Re: [Plplot-devel] Removed dynamic casts from Qt driver
On Mon, Jul 20, 2009 at 03:29:56PM +0100, Rochel, Alban wrote: Hello all, Due to the issues related to dynamic_casts, and as those were only used to make the code compacter, I`ve changed the way the driver works as follows: - Before: 1 callback function common to all Qt devices (e.g. plD_line_qt), calling the proper implementations making tests on the results of dynamic_casts - After: 1 callback function for every Qt device (e.g. plD_line_svgqt), using simple C-style casts. If you prefer C++ reinterpret_casts, just ask. This is by far the simplest solution I can think about. If this suits your needs, I will then proceed with other updates later. While I`m working on this, Dmitri Gribenko complained about the fact that all the class definitions were bulked in one header, which is not conventional in C++, but is functional and avoids having many header files for a single driver. I can change that of course, I just want an official green light for this, as it remained as the state of a discussion. Dear Alban, As an aside, make sure you get the latest update. I've been looking at fixing the qt visibility issues today and have made a few changes as a result. Looking at the code I only see 2 versions of all the callback functions a standard one and an extended one for extqt and qtwidget. Is this correct? If so, then it is not a large increase in code size to remove the dynamic casting if it is problematic. In terms of splitting up qt.h into lots of class specific files, I know this is the normal C++ way of doing things, but this will generate a lot of extra headers for just 1 driver. I think the clutter this makes will outway any C++ elegance benefits. That's my opinion. Others may disagree. Slightly off topic - I've now got gcc -fvisibility=hidden working with the qt driver and with the C++ qt_example.cpp. It does not work with the pyqt4 bindings yet. There are a couple of issues here (I think). 1) The sip generated code does not export the symbols that python requires. It should be possible to massage the source code if required. 2) The moc generated code for qt also is not visibility aware and I think this is what is leading to the missing symbols in plplot_pyqt4.so at the moment. I've not yet worked out how (or indeed even if it is possible) to massage the code by hand to fix this. Andrew -- Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge ___ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
[Plplot-devel] QT rpath bug
Alan, When implementing rpath support for the libplplotqtd library you have added the qt path in front of the rest of the rpath directories. From the comments you added I think this is to ensure that you pick up your custom built qt libraries rather than the system ones. Unfortunately if you are using the system qt libraries this adds /usr/lib at the front of the rpath directory list and therefore you end up linking with system version of the plplot libraries rather than those in the install location. For me this is a headache. Of course the proper solution is to order the rpath variables so that the install prefix always appears first, but I can't see an easy way of doing that with your current approach. Any thoughts? Andrew -- Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge ___ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
[Plplot-devel] RE : Removed dynamic casts from Qt driver
Hello Andrew, Here is a version of the same patch based on R10166. There were indeed conflicts with your changes. There should be about 5 versions for the callbacks: - rasterqt (for png, bmp...), sometimes specialised as pngqt, bmpqt... where needed - svgqt - epspdfqt (sometimes epsqt and pdfqt) - qtwidget - extqt I`m afraid I cannot help a lot with the Python bindings as I have absolutely no experience with them (or with Python more generally). But if you have ideas of what should be changed in the C++ aspects of the driver, feel free to ask. Alban De : an.r...@virgin.net [an.r...@virgin.net] de la part de Andrew Ross [andrewr...@users.sourceforge.net] Date d'envoi : lundi 20 juillet 2009 15:55 À : Rochel, Alban Cc : PLplot development list Objet : Re: [Plplot-devel] Removed dynamic casts from Qt driver On Mon, Jul 20, 2009 at 03:29:56PM +0100, Rochel, Alban wrote: Hello all, Due to the issues related to dynamic_casts, and as those were only used to make the code compacter, I`ve changed the way the driver works as follows: - Before: 1 callback function common to all Qt devices (e.g. plD_line_qt), calling the proper implementations making tests on the results of dynamic_casts - After: 1 callback function for every Qt device (e.g. plD_line_svgqt), using simple C-style casts. If you prefer C++ reinterpret_casts, just ask. This is by far the simplest solution I can think about. If this suits your needs, I will then proceed with other updates later. While I`m working on this, Dmitri Gribenko complained about the fact that all the class definitions were bulked in one header, which is not conventional in C++, but is functional and avoids having many header files for a single driver. I can change that of course, I just want an official green light for this, as it remained as the state of a discussion. Dear Alban, As an aside, make sure you get the latest update. I've been looking at fixing the qt visibility issues today and have made a few changes as a result. Looking at the code I only see 2 versions of all the callback functions a standard one and an extended one for extqt and qtwidget. Is this correct? If so, then it is not a large increase in code size to remove the dynamic casting if it is problematic. In terms of splitting up qt.h into lots of class specific files, I know this is the normal C++ way of doing things, but this will generate a lot of extra headers for just 1 driver. I think the clutter this makes will outway any C++ elegance benefits. That's my opinion. Others may disagree. Slightly off topic - I've now got gcc -fvisibility=hidden working with the qt driver and with the C++ qt_example.cpp. It does not work with the pyqt4 bindings yet. There are a couple of issues here (I think). 1) The sip generated code does not export the symbols that python requires. It should be possible to massage the source code if required. 2) The moc generated code for qt also is not visibility aware and I think this is what is leading to the missing symbols in plplot_pyqt4.so at the moment. I've not yet worked out how (or indeed even if it is possible) to massage the code by hand to fix this. Andrew qt_patch Description: qt_patch -- Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge___ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
Re: [Plplot-devel] unexported symbols for libplplotqtd.so (and plplot_pyqt4.so)
On Mon, Jul 20, 2009 at 11:22:32AM -0700, Alan Irwin wrote: Hi Andrew: I have changed the subject line appropriately. On 2009-07-20 15:55+0100 Andrew Ross wrote: Slightly off topic - I've now got gcc -fvisibility=hidden working with the qt driver and with the C++ qt_example.cpp. It does not work with the pyqt4 bindings yet. There are a couple of issues here (I think). 1) The sip generated code does not export the symbols that python requires. It should be possible to massage the source code if required. 2) The moc generated code for qt also is not visibility aware and I think this is what is leading to the missing symbols in plplot_pyqt4.so at the moment. I've not yet worked out how (or indeed even if it is possible) to massage the code by hand to fix this. Thanks very much for dealing with the visibility issues caused by splitting the old qt.cpp source code into qt.cpp and plqt.cpp. I have now added one additional cleanup there (revision 10167) to remove some unneeded logic in pldll.h.in. I agree with your comment (1) above that the visibility of the symbols defined in plplot_pyqt4.so will be an issue when python attempts to dynamically load that shared object. But I believe the first priority is to deal with the remaining unexported symbols in libplplotqtd.so that are causing undefined symbols for plplot_pyqt4.so. The current undefined symbol status is as follows. softw...@raven ldd -r drivers/qt.so 21 |grep undefined softw...@raven ldd -r bindings/qt_gui/libplplotqtd.so 21 |grep undefined softw...@raven ldd -r bindings/qt_gui/pyqt4/plplot_pyqt4.so 21 |grep undefined undefined symbol: _ZTI11QtExtWidget (bindings/qt_gui/pyqt4/plplot_pyqt4.so) undefined symbol: _ZTI10QtPLWidget (bindings/qt_gui/pyqt4/plplot_pyqt4.so) undefined symbol: _ZN11QtExtWidget16staticMetaObjectE (bindings/qt_gui/pyqt4/plplot_pyqt4.so) undefined symbol: _ZN10QtPLWidget16staticMetaObjectE (bindings/qt_gui/pyqt4/plplot_pyqt4.so) undefined symbol: _ZN10QtPLWidget11qt_metacastEPKc (bindings/qt_gui/pyqt4/plplot_pyqt4.so) undefined symbol: _ZN11QtExtWidget11qt_metacallEN11QMetaObject4CallEiPPv(bindings/qt_gui/pyqt4/plplot_pyqt4.so) undefined symbol: _ZN10QtPLWidget11qt_metacallEN11QMetaObject4CallEiPPv (bindings/qt_gui/pyqt4/plplot_pyqt4.so) undefined symbol: _ZN11QtExtWidget11qt_metacastEPKc (bindings/qt_gui/pyqt4/plplot_pyqt4.so) I have double-checked that all these eight symbols are defined in but not exported from libplplotqtd. You are probably aware of this already, but I just discovered you can correlate these symbols with the demangled symbols using the appropriate nm options and grep. For example: softw...@raven nm --defined-only bindings/qt_gui/libplplotqtd.so |grep _ZTI11QtExtWidget 00214b50 d _ZTI11QtExtWidget softw...@raven nm --defined-only --demangle bindings/qt_gui/libplplotqtd.so |grep 00214b50 00214b50 d typeinfo for QtExtWidget That demangled symbol information for all eight symbols may help you to figure out what is really going on with visibility. I am curious about your comment (2) above. Is there an example of a externally required symbol in the generated moc_qt.cxx that cannot be exported by some qt.h change involving the PLDLLIMPEXP_QT macro? Alan, I've also done the same thing with ldd / nm. I get the same missing symbols as you. When you run moc it generates some extra functions in the the C++ class to deal with the qt signals / slots. The Q_OBJECT line in the C++ classes denotes the classes that this applies to. If you look through qt.h you will see that there are unexported typeinfo information for _only_ those classes which contain Q_OBJECT, even though everything else is that same (e.g. c.f. the typeinfo entries for QtPLDriver and QTPLWidget). The other entries, those containing metaObject or metacast are generated by moc (look in the source file). I've tried manually adding in the required attribute information to these files but it doesn't compile. It seems that these are being defined elsewhere through QT macros - I imagine an expansion of the Q_OBJECT tag - and the definitions conflict. I will need to look into this further. I'm not quite sure why this only affects the pyqt4 bindings. Presumable qt.cpp is not using the typeinfo information. Andrew -- Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge ___ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
Re: [Plplot-devel] QT rpath bug
On Mon, Jul 20, 2009 at 11:28:31AM -0700, Alan Irwin wrote: On 2009-07-20 15:59+0100 Andrew Ross wrote: Alan, When implementing rpath support for the libplplotqtd library you have added the qt path in front of the rest of the rpath directories. From the comments you added I think this is to ensure that you pick up your custom built qt libraries rather than the system ones. Unfortunately if you are using the system qt libraries this adds /usr/lib at the front of the rpath directory list and therefore you end up linking with system version of the plplot libraries rather than those in the install location. For me this is a headache. Of course the proper solution is to order the rpath variables so that the install prefix always appears first, but I can't see an easy way of doing that with your current approach. Any thoughts? Leave this with me for now. I agree the described behaviour would be most annoying, and I will try to figure out a fix. Thanks Andrew -- Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge ___ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
Re: [Plplot-devel] RE : Removed dynamic casts from Qt driver
On 2009-07-20 16:46+0100 Rochel, Alban wrote: Hello Andrew, Here is a version of the same patch based on R10166. There were indeed conflicts with your changes. There should be about 5 versions for the callbacks: - rasterqt (for png, bmp...), sometimes specialised as pngqt, bmpqt... where needed - svgqt - epspdfqt (sometimes epsqt and pdfqt) - qtwidget - extqt This patch only makes the code 20 per cent larger, and I think that is a completely acceptable price to pay for peace of mind about issues with how dynamic casting is implemented on Linux. Furthermore, qt.so built without issues and make test_noninteractive in the installed examples tree generated epsqt, pdfqt, bmpqt, jpgqt, pngqt, ppmqt, tiffqt, and svgqt results without issues. Also, c++/qt_example (which exercises extqt) worked fine. Therefore, I have committed this as revision 10168. Thanks, Alban! Alan __ Alan W. Irwin Astronomical research affiliation with Department of Physics and Astronomy, University of Victoria (astrowww.phys.uvic.ca). Programming affiliations with the FreeEOS equation-of-state implementation for stellar interiors (freeeos.sf.net); PLplot scientific plotting software package (plplot.org); the libLASi project (unifont.org/lasi); the Loads of Linux Links project (loll.sf.net); and the Linux Brochure Project (lbproject.sf.net). __ Linux-powered Science __ -- Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge ___ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
[Plplot-devel] [ plplot-Feature Requests-2824647 ] Qt designer plugin for plplot widget
Feature Requests item #2824647, was opened at 2009-07-21 04:40 Message generated for change (Tracker Item Submitted) made by nobody You can respond by visiting: https://sourceforge.net/tracker/?func=detailatid=352915aid=2824647group_id=2915 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: None Group: None Status: Open Priority: 5 Private: No Submitted By: Nobody/Anonymous (nobody) Assigned to: Nobody/Anonymous (nobody) Summary: Qt designer plugin for plplot widget Initial Comment: Create a plplot Qt Designer plugin. Attached, please find the a file with the designer.pro, qtplplotplugin.h, qtplplotplugin.cpp and Makefiles. Also, I have include the release binaries. You will need to make sure plplotd.dll, qsastime.dll and qt.dll (this is the Plplot Qt driver are in your path). I simply copied these to $QTDIR\bin. Then place qtplplotplugin.dll, qtplplotplugin.lib in $QTDIR\plugins\designer. The module is loaded under Display Widgets with name QtPLWidget -- You can respond by visiting: https://sourceforge.net/tracker/?func=detailatid=352915aid=2824647group_id=2915 -- Enter the BlackBerry Developer Challenge This is your chance to win up to $100,000 in prizes! For a limited time, vendors submitting new applications to BlackBerry App World(TM) will have the opportunity to enter the BlackBerry Developer Challenge. See full prize details at: http://p.sf.net/sfu/Challenge ___ Plplot-devel mailing list Plplot-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/plplot-devel
