Re: Crash of GDB with gdbserver btrace enabled [Re: [patch v9 00/23] branch tracing support for Atom]

[Jan Kratochvil <jan dot kratochvil at redhat dot com>]

On Thu, 07 Mar 2013 16:31:23 +0100, Metzger, Markus T wrote:
> On a related topic, remote.c calls pop_target at various places, e.g.
> in readchar when the communication breaks down.  It looks to me
> as if it assumes that it is the topmost target.
> 
> If communication breaks down during remote recording, wouldn't
> this pop the record target instead of the remote target?

It seems so, in fact when I was playing here with btrace and gdbserver I had
to sometimes 'killall gdb' after gdbserver died.  I did not put much attention
to it.

Explicit unpush_target for remote_ops or extended_remote_ops as appropriate
would be hopefully a fix.


Thanks,
Jan