[Spice-devel] [PATCH spice-gtk 1/5] Block sending clipboard data > max-clipboard

[Uri Lublin]

On 11/08/2013 01:06 AM, Marc-André Lureau wrote:
> On Thu, Nov 7, 2013 at 11:47 PM, Uri Lublin <uril at redhat.com> wrote:
>>
>>> diff --git a/gtk/channel-main.c b/gtk/channel-main.c
>>> index 5c33e67..dbcaff8 100644
>>> --- a/gtk/channel-main.c
>>> +++ b/gtk/channel-main.c
>>> diff --git a/gtk/spice-gtk-session.c b/gtk/spice-gtk-session.c
>>> index 71ed300..eab7e2f 100644
>>> --- a/gtk/spice-gtk-session.c
>>> +++ b/gtk/spice-gtk-session.c
>>> @@ -748,16 +748,19 @@ static void clipboard_received_cb(GtkClipboard
>>> *clipboard,
>>>        gchar* name;
>>>        GdkAtom atom;
>>>        int selection;
>>> +    int max_clipboard;
>>>          selection = get_selection_from_clipboard(s, clipboard);
>>>        g_return_if_fail(selection != -1);
>>>    +    g_object_get(s->main, "max-clipboard", &max_clipboard, NULL);
>>>        len = gtk_selection_data_get_length(selection_data);
>>> -    if (len == -1) {
>>> +    if (len == 0 || (max_clipboard != -1 && len > max_clipboard)) {
>>> +        g_warning("discarded clipboard of size %d (max: %d)", len,
>>> max_clipboard);
>>> +        return;
>>> +    } else if (len == -1) {
>>>            SPICE_DEBUG("empty clipboard");
>>>            len = 0;
>>> -    } else if (len == 0) {
>>> -        SPICE_DEBUG("TODO: what should be done here?");
>>>        } else {
>>>            atom = gtk_selection_data_get_data_type(selection_data);
>>>            name = gdk_atom_name(atom);
>>
>> Why did you switch the logic of len==0 and len==-1.
>> I think we can drop the clipboard for both cases.
>> Is it possible to get such values (0 and -1) for len ?
> -1 is for empty clipboard, or fail to retrieve clipboard data (for
> various reasons).
>
> 0 could eventually happen, but when? an app claims to have clipboard
> data, but provides 0 data? I don't know, I thought a warning made
> sense, instead of silently sending 0 data, but perhaps the current
> behaviour is more correct. I can split that in a separate patch, or
> drop that little change. And perhaps we should always send empty
> clipboard msg on error to the guest actually. Afaik, there is no other
> way to tell to the guest app to stop waiting for clipboard data. Any
> idea?

 From a quick search I could not tell when gtk_selection_data_get_length 
returns
-1 and when it returns 0.
I expect -1 (<0) for failures, and 0 for empty clipboard.
Maybe a clipboard selection from a program that later exited is of length 0.
Or maybe the user can "clear clipboard" which results in a claim to have 
a selection of length 0.
I'm just guessing, need to read more documentation.

It may be good to send 0 data for failed clipboard too.
Another way is to send CLIPBOARD_RELEASE message.
Also after timeout vdagent stops waiting (but better not wait if it can 
let know if should not).