This is the mail archive of the
mailing list for the GCC project.
Re: [debug/13267] proposed patch to dwarf2out.c
On Fri, 2004-03-05 at 13:26, DJ Delorie wrote:
> Ok, I did some debugging. At the time the code in
> rtl_for_decl_location is called,the rtl for that parameter is a REG.
> The code checks for MEMs (and the bug only happens when it's a MEM
> anyway), so never catches that case.
So where did the MEM come from then? How come the value is a REG in
rtl_for_decl_location and then magically a MEM in
add_location_or_const_value_attribute? That doesn't make sense. Please
Was it the eliminate_regs call? If so, why did it convert a REG to a
MEM? Maybe the problem is that the big-endian correction in
rtl_for_decl_location is done in the wrong place? Maybe it needs to be
after the eliminate_regs call?
Or maybe one of the other functions converted it from a REG to a MEM?
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com