Re: [PATCH] gdbsupport: fix scoped_debug_start_end's move constructor

[Simon Marchi <simon.marchi@polymtl.ca>]

On 1/5/23 15:05, Tom Tromey wrote:
>>>>>> "Simon" == Simon Marchi via Gdb-patches <gdb-patches@sourceware.org> writes:
> 
>>> +  scoped_debug_start_end (scoped_debug_start_end &&other)
>>> +    : m_debug_enabled (other.m_debug_enabled),
>>> +      m_module (other.m_module),
>>> +      m_func (other.m_func),
>>> +      m_end_prefix (other.m_end_prefix),
>>> +      m_msg (other.m_msg),
> 
> Simon> Just found this nit... not that it changes anything (because this ctor
> Simon> isn't called in practice), but we should std::move m_msg.  I'll change
> Simon> it locally.
> 
> I think it's also fine to just leave it as-is.
> 
> Simon> Well, we could std::move all fields, but it would be a bit verbose.
> 
> If we think we may need this kind of behavior again, one way would be a
> sort of "move token" object that wraps a bool, and that sets the bool on
> construction and clears it on move.  Then scoped_debug_start_end could
> just use the default move constructor again, and check the token's value
> in its destructor.

Interesting, I'll keep this in mind.  I think it could apply to some
scoped_restore_* objects that have global side effects, that we want to
be movable, because we want to be able to have functions that return
them.

> Probably overkill for just the one case.  I think your patch is ok.

Ok, thanks, will push.

Simon