This is the mail archive of the
gdb-patches@sources.redhat.com
mailing list for the GDB project.
Re: [RFA] Link remote target with svr4 solibs.
- To: jtc at redback dot com
- Subject: Re: [RFA] Link remote target with svr4 solibs.
- From: Michael Snyder <msnyder at cygnus dot com>
- Date: Mon, 12 Mar 2001 16:44:45 -0800
- CC: gdb-patches at sources dot redhat dot com, cagney at redhat dot com, kevinb at redhat dot com, hunt at redhat dot com, jsmith at redhat dot com
- Organization: Red Hat
- References: <3AAD635C.DAE5D340@cygnus.com> <5m1ys2cxxt.fsf@jtc.redback.com>
"J.T. Conklin" wrote:
>
> >>>>> "Michael" == Michael Snyder <msnyder@cygnus.com> writes:
> Michael> The following change is necessary in order to activate svr4
> Michael> solib debugging for remote targets. With it, GDB will detect
> Michael> when the remote target loads a new shared library (using the
> Michael> method implemented in solib-svr4.c), and GDB will load the
> Michael> symbols for the library (provided that GDB has been told
> Michael> where to find the symbol files, using the "set
> Michael> solib-search-path" and "set solib-absolute-prefix"
>
> If I understand the shared library implementation correctly, a magic
> breakpoint is inserted in the dynamic loader, which when hit GDB does
> normal reads and writes to determine what shared library has been
> loaded or unloaded, etc. If so, practically any remote back end
> (perhaps a ROM monitor interface) will provide the infrastructure
> required for this to work.
>
> Doesn't this argue that this should reside somewhere above
> remote_open()?
I'm open to suggestions. The other place from which it is called
is fork_inferior. Remote targets do not fork. Open is the best
place I could think of.