Christopher Faylor wrote:
> On Tue, Aug 14, 2007 at 01:23:28AM +0100, Pedro Alves wrote:
>> Here is the new version of the patch that converts native win32
>> debugging to use the new solib-target.c.
> 
> I have a few of questions/observations wrt the win32-nat.c changes.
> 

Thanks for such a quick review.

> 1) Does it still properly handle the "exceptions" that are thrown by
> cygwin which must be ignored?  I believe that the most populr source
> of complaints about those came from people who were using pthreads
> functions.
> 

Yes.  I tried with this:
http://www.cygwin.com/ml/cygwin/2006-05/msg00650.html

'set cygwin-exceptions 1' shows the SEGV, 'set cygwin-exceptions 0'
doesn't.

> 2) Is there some reason you didn't record the cygwin load/start address
> in win32_make_so?  Isn't all of the information you need available when
> that function is called?
> 

I could open a bfd and look for .text like old solib_symbols_add
does, but the so_list of the main list will have one open, so we can
use that instead.  I'll want to move these bits of cygwin detection and
exception ignoring into a win32-tdep.c file that can be reused
when remote debugging.  This is step in that direction, but I'll
understand if you still prefer the old way.

> 3) If the answer to the above question is no, then it seems like
> cygwin_load_start and cygwin_load_end should be static variables local
> to ignore_access_violation_p.  

Storing the cygwin1.dll addresses and reusing
it across runs should be ok if we assume that we are always
debugging cygwin apps, but it could mask an access exception in
a non-cygwin app that happened to occur in a dll loaded in that
range by coincidence.  I was going to leave it for later, but
since you've asked, I'm now clearing the start/end addresses
in do_initial_win32_stuff.  This shows why they can't be local
static.

 > I wonder if that logic should even be
 > further broken out into its own inside_cygwin(addr) function.

Wonder no more.  Done.

> 4) I'd prefer it if you dropped the _p from "ignore_access_violation".
> 

Done.


I gave it another testsuite spin, and it looks good.

-- 
Cheers,
Pedro Alves