This is the mail archive of the gdb-patches@sourceware.org mailing list for the GDB project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [PATCH] gdb.ada/catch_ex.exp, gdb.ada/mi_catch_ex.exp and unsupported catchpoints


On Wednesday 07 December 2011 22:01:19, Joel Brobecker wrote:
> (to answer Pedro's questions, the files in question are libgnat.a and
> libgnat.so)
> 
> > Fedora strips everything to a separate debuginfo file in an automated
> > way.  There might be an exception for one library, but last time I asked
> > about this, I was told that exceptions are a pain and could we please do
> > something else.  Also, the granularity of debuginfo packages is quite
> > large.
> 
> I can see the reason for this, but it is kind of a pain for Ada.
> I believe the makefiles already build the runtime without debugging
> info except for the few files that we need (I think it's basically
> a-tags.adb and a-except.adb)...
> 
> > If you can make it so this can work without debuginfo somehow, that
> > would be much better for users.
> 
> The problem is that we need to get to the argument of function we are
> breaking on in order to determine which exception has been raised.
> That's where we really need the debug info. I think I even added
> a comment about that in the code...

GDB knows the target's function call ABI (for infcalls), and supposedly
this function's prototype is cast in stone as part of the ABI too.
GDB could just know where to get the arguments from?  You'd still need
minimal symbols in libgnat, but that doesn't look much different
from needing minimal symbols for libpthread, from the packager's
perspective -- it should be simpler than caring about stripping the library
except a couple objects, due to size considerations.

-- 
Pedro Alves


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]