Hi, On 06/21/2013 12:16 AM, Pete Batard wrote: > On 2013.06.20 12:11, Hans de Goede wrote: >> Done (more or less, see git). > > Even with the dual set of patches, I'm still not sure we wouldn't want > to at least print a debug message if we find out that > libusb_cancel_transfer(...) failed. But I guess we can go with the > current proposal.
FWIW I deliberately did not add a log message for the cancel failing, since if we ever enter this code-path things are already falling apart pretty badly, so the cancel will likely fail too. libusb_handle_events should simply never fail. Note that the bug-report which made us look at this error handling path in the first place, involves an EFAULT, any other way of referencing the bad memory the app is passing us would result in instant dead of the program, instead we got "lucky" and only got an EFAULT (I would have preferred a segfault TBH). This is also the reason why I log an error on the first failure of libusb_handle_events, rather then only in the retry case. > On a side note, I'd very much prefer a cool-off period of at least one > day between the last non-minor commit being pushed and RC (the patches > you pushed against sync.c don't look that minor to me). Understood, although I think it would not have helped, since as said this error handling path (almost) never gets executed. > > I know we're all in a hurry to get RC off the ground, but I've had too > many bad surprises from seemingly innocuous fixes breaking Windows, and > I can't help getting nervous seeing something flagged as RC before it > has been tested on all the Windows platforms. > > This being said, I have just re-run the usual set of VS2012+Code > Analysis, MinGW-Clang, cygwin, WDK (also threw in MinGW-w64 for good > measure) and things look OK as far as compilation is concerned. As always thanks for all your work on libusb. Regards, Hans ------------------------------------------------------------------------------ This SF.net email is sponsored by Windows: Build for Windows Store. http://p.sf.net/sfu/windows-dev2dev _______________________________________________ libusbx-devel mailing list libusbx-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/libusbx-devel
