On 2013.01.16 17:08, Toby Gray wrote: > There are a couple of changes that are needed to get it to build on WinCE. > See the attached patch.
Ah, my bad. I had to apply a few changes to poll_windows, and decided I might as well use the same naming conventions for the attributes you added, but it looks like I missed a couple changes. I'll try to include the necessary fixes in the existing commit. > I don't quite understand why poll_windows.h now needs to explicitly include > <libusb.h> (for ssize_t). I assume it must have just been luck before > that all files which #included poll_windows.h happened to have already > included libusb.h. Well that's relevant to one of the issue I had to fix in your code actually, and part of what I wanted to mention. You added a struct usbi_transfer attribute in poll_windows.h's struct winfd. However, struct usbi_transfer is defined in libusbi.h, which itself included poll_windows.h *before* the definition of struct usbi_transfer, so the compiler was (rightfully) complaining about it. Thus I had to move the #include of poll_windows.h after struct usbi_transfer in libusbi.h [1]. I was also able to remove the need for the libusb.h include there, and I would prefer to keep that header away if it's just for a ssize_t definition (which is indeed defined in libusb.h for platforms that don't have it). Looking at the code, I think the reason you get the ssize_t issue (and other Windows platforms don't) is because you don't include libusbi.h (which in turn includes libusb.h) before poll_windows.h, in wince_usb.h. I'd also prefer if the only include in wince_usb.h was for windows_common.h, and the other ones moved to wince_usb.c (before the #include wince_usb.h), though I don't really care about this kind of cleanup occurring after the current proposal have been added to mainline, as there's some other cleanup that I'd like to perform (libusb.h does include <sys/types.h> if available, so I think we should be able to remove some of the #ifdef'd code from the samples). The other issue I had to take care of was that thread_windows is not being used on all Windows platforms: cygwin uses thread_posix. Therefore the usbi_sleep() was causing us issues there, and I duplicated as a simple one liner define in both poll_windows and threaf_windows. Finally, to remove another WinCE ifdef in core.c, I'm wondering if we couldn't define an inline GetSystemTimeAsFileTime() for CE in missing.h. I'm also thinking about defining a fake SleepEx() there, to remove the need for usbi_sleep altogether. Again, these are two items that I'd see better suited to do post integration, as I don't see much of a problem leaving extra cleanup for later. For now, if you can confirm that the inclusion of libusbi.h, either in wince_usb.h or wince_usb.c removes the need for the libusb.h in poll_windows, I think I'll have enough to proceed with the integration, though I'll probably wait a few days to try to run some more tests before I push anything. Regards, /Pete [1] https://github.com/pbatard/libusbx/commit/29c5c85f4c77ee965d51492cb84da96d13da528d ------------------------------------------------------------------------------ Master Java SE, Java EE, Eclipse, Spring, Hibernate, JavaScript, jQuery and much more. Keep your Java skills current with LearnJavaNow - 200+ hours of step-by-step video tutorials by Java experts. SALE $49.99 this month only -- learn more at: http://p.sf.net/sfu/learnmore_122612 _______________________________________________ libusbx-devel mailing list libusbx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libusbx-devel
