Re: [Qemu-devel] [PATCH 4/5] Switch migrate_set_speed() to take an 'o' argument rather than a float.

[Markus Armbruster <armbru@redhat.com>]

Jes.Sorensen@redhat.com writes:

> From: Jes Sorensen <Jes.Sorensen@redhat.com>
>
> Signed-off-by: Jes Sorensen <Jes.Sorensen@redhat.com>
> ---
>  migration.c     |    4 ++--
>  qemu-monitor.hx |    2 +-
>  2 files changed, 3 insertions(+), 3 deletions(-)
>
> diff --git a/migration.c b/migration.c
> index 468d517..9ee8b17 100644
> --- a/migration.c
> +++ b/migration.c
> @@ -132,10 +132,10 @@ int do_migrate_cancel(Monitor *mon, const QDict *qdict, QObject **ret_data)
>  
>  int do_migrate_set_speed(Monitor *mon, const QDict *qdict, QObject **ret_data)
>  {
> -    double d;
> +    int64_t d;
>      FdMigrationState *s;
>  
> -    d = qdict_get_double(qdict, "value");
> +    d = qdict_get_int(qdict, "value");
>      d = MAX(0, MIN(UINT32_MAX, d));
>      max_throttle = d;

This isn't backwards bug-compatible.

Before, a client could send any number.  Any fractional part was
ignored.

Now, the number must be an integer.  Other numbers are rejected.

I don't care myself, but others have argued most forcefully for keeping
QMP fully backward compatible from 0.13 on, so they might object.

> diff --git a/qemu-monitor.hx b/qemu-monitor.hx
> index 49bcd8d..7f58fb2 100644
> --- a/qemu-monitor.hx
> +++ b/qemu-monitor.hx
> @@ -1093,7 +1093,7 @@ EQMP
>  
>      {
>          .name       = "migrate_set_speed",
> -        .args_type  = "value:f",
> +        .args_type  = "value:o",
>          .params     = "value",
>          .help       = "set maximum speed (in bytes) for migrations",
>          .user_print = monitor_user_noop,

Doesn't this change the interpretation of "42" from 42 to (42 << 20)?