On Tue, Dec 03, 2013 at 04:56:27PM +0000, Jon TURNEY wrote: >Tried bisecting through the X server versions for the past 6 months, and it >seems that this problem first appears in X server 1.14.3-2 > >(As an aside, it's probably relevant to the recent discussion, that I can't >find the thread for, about how many previous versions we should keep around >that it's not possible to do this bisection without 'Secret Knowledge' at the >moment) > >This does contain a few clipboard changes, and in particular it changes the >way the messages get processed so we will return to the select() more often >(after each stage of the conversion operation), which it looks like it could >be incorrect sometimes, but I'd expect this to cause unneeded blocking >(waiting in select() for a message which has already been placed on the event >queue by XPending() rather than the observed behaviour. > >Anyhow, I guess I need to look at this some more... > >> Talking about wndproc.c, it only checks iReturn for being < 0. After >> that, we don't really know which value it has, we only know that >> FD_ISSET(iConnNumber, &fdsRead) returns 0. The value of iReturn should >> be printed in the debug output at line 133. > >It's 0 when I inspect it with the debugger, but yes, I'll change that. > >> What kind of object is the iConnNumber descriptor? Pipe? Fifo? >> Socket? /dev/windows? We really need a simple testcase without the >> X and emacs overhead... > >It's a socket connected to the X server.
I added an ugly hack to work around this symptom in the latest cygwin. It shouldn't have any big impact on anything but this particular scenario but I would appreciate it if people downloaded today's snapshot and verified that things are still working ok. I plan on addressing the actual problem for Cygwin 1.7.28. 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
