[Xcb] [PATCH 1/1] read from connection when polling special events and replies

[Uli Schlachter]

On 29.04.2017 15:14, Uli Schlachter wrote:
> On 19.04.2017 04:58, David McFarland wrote:
>> Using the mesa vulkan driver, if you acquire an image from a
>> swapchain using a finite timeout (x11_acquire_next_image_poll_x11),
>> it will occasionally lock, calling xcb_poll_for_special_event in
>> a loop until the timeout expires.
>>
>> Call _xcb_in_read() once from the polling functions for special
>> events and replies, in the same way as xcb_poll_for_event.
>>
>> Signed-off-by: David McFarland <corngood at gmail.com>
>> ---
>>  src/xcb_in.c | 6 ++++++
>>  1 file changed, 6 insertions(+)
>>
>> diff --git a/src/xcb_in.c b/src/xcb_in.c
>> index bab4bc7..73209e0 100644
>> --- a/src/xcb_in.c
>> +++ b/src/xcb_in.c
>> @@ -661,6 +661,8 @@ int xcb_poll_for_reply(xcb_connection_t *c, unsigned int request, void **reply,
>>      assert(reply != 0);
>>      pthread_mutex_lock(&c->iolock);
>>      ret = poll_for_reply(c, widen(c, request), reply, error);
>> +    if(!ret && c->in.reading == 0 && _xcb_in_read(c)) /* _xcb_in_read shuts down the connection on error */
>> +        ret = poll_for_reply(c, widen(c, request), reply, error);
>>      pthread_mutex_unlock(&c->iolock);
>>      return ret;
>>  }
>> @@ -678,6 +680,8 @@ int xcb_poll_for_reply64(xcb_connection_t *c, uint64_t request, void **reply, xc
>>      assert(reply != 0);
>>      pthread_mutex_lock(&c->iolock);
>>      ret = poll_for_reply(c, request, reply, error);
>> +    if(!ret && c->in.reading == 0 && _xcb_in_read(c)) /* _xcb_in_read shuts down the connection on error */
>> +        ret = poll_for_reply(c, request, reply, error);
>>      pthread_mutex_unlock(&c->iolock);
>>      return ret;
>>  }
>> @@ -768,6 +772,8 @@ xcb_generic_event_t *xcb_poll_for_special_event(xcb_connection_t *c,
>>          return 0;
>>      pthread_mutex_lock(&c->iolock);
>>      event = get_special_event(c, se);
>> +    if(!event && c->in.reading == 0 && _xcb_in_read(c)) /* _xcb_in_read shuts down the connection on error */
>> +        event = get_special_event(c, se);
>>      pthread_mutex_unlock(&c->iolock);
>>      return event;
>>  }
>>
> 
> Reviewed-by: Uli Schlachter <psychon at znc.in>
> 
> (Anyone wants to guess how long it takes for the patch adding the
> `_queued`-variant of these function to appear?)

Pushed to git.

-- 
“Cold. Cold. Cold. Cold. Cold. Cold. Cold. Cold. Cold. Cold.” – Anna