This is the mail archive of the
gdb-patches@sourceware.org
mailing list for the GDB project.
Re: qXfer:exec-file:read and non multiprocess target
- From: Philippe Waroquiers <philippe dot waroquiers at skynet dot be>
- To: Gary Benson <gbenson at redhat dot com>
- Cc: gdb-patches at sourceware dot org
- Date: Tue, 05 May 2015 22:45:44 +0200
- Subject: Re: qXfer:exec-file:read and non multiprocess target
- Authentication-results: sourceware.org; auth=none
- References: <1430560118 dot 11263 dot 9 dot camel at soleil> <20150505110207 dot GA17684 at blade dot nx>
On Tue, 2015-05-05 at 12:02 +0100, Gary Benson wrote:
> The PID is fake because vgdb does not support multiprocess extensions.
> I don't like sending a fake/zero PID over the wire, but how about I
> change qXfer:exec-file:read to send a NULL annex if the remote does
> not have multiprocess extensions? Can you make your side work with
> the patch inlined below? If so I'll tidy and document it and submit
> it for review.
It looks effectively better to not send the fake or 0 pid
(e.g. similar to the qAttached packet).
The valgrind gdbserver side works ok with the patch to send an empty
annex.
This is the resulting exchange, traced at Valgrind side:
...
--4980:3: gdbsrv getpkt ("qAttached"); [no ack]
--4980:3: gdbsrv putpkt ("$1#31"); [no ack]
--4980:3: gdbsrv getpkt ("qXfer:exec-file:read::0,fff"); [no ack]
--4980:3: gdbsrv putpkt ("$l/home/philippe/valgrind/better_stats/memcheck/tests/trivialleak#2c"); [no ack]
--4980:3: gdbsrv getpkt ("vFile:open:2f686f6d652f7068696c697070652f76616c6772696e642f6265747465725f73746174732f6d656d636865636b2f74657374732f7472697669616c6c65616b,0,0"); [no ack]
--4980:3: gdbsrv putpkt ("$#00"); [no ack]
...
and then GDB could properly use the returned exec-file
(even if vFile is not supported by Valgrind gdbserver).
Thanks
Philippe