Re: Speaking of DWARF-2 macro info

[Daniel Berlin <dan at dberlin dot org>]

On Sun, 17 Mar 2002, Carlo Wood wrote:

> On Sun, Mar 17, 2002 at 05:51:06PM +0000, Neil Booth wrote:
> > Neil Booth wrote:-
> > 
> > > The first /dev/null is so that re-compiling preprocessed output can get
> > > the original source file from the first line [the .i -> .c translation].
> > 
> > You may remember my asking you (and maybe others) about this around 8
> > months ago.  The only reason we need to do this is for the debug code in
> > GCC - it insists on having to know the name of the main input file before
> > anything else (note how, when I cleaned up toplev.c initialization,
> > debug info is about the last subsystem to be initialized, because it
> > needs to wait for this information right until CPP starts preprocessing].
> > If you can avoid this, then we can change CPP to not have to output the
> > initial line marker.
> 
> Neil, just a question.  I have this in the code of my c++ debugging support lib:
> 
> // This assumes that DW_TAG_compile_unit is the first tag for each compile unit
> // which is not strictly garanteed in the standard but seems to be the case.
> // Moreover it assumes that DW_TAG_compile_unit is the first entry in the
> // .debug_info section for each compile unit.
> #define DWARFSPEEDUP
> 
> would it be possible that if you change the above that then I can't
> rely anymore on DW_TAG_compile_unit being the first entry in the .debug_info
> section for each compile unit?

I'm pretty sure it's guaranteed by the standard to be the first entry.
> 
> Sorry if this is not related, I am not sure.
> 
>