On Thu, Jul 21, 2011 at 03:57:46PM -0400, Christopher Faylor wrote: >On Thu, Jul 21, 2011 at 07:47:17PM +0200, Corinna Vinschen wrote: >>On Jul 21 13:05, Christopher Faylor wrote: >>> On Thu, Jul 21, 2011 at 04:39:31PM +0200, Corinna Vinschen wrote: >>> >On Jul 20 17:37, Yaakov (Cygwin/X) wrote: >>> >> On Wed, 2011-07-20 at 17:46 +0200, Corinna Vinschen wrote: >>> >> > On Jul 20 03:11, Yaakov (Cygwin/X) wrote: >>> >> > > On Linux, ioctl(2) returns several different errors[1]: >>> >> > > >>> >> > > EBADF d is not a valid descriptor. >>> >> > > EFAULT argp references an inaccessible memory area. >>> >> > > EINVAL Request or argp is not valid. >>> >> > > ENOTTY d is not associated with a character special device. >>> >> > > ENOTTY The specified request does not apply to the kind of object >>> >> > > that >>> >> > > the descriptor d references. >>> >> > > >>> >> > > In the case of FIONREAD, Cygwin doesn't seem to distinguish between >>> >> > > EINVAL and ENOTTY, and this causes at least one major bug: >>> >> > > >>> >> > > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=35536 >>> >> > > >>> >> > > I have patched GCJ and GNU classpath to work around it, but this >>> >> > > really >>> >> > > needs to be fixed in Cygwin itself. >>> >> > >>> >> > Would this patch be sufficient? >>> >> > >>> >> > Index: fhandler.cc >>> >> > =================================================================== >>> >> > RCS file: /cvs/src/src/winsup/cygwin/fhandler.cc,v >>> >> > retrieving revision 1.397 >>> >> > diff -u -p -r1.397 fhandler.cc >>> >> > --- fhandler.cc 5 Jul 2011 12:02:10 -0000 1.397 >>> >> > +++ fhandler.cc 20 Jul 2011 15:46:40 -0000 >>> >> > @@ -1151,6 +1151,10 @@ fhandler_base::ioctl (unsigned int cmd, >>> >> > set_nonblocking (*(int *) buf); >>> >> > res = 0; >>> >> > break; >>> >> > + case FIONREAD: >>> >> > + set_errno (ENOTTY); >>> >> > + res = -1; >>> >> > + break; >>> >> > default: >>> >> > set_errno (EINVAL); >>> >> > res = -1; >>> >> >>> >> Given my testcases, give me a few days to check this out. >>> > >>> >Ok, but the patch doesn't build as is. You have to add >>> > >>> > #include <asm/socket.h> >>> > >>> >to get the FIONREAD definition. Sorry about that. >>> >>> Btw, I don't think the above is sufficient. I think you'll have to add a >>> similar test to other fhandlers (like fhandler_windows) which completely >>> implement ioctl without reverting to fhandler_base. >> >>Hmm, isn't that just a bug in these ioctl methods? I think they >>all should fall through to fhandler_base::ioctl. > >Yep. I think you're right. I'll handle that if you want.
I made this modification and added the above patch as well. We can always pull it out or modify it if it doesn't do the right thing. A snapshot is building now. cgf -- Problem reports: http://cygwin.com/problems.html FAQ: http://cygwin.com/faq/ Documentation: http://cygwin.com/docs.html Unsubscribe info: http://cygwin.com/ml/#unsubscribe-simple
