Hi Pekka,
On 21 November 2016 at 13:30, Pekka Paalanen wrote:
> On Mon, 21 Nov 2016 12:19:43 + Daniel Stone wrote:
>> I'm would suggest we document that invoking wl_display_flush_clients() /
>> wl_connection_flush() from a request handler may cause undefined behaviour,
>> and that it must on
On Mon, 21 Nov 2016 12:19:43 +
Daniel Stone wrote:
> Hi Hyun Kook,
>
> On 23 September 2016 at 00:40, Hyun Kook Khang
> wrote:
>
> > in wl_client_connection_data(),
> > there is no guarantee that client would never be destroyed in the process
> > of handling the pending input.
> >
> > For
Hi Hyun Kook,
On 23 September 2016 at 00:40, Hyun Kook Khang
wrote:
> in wl_client_connection_data(),
> there is no guarantee that client would never be destroyed in the process
> of handling the pending input.
>
> For example,
> 1. It comes to wl_client_connection_data()
> 2. wl_closure_invoke(
Hi,
could you write how to trigger the crash, also in the commit message
maybe? Besides that, i have a comment inline below.
2016-09-21 9:08 GMT+02:00 Hyunkook Khang :
> While processing pending data, client could be destroyed in the middle of
> the process. (e.g. by invoking wl_display_flush_cli
While processing pending data, client could be destroyed in the middle of
the process. (e.g. by invoking wl_display_flush_clients()).
In this case, client will be freed, but we are still in the processing data
of the client, so it could cause a crash.
To avoid this, instead of destroying the clien
