On 2013.04.11 13:41, Toby Gray wrote: > There is also an assumption that the reading of ints is > atomic as it checks the fd value at the same time as it might be copied > to by the submission of the transfer. However that's a bigger greater > problem to fix (Peter Stuge helpfully pointed me at the discussion here: > http://libusb.org/ticket/81).
Right, I kind of remember that one. > I've attached a patch for the usbi_free_fd change you suggested. This actually goes further than what I suggested, as I wasn't looking past providing a pointer to wfd.fd to usbi_free_fd(). But now that I think about it, we're really done with all our wfd operations there, so we might as reset the whole thing wholesale as you suggest. usbi_create_fd() returns a wfd, so it makes sense that its counterpart takes a wfd. > I think > it still doesn't solve the read of the fd while trying to find the > transfer in the flying transfers, but I'm not sure any of the current > locks can help with that. Yeah, the way I see it we still have some underlying issues in core, but I doubt we're going to have a chance to fix these any time soon due to resource constraints (as well as people pushing for new developments rather than bugfixes ;)). Your patch has now been pushed. Regards, /Pete ------------------------------------------------------------------------------ Precog is a next-generation analytics platform capable of advanced analytics on semi-structured data. The platform includes APIs for building apps and a phenomenal toolset for data science. Developers can use our toolset for easy data analysis & visualization. Get a free account! http://www2.precog.com/precogplatform/slashdotnewsletter _______________________________________________ libusbx-devel mailing list libusbx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libusbx-devel
