Re: [PATCH] Do not record a rejected target description

[Andrew Burgess <aburgess@redhat.com>]

Tom Tromey <tromey@adacore.com> writes:

>>>>>> "Andrew" == Andrew Burgess <aburgess@redhat.com> writes:
>
>>> info.target_desc = tdesc_info->tdesc;
>>> if (!gdbarch_update_p (info))
>>> -	warning (_("Architecture rejected target-supplied description"));
>>> +	{
>>> +	  warning (_("Architecture rejected target-supplied description"));
>>> +	  tdesc_info->tdesc = nullptr;
>>> +	}
>
> Andrew> This seems fine, but it is possible that an exception could take us out
> Andrew> of this function too, for example, a misbehaving remote target can cause
> Andrew> target_read_description_xml to throw an exception.  I guess that any of
> Andrew> the issues you were originally seeing would also trigger if we throw an
> Andrew> exception.
>
> I don't think this can really happen, because that call is part of the
> assignment to the field in question:
>
>   if (tdesc_info->tdesc == nullptr)
>     tdesc_info->tdesc = target_read_description
>       (current_inferior ()->top_target ());
>
> So, if it does throw, the tdesc_info won't be updated.

You are, of course, correct.  I guess given we know nothing after the
assignment can throw the what you have is fine.  Hopefully we'll not add
an error() call later.

Sorry for the noise.

Andrew