I believe there was an agreement that it was not a good idea to send a =breakpoint-modified event each time a dprintf was hit. When a dprintf is in a loop, these event can become quite inefficient. Furthermore, these events are not sent when using 'agent' dprintf-style. Please see discussion and agreement here: http://sourceware.org/ml/gdb-patches/2013-03/msg00260.html http://sourceware.org/ml/gdb-patches/2013-03/msg00380.html (note that if this change is made, it would nullify Bug 15694)
I'm still worried about this =breakpoint-modified being sent every time a dprintf hits. Sounds like excessive noise towards the frontend.
Proposed fix: https://sourceware.org/ml/gdb-patches/2014-04/msg00601.html
It turns out that it was not the =breakpoint-modified event causing a performance degradation but the extra *running events reported by bug 15693. I'm marking this bug as WONTFIX based on Pedro's arguments: https://sourceware.org/ml/gdb-patches/2014-04/msg00623.html