Re: [RFA]: LOC_COMPUTED + abstracted dwarf2 evaluator, again

[Daniel Jacobowitz <drow at mvista dot com>]

On Mon, Jul 08, 2002 at 01:02:59PM -0400, Daniel Berlin wrote:
> For yer convenience,  I rediffed it against the current gdb sources.
> I'll update the changelog dates if it ever gets reviewed or approved.
> Slightly updated accompanying text (from the april submission):
> "
> The only real things left to do is give some *real* description in 
> locexpr_describe_location, add tracepoint support, and remove the 
> dwarf2cfi.c evaluator .  
> None of these should be blockers to committing it or reviewing it.
> 
> It would also be nice to use location expressions in *all* cases 
> (currently, it only does it for variables) so you can get rid of 
> decode_loc_desc, but i'll leave that for the next patch.
>  
> There are a few lines that are too long (<10, i *think*), sorry about 
> that, i couldn't see an easy way to break them up and have them make 
> sense.
> 
> Splitting this patch into loc_computed addition, and dwarf2 implementation 
> of loc_computed, doesn't make it much smaller (loc_computed_addition is 
> maybe 5% of the patch), so if size is a concern, sorry, but it's pretty 
> much unsplittable beyond that (the biggest portion is the addition of the 
> abstracted evaluator, but the dwarf2 implementation doesn't work without 
> it).
> 
> This patch has been tested on both powerpc-linux and 686-linux.
> "

FYI,

I've picked up this patch (well, an updated version Daniel sent me) and
I intend to get it working and resubmitted.  I've got it mostly passing
the testsuite now (there were a couple of problems, with frames and
such).  The tracepoint test will fail until that bit is implemented,
which I don't see as a blocker; I'll try to do that bit as a separate
patch afterwards.

I'm still working on cleaning up some bogosities, etc. etc., so it will
be another day or two.

-- 
Daniel Jacobowitz
MontaVista Software                         Debian GNU/Linux Developer