Hi Pete, Am 08.09.2014 19:31, schrieb Pete Zaitcev: > On Sat, 06 Sep 2014 16:06:28 +0200 > Alexander Kurpiers <[email protected]> wrote: > >> Next try... > Looks good as far as general structure goes. Style is a little odd, > perhaps. For example, /* Check the error code. */ comment is useless, > straight from the book of "a=b; /* assign a value of b to a */".
True - I left that from the original code. Could indeed be removed - and a command added instead about moving of dev->async_status = RTLSDR_RUNNING; which is necessary (otherwise the thread does not stop despite the error returned...). > When you unroll, the ideom is while(i--). If you hide the decrement > inside a [] inside the loop body, it's liable to get missed one day. Also true, but would need to be --i>=0, as i is the index of an entry that is already NULL. Not sure if that is nicer. I have not checked what the libusb_free_transfer() does if called with NULL. At least for v1.0 it seems ok (http://libusb.sourceforge.net/api-1.0/group__asyncio.html#ga6ab8b2cff4de9091298a06b2f4b86cd6) If that works for all variants of libusb, memset'ing after malloc (or calloc?) to zero the pointers and calling _rtlsdr_free_async_buffers() may be easier than unrolling (less code duplication). > But I'll let the maintainer take issues with style. You got the gist > of it right, I think. > Thanks ;-) Yet I would prefer a comment on patch 3 - that's the only one that I'm really interested in. This one is only to port all the changes between Leif's code and the current git code. Regards, Alexander
