OK. We should probably figure out a portable way to deal with the PDL datatypes issue since it will become important when we eventually support 64bit indexing for piddles. Thanks for testing the fix.... I think it is about time to have another swing at building PLplot on cygwin.
--Chris On Fri, Nov 18, 2011 at 1:02 PM, Doug Hunt <[email protected]> wrote: > Hi Chris: Your latest fix works for both Linux and Mac and looks more > general (IE more likely to work on Windows) than my earlier fix. > > Lets go with that! > > --Doug > > [email protected] > Software Engineer > UCAR - COSMIC, Tel. (303) 497-2611 > > On Fri, 18 Nov 2011, Chris Marshall wrote: > >> Hmm. The 'long' for $im_ptr_type will break for platforms with >> 32bit long but 64bit pointers. I think the C type should be using >> the values from %Config and not the PDL types which I did not >> understand. I've pushed a modified fix which should work. >> Please let me know. >> >> Thanks, >> Chris >> >> On Fri, Nov 18, 2011 at 11:16 AM, Doug Hunt <[email protected]> wrote: >>> >>> Hi Chris: Yeah, that breaks it. >>> >>> the $pp_ptr_type needs to be 'longlong' after the PDL data type. Also, >>> I64 >>> is not defined on my (Linux) system. >>> >>> This is what I can get working: >>> >>> my $ptrsize = $Config{'ptrsize'}; >>> my $int_ptr_type = ($ptrsize == 4) ? 'I32' : 'long'; # pointer type >>> in C >>> code >>> my $pp_ptr_type = ($ptrsize == 4) ? 'int' : 'longlong'; # PDL data type >>> in >>> pp_def 'Pars' >>> >>> I don't see any problem with leaving it like this. >>> >>> Regards, >>> >>> Doug >>> >>> [email protected] >>> Software Engineer >>> UCAR - COSMIC, Tel. (303) 497-2611 >>> >>> On Fri, 18 Nov 2011, Chris Marshall wrote: >>> >>>> Hi Doug and Derek- >>>> >>>> I just pushed a slight modification of the detection of ptrsize >>>> et. al. in plplot.pd. It uses the explicit values from Config which >>>> should be more robust. Would you mind confirming that I did >>>> not break anything? >>>> >>>> Thanks much, >>>> Chris >>>> >>>> On Thu, Nov 17, 2011 at 5:57 PM, Doug Hunt <[email protected]> wrote: >>>>> >>>>> Hi Derek: Nice meeting you this morning. >>>>> >>>>> I've put in a change to the PDL git that seems to fix the problem and >>>>> works >>>>> both on Linux and Mac. It was simple. The pp_def for plAllocGrid and >>>>> plAlloc2dGrid passed out a 64 bit pointer using a 32 bit PDL 'int' >>>>> value: >>>>> >>>>> (in plplot.pd) >>>>> >>>>> pp_def ('plAllocGrid', >>>>> NoPthread => 1, >>>>> Pars => "double xg(nx); double yg(ny); int [o] grid()", >>>>> >>>>> I just changed this to: >>>>> >>>>> pp_def ('plAllocGrid', >>>>> NoPthread => 1, >>>>> Pars => "double xg(nx); double yg(ny); $pp_ptr_type [o] grid()", >>>>> >>>>> where $pp_ptr_type is either 'int' (for 32 bit systems) or 'longlong' >>>>> for >>>>> 64 >>>>> bit systems. >>>>> >>>>> Let me know if you have any more troubles on your Mac after pulling the >>>>> latest changes from PDL git. >>>>> >>>>> Regards, >>>>> >>>>> Doug >>>>> >>>>> [email protected] >>>>> Software Engineer >>>>> UCAR - COSMIC, Tel. (303) 497-2611 >>>>> >>>>> On Tue, 15 Nov 2011, Derek Lamb wrote: >>>>> >>>>>> Back when this was bothering me I think I had PLplot 5.9.6 (or 7) >>>>>> installed. I'll get 5.9.9 installed and the latest PDL git built >>>>>> against >>>>>> it--I have just been using whatever P::G::P is in the PDL source tree, >>>>>> which >>>>>> looks right now to be 0.61. None of the PLplot library examples fail, >>>>>> just >>>>>> the PDL bindings. >>>>>> >>>>>> Does Thursday morning work? I have a lunch appointment but hopefully >>>>>> we >>>>>> will be done before then! I'm coming from Broomfield, I'll take the >>>>>> bus >>>>>> to >>>>>> Table Mesa and then the UCAR shuttle from there. I'll shoot for the >>>>>> 9:13 >>>>>> shuttle that should have me to the FL complex around 9:30. If for >>>>>> some >>>>>> reason I miss that then I'll just bike up, it's easier than waiting >>>>>> for >>>>>> the >>>>>> shuttle and Thursday looks to be a nice day for a ride. I'm assuming >>>>>> there's a receptionist at the entrance like at HAO so they could >>>>>> direct >>>>>> me >>>>>> your way? >>>>>> >>>>>> cheers, >>>>>> Derek >>>>>> >>>>>> On Nov 14, 2011, at 4:43 PM, Doug Hunt wrote: >>>>>> >>>>>>> Hi Derek: I have not looked into this bug since (until recently) I >>>>>>> did >>>>>>> not have a Mac. If you've got a laptop that can duplicate the >>>>>>> problem, >>>>>>> perhaps you could come to my office and we could take a look at it. >>>>>>> >>>>>>> I've done a fair amount of debugging in the plplot library and just >>>>>>> refactored the PDL::Graphics::PLplot interface so I'm still fairly >>>>>>> familiar with the code and might be able to help. >>>>>>> >>>>>>> Have you got the plplot library source installed and compiled >>>>>>> manually? >>>>>>> It will probably come down to putting print statements in >>>>>>> plplot--it's >>>>>>> good >>>>>>> to have a PDL version working that uses a locally compiled version of >>>>>>> plplot >>>>>>> so you can poke around both in PDL::Graphics::PLplot and in plplot >>>>>>> itself. >>>>>>> >>>>>>> What versions of PDL, PDL::Graphics::PLplot and plplot are you using? >>>>>>> >>>>>>> I'm in the new 'FLA' building at UCAR Foothills lab--3375 Mitchell >>>>>>> Lane--the old curving Wild Oats building. What time would be good >>>>>>> for you to meet? >>>>>>> >>>>>>> Regards, >>>>>>> >>>>>>> Doug Hunt >>>>>>> >>>>>>> [email protected] >>>>>>> Software Engineer >>>>>>> UCAR - COSMIC, Tel. (303) 497-2611 >>>>>>> >>>>>>> On Mon, 14 Nov 2011, Derek Lamb wrote: >>>>>>> >>>>>>>> Hi Doug, >>>>>>>> >>>>>>>> Since the PDL bug "PLplot segfaults on plshades call on OS X" was >>>>>>>> reported by me, but I've been unable to figure out just what is >>>>>>>> screwy, and >>>>>>>> since you are the current maintainer of PDL::Graphics::PLplot and >>>>>>>> probably >>>>>>>> know more about the inner PLplot workings than I do, and since that >>>>>>>> bug has >>>>>>>> a severity of 9, and since we both work in Boulder, would you be >>>>>>>> interested >>>>>>>> in getting together sometime maybe before Thanksgiving and trying to >>>>>>>> hammer >>>>>>>> that out? I can easily get to UCAR's Foothills lab by bike from >>>>>>>> SwRI >>>>>>>> if I >>>>>>>> know to bring the bike with me, or I can get to the Mesa lab first >>>>>>>> thing in >>>>>>>> the morning some day via the UCAR shuttle. >>>>>>>> >>>>>>>> cheers, >>>>>>>> Derek >>>>>>>> >>>>>>> >>>>>> >>>>>> >>>>> >>>>> _______________________________________________ >>>>> Perldl mailing list >>>>> [email protected] >>>>> http://mailman.jach.hawaii.edu/mailman/listinfo/perldl >>>>> >>> > _______________________________________________ Perldl mailing list [email protected] http://mailman.jach.hawaii.edu/mailman/listinfo/perldl
