On Tue, May 25, 2010 at 12:32:37PM +0100, Leo wrote: > On 2010-05-25 10:54 +0100, Andrew Ross wrote: > > So just to be clear, you have only observed the crash with the following > > interactive devices: xcairo, qtwidget > > > > Have you tried xwin on Linux? This has far fewer external dependencies > > and the thread support for screen refresh is in the driver (so we have > > more chance of debugging!) > > I have been doing a round trip. Some time ago xwin and xcairo crashed > lisp due to closing the plot window by clicking on the close button. So > I moved to qtwidget. > > But qtwidget and some others crashed lisp when running in another thread > (which is what happens when running in slime). > > XWIN seems to be working fine. See this: http://imagebin.org/98248 > except I need to make sure I remember to exit the plot window with SPC. > Otherwise the plot window will not respond to any user interaction and > will remain on the screen until lisp exits.
Right. That helps. Alan's comments about using the svn version apply here. I've recently fixed up some of the issues related to closing the window with the close button for the xwin driver so you may find that works. > > You have not observed problems with psc or svg drivers. What about the > > qt or cairo file based drivers (e.g. pscairo / epsqt)? > > EPSQT has the same issue as qtwidget. PSCAIRO has the same issue as > xcairo. Right, that is useful to know. It's not an issue with the interactive drivers then, but something deeper. Qt in particular has it's own thread support, and it is possible that something in there is interfering with your lisp thread support. I will look again at the qt mutex code to check for possible races. > > > You might try running with valgrind as an alternative to gdb. It tends > > to be quite good at finding issues related to memory access / memory > > allocation. > > I am not familiar with valgrind and i got "Unsupported arch_prtctl > option" trying to use it by running: > > valgrind devel/ccl/lx86cl64 OK. It's probably more tricky with lisp. Your backtrace already provides some clues. Andrew ------------------------------------------------------------------------------ _______________________________________________ Plplot-general mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/plplot-general
