wrote:
>
> > here is a patch that adds EmitCOFFSymbolDef to the MCStreamer interface
> this allows a symbols storage class & type to be set from the AsmPrinter
> without using EmitRawText. <EmitCOFFSymbolDef.patch>
>
> Are you sure that this is better than emitting each directive individually?
>  Should def/scl/type/endef each be their own callbacks?
>
> The MC design assumes that we want to build real native assemblers (if for
> no other reason, this makes it really easy to verify that the -filetype=obj
> and -filetype=asm paths are generating the same code).  If you start
> aggregating callbacks like this, we can't do this reasonably.
>
> Another way of asking the question: if gas can handle these directives
> individually, why does the COFF emitter need them aggregated together?
>
> -Chris


OK, I reworked the patch to break out the components of the symbol
definition into separate callbacks.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20100508/0377fb3b/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: EmitCOFFSymbolDef.patch
Type: application/octet-stream
Size: 8292 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-commits/attachments/20100508/0377fb3b/attachment.obj>