[PATCH 1/2] new memory-changed MI notification.

[dje@google.com]

Yao Qi writes:
 > On 09/29/2012 01:17 AM, Tom Tromey wrote:
 > > Usually I think it would be preferable to have a flag correspond to a
 > > notification and not a command; but this would not work so well if a
 > > command needed to suppress two different messages.  (Though if that
 > > happens then maybe we should have a slightly different approach based on
 > > bitmasks.)
 > > 
 > 
 > I agree with you that one flag should correspond to a notification.  I
 > revised my patch a little bit to get rid of suppression flag
 > 'var_assign'.

Hi.

For my own education, is this suppression just an optimization, or is there a correctness issue here?
I can imagine that it's an optimization, why notify the frontend something changed when it's the frontend that requested the change.
But there is *zero* documentation in mi-main.h on *why* struct mi_suppress_notification exists, so it's hard to tell. :-(
[I realize your patch is just adding an entry, but I'd like to learn what the reason for it is.]