Hi.
I noticed there is a 1 second delay in librtlsdr between calling rtlsdr_cancel_async and releasing from rtlsdr_read_async. I'm writing a radio scanner and this delay limits scanning speed considerably. I tracked the delay down to this piece of code in librtlsdr.c:
while (RTLSDR_INACTIVE != dev->async_status) { r = libusb_handle_events_timeout(dev->ctx, &tv); [...] if (RTLSDR_CANCELING == dev->async_status) { next_status = RTLSDR_INACTIVE; [...] if (dev->dev_lost || RTLSDR_INACTIVE == next_status) {
libusb_handle_events_timeout(dev->ctx, &tv);
break; }
The marked line is almost always(?) invoked when all the events are already processed, hence it blocks for tv = 1 second. In fact I don't understand what this `if' is supposed to do at all. I tried to modify this line by changing the delay to 0 making it a non-blocking call (leaving the timeout of 1 sec for the former libusb_handle_events_timeout call). This worked with my project and the 1 second delay disappeared. So, I suppose, it's better to either a) remove the whole `if' or b) make the latter libusb_handle_events_timeout non-blocking as I did. I checked that b) works for me, and didn't test a) yet.
The question is: what that `if (dev->dev_lost ...' is supposed to do?
And the two possible fixes for 1 second delay bug depending on the answer are (diffs): === 1630a1631
struct timeval tv_nb = { 0, 0 };
1696c1697 < libusb_handle_events_timeout(dev->ctx, &tv); ---
libusb_handle_events_timeout(dev->ctx, &tv_nb);
===
=== 1694,1698d1693 < < if (dev->dev_lost || RTLSDR_INACTIVE == next_status) { < libusb_handle_events_timeout(dev->ctx, &tv); < break; < } ===
Hi.
Since there's no discussion, I'll write a little more about that extra delay issue.
I noticed there is a 1 second delay in librtlsdr between calling rtlsdr_cancel_async and releasing from rtlsdr_read_async. I'm writing a
1) I suggest to incorporate the first of my two fixes into the master branch. The first fix sholdn't break anything, it works well for me and fixes the delay bug. I moved on to my other projects, and ain't going to investigate this problem any more.
And the two possible fixes for 1 second delay bug depending on the answer are (diffs): === 1630a1631
struct timeval tv_nb = { 0, 0 };
1696c1697
< libusb_handle_events_timeout(dev->ctx, &tv);
libusb_handle_events_timeout(dev->ctx, &tv_nb);
===
2) There was an error in my second fix: I obviously should have removed only the libusb_handle_events_timeout line, not the whole `if'.
=== 1694,1698d1693 < < if (dev->dev_lost || RTLSDR_INACTIVE == next_status) { < libusb_handle_events_timeout(dev->ctx, &tv); < break;
< }
-- GPG fp: 6DFA 1186 7576 DE60 6CCB EB39 AD81 1733 EEBB 970A