On 20.02.2018 22:34, Raman Shishniou wrote:
> On 02/20/2018 11:04 PM, Georg Chini wrote:
>> On 20.02.2018 19:49, Raman Shishniou wrote:
>>> On 02/20/2018 07:02 PM, Georg Chini wrote:
>>>> On 20.02.2018 16:38, Raman Shyshniou wrote:
>>>>> Currently the pipe-source will remain running even if no
>>>>> writer is connected and therefore no data is produced.
>>>>> This patch adds the autosuspend=<bool> option to prevent this.
>>>>> Source will stay suspended if no writer is connected.
>>>>> This option is enabled by default.
>>>>> ---
>>>>> src/modules/module-pipe-source.c | 279 +++++++++++++++++++++++++++++----------
>>>>> 1 file changed, 212 insertions(+), 67 deletions(-)
>>>>>
>>> I think I need post a simple pseudo code of new thread loop because it
>>> was completely rewritten. There are too many changes in one patch.
>>> It can be difficult to see the whole picture of new main loop.
>> Well, I applied the patch and looked at the result. I still don't like the approach.
>>
>> I would propose this:
>>
>> auto_suspended = false;
>> revents = 0
>> events = POLLIN
>>
>> for (;;) {
>>
>> /* This is the part that is run when the source is opened
>> * or auto suspended
>> if (SOURCE_IS_OPENED(source) || auto_suspended) {
>>
>> /* Check if we wake up from user suspend */
>> if (corkfd >= 0 && !auto_suspended) {
>> len = 0
>> close pipe for writing
>> }
>>
>> /* We received POLLIN or POLLHUP or both */
>> if (revents) {
>>
>> /* Read data from pipe */
>> len = read data
>>
>> /* Got data, post it */
>> if (len > 0) {
>> if (auto_suspend) {
>> send unsuspend message
>> auto_suspend = false
>> }
>> post data
> We cannot post data here because source still suspended. Sending resume message is not enough
> to immediately resume the source. We need to wait several poll runs until it will be resumed.
> (source->thread_info.state changed in this thread, i.e. during poll run). But we will see
> POLLIN and/or POLLHUP each run if we don't remove pipe fd from polling.
Why do we have to wait? The source will be unsuspended on the next
rtpollrun.
I do not see why we cannot already push data. Or does something get lost?
