Re: [RFA] dwarf2_physname

[Keith Seitz <keiths at redhat dot com>]

On 09/15/2009 09:11 AM, Daniel Jacobowitz wrote:
> So, if icc is failing to mark things artificial, does it make up for
> it by having DW_AT_object_pointer?  I don't believe GCC generates that
> attribtue (guess it should!).

Unfortunately, ICC does not emit DW_AT_object_pointer. It outputs only 
what I've previously posted.

> This is an area where, pain or not, I'd really prefer to have only one
> way to do things.

I can certainly understand that (and I agree): I was just throwing that 
out as an option. Afterall, there is a bigger picture (for you) to 
consider -- I'm just a linux-centered, who-cares-about-anything-else 
hacker. O:-)

> I looked over your patch again.  I have no objection and Tom's already
> approved it upthread; I think you're good to go.

Tom has suggested to me to conform and start posting bigger bits of the 
patches. Unless I hear anything to the contrary, I'll consider this a 
general approval of the approach, and I will attempt to submit any 
outstanding isolated patches before posting the whole (ginormous) patch 
which implements everything.

I've also added a page to the wiki about running the test suite on 
non-gcc compilers (specifically ICC and RealView). It's not complete in 
the sense that I think some careful Dejagnu configury and/or compiler 
wrapper trickery might produce better (or more meaningful) results, but 
it's a start.

For example, I completely ignore the whole multilib issue. In my case, I 
was really just interested in the C++ DWARF implementations of these 
compilers, and I believe the setup is sufficient for this purpose. In a 
real world scenario, it probably isn't. I hope that people who 
remember/know more about this stuff add their two bits to the page.

Keith