Re: [PATCH v2 1/7] Add target method for converting thread handle to thread_info struct pointer

[Kevin Buettner <kevinb at redhat dot com>]

On Thu, 04 May 2017 23:26:20 -0400
Simon Marchi <simon.marchi@polymtl.ca> wrote:

> On 2017-04-09 02:06, Kevin Buettner wrote:
> > --- a/gdb/linux-thread-db.c
> > +++ b/gdb/linux-thread-db.c
> > @@ -1410,6 +1410,29 @@ thread_db_extra_thread_info (struct target_ops 
> > *self,
> >    return NULL;
> >  }
> > 
> > +/* Return pointer to the thread_info struct which corresponds to
> > +   THREAD_HANDLE (having length HANDLE_LEN).  */
> > +static struct thread_info *
> > +thread_db_thread_handle_to_thread_info (struct target_ops *ops,
> > +					const gdb_byte *thread_handle,
> > +					int handle_len)
> > +{
> > +  struct thread_info *tp;
> > +  thread_t handle_tid;
> > +
> > +  gdb_assert (handle_len == sizeof (handle_tid));  
> 
> I assume this is always true, because if you are using libthread_db, it 
> implies that GDB is of the exact same architecture (32 bits vs 64 bits) 
> as the inferior?

I'm about to submit a new patch set, but I decided that I ought to
address this question...

I had first thought that the answer to your question was "yes", but
after thinking about it some more, I decided that it's not appropriate
to generate an internal error (via a failed assert) for this case. 
It's possible to get here via a call from the python interface.  We
don't want user written code to generate internal errors.  Therefore,
I changed it to call error() when there's a handle size mismatch.

Thanks for this review and all of the others too.  They were very much
appreciated.

Kevin