Re: [Qemu-devel] [PATCH] hmp: delvm: use hmp_handle_error

[Markus Armbruster]

Cole Robinson <address@hidden> writes:

> This gives us the consistent 'Error:' prefix added in 66363e9a43f,
> which helps users like libvirt who still need to scrape hmp error
> messages to detect failure.
>
> Signed-off-by: Cole Robinson <address@hidden>
> ---
>  hmp.c | 7 ++++---
>  1 file changed, 4 insertions(+), 3 deletions(-)
>
> diff --git a/hmp.c b/hmp.c
> index 8eec768088..74a4bfc1f9 100644
> --- a/hmp.c
> +++ b/hmp.c
> @@ -1481,10 +1481,11 @@ void hmp_delvm(Monitor *mon, const QDict *qdict)
>      const char *name = qdict_get_str(qdict, "name");
>  
>      if (bdrv_all_delete_snapshot(name, &bs, &err) < 0) {
> -        error_reportf_err(err,
> -                          "Error while deleting snapshot on device '%s': ",
> -                          bdrv_get_device_name(bs));
> +        error_prepend(&err,
> +                      "Error while deleting snapshot on device '%s': ",
> +                      bdrv_get_device_name(bs));
>      }
> +    hmp_handle_error(mon, &err);
>  }
>  
>  void hmp_info_snapshots(Monitor *mon, const QDict *qdict)

No objection to this patch, just apropos hmp_handle_error().

HMP command handlers look like this:

    void hmp_FOO(Monitor *mon, const QDict *qdict)

They can report errors however they like.  The monitor core has no
notion of HMP command failure.

Commonly, hmp_FOO() wraps around some qmp_FOO(), or some helper(s) it
shares with qmp_FOO().  These will return errors through an Error **
argument.  The sane way for hmp_FOO() to report them is with
hmp_handle_error().

In other words, we get an hmp_handle_error() on most[*] failure paths.

Why not move it into the monitor core?

    bool hmp_FOO(Monitor *mon, const QDict *qdict, Error **errp)

While at it, ditch the @mon parameter, because it's always cur_mon
anyway:

    bool hmp_FOO(const QDict *qdict, Error **errp)


[*] Common exceptions are failures in code that add convenience over
QMP.  These need not produce an Error object.  Instead, they may report
with error_report(), or even monitor_printf().  The latter would be in
bad taste.