于 2013-5-2 20:02, Luiz Capitulino 写道:
On Thu, 02 May 2013 10:05:08 +0800
Wenchao Xia <xiaw...@linux.vnet.ibm.com> wrote:
于 2013-4-30 3:05, Luiz Capitulino 写道:
On Fri, 26 Apr 2013 16:46:57 +0200
Stefan Hajnoczi <stefa...@gmail.com> wrote:
On Fri, Apr 26, 2013 at 05:31:15PM +0800, Wenchao Xia wrote:
@@ -2586,10 +2585,12 @@ void do_info_snapshots(Monitor *mon, const QDict *qdict)
}
if (total > 0) {
- monitor_printf(mon, "%s\n", bdrv_snapshot_dump(buf, sizeof(buf),
NULL));
+ bdrv_snapshot_dump(NULL);
+ monitor_printf(mon, "\n");
Luiz: any issue with mixing monitor_printf(mon) and
monitor_vprintf(cur_mon) calls? I guess there was a reason for
explicitly passing mon instead of relying on cur_mon.
where are they being mixed?
bdrv_snapshot_dump() used a global variable "cur_mon" inside, instead
of let caller pass in a explicit montior* "mon", I guess that is the
question.
I'd have to see the code to tell, but yes, what Stefan described is the
best practice for the Monitor.
I think this would not be a problem until qemu wants more than one
human monitor console, and then we may require a data structure to tell
where to output the string: stdout, *mon, or even stderr, and
error_printf() also need to be changed.
--
Best Regards
Wenchao Xia
