Re: [Xen-devel] [PATCH 28/28] Separate runtime debug output from debug symbols

[Tim Deegan <tim@xxxxxxx>]

At 09:06 +0100 on 02 Oct (1412237217), Jan Beulich wrote:
> >>> On 02.10.14 at 10:00, <olaf@xxxxxxxxx> wrote:
> > On Thu, Oct 02, Jan Beulich wrote:
> > 
> >> >>> On 01.10.14 at 18:41, <olaf@xxxxxxxxx> wrote:
> >> > --- a/xen/Rules.mk
> >> > +++ b/xen/Rules.mk
> >> > @@ -43,7 +43,8 @@ ALL_OBJS-$(x86)          += 
> >> > $(BASEDIR)/crypto/built_in.o
> >> >  
> >> >  CFLAGS += -fno-builtin -fno-common
> >> >  CFLAGS += -Werror -Wredundant-decls -Wno-pointer-arith
> >> > -CFLAGS += -pipe -g -D__XEN__ -include $(BASEDIR)/include/xen/config.h
> >> > +CFLAGS-$(debug_symbols) += -g
> >> > +CFLAGS += -pipe -D__XEN__ -include $(BASEDIR)/include/xen/config.h
> >> >  CFLAGS += -nostdinc
> >> >  
> >> >  CFLAGS-$(XSM_ENABLE)    += -DXSM_ENABLE
> >> 
> >> I'm fine with this change, but will make my ack depend on Andrew at
> >> least not objecting (iirc he had reservations on a previous iteration).
> > 
> > Does that still stand? Quote:
> 
> Is that a question to me or to him?
> 
> > ...
> >> I am not sure this is correct.  Xen needs to embed its full symbol
> >> table for backtraces, so does need to be built unconditionally with -g
> > 
> > What does -g have to do with backtrace generation? If you're
> > after .debug_frame, that can be had as .eh_frame without -g (and
> > with -fasynchronous-unwind-tables passed instead). Yet whether
> > everyone wants this is not at all clear, so providing a knob to turn  
> > it on/off seems quite desirable to me.

I think Andrew is talking about the symbol table rather than the debug
info - xen's backtrace walker doesn't speak DWARF.  AFAICT the symbol
tables generated without '-g' are just fine for xen's own use printing
backtraces &c, though possibly post-mortem tools for crash analysis
might need more?

Tim.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel