Re: [Plplot-devel] Qt5 suport, strange driver output

[Alan W. I. <ir...@be...>]

On 2014-07-21 13:16+0100 António Rodrigues Tomé wrote:

> sorry  opensuse 13.1 not 13.2
>
>
>
> On Mon, Jul 21, 2014 at 1:14 PM, António Rodrigues Tomé <ar...@gm...>
> wrote:
>
>> Hi Alan,
>> thanks for the quick reply.
>> I'm using opensuse 13.2 x86_64 system.
>> my qt is 5.3.1  binary packages provided by the opensuse repository.
>>
>> I built the plplot with the command
>>  atome@linux-9gx6:~/StatisticalForecast/bibucket/plplot/plplot-code-13137-trunk/built-dir>
>> cmake -DDEFAULT_NO_CAIRO_DEVICES=ON -DENABLE_DYNDRIVERS=OFF
>> -DDEFAULT_NO_BINDINGS=ON -DENABLE_cxx=ON -DENABLE_qt=ON -DPLPLOT_USE_QT5=ON
>> -DCMAKE_INCLUDE_PATH=/usr/include/qt5/
>> -DCMAKE_INSTALL_PREFIX=/home/atome/StatisticalForecast/bibucket/plplot/plplotLibs/
>> ../ >& cmake.out

Hi António;

Thanks for the above information.  In general I would advise adopting
-DENABLE_DYNDRIVERS=ON (or dropping -DENABLE_DYNDRIVERS=OFF since ON
is the default) since that build is much less tricky than the
-DENABLE_DYNDRIVERS=OFF case.  Also, if you want to do build tree
tests like the test_all_qt target I suggest below, then you should
also use the cmake option -DBUILD_TEST=ON.

SuSe has a long historical record of distribution expertise concerning
KDE and Qt.  So perhaps they (or the upstream Qt developers) have
already fixed in Qt5.3.1 the character offset issue I empirically
corrected for the Qt-5.2.1 case.  What happens in the pdf case when
you remove that PLplot "correction"?  (Look for empirical_yOffset in
bindings/qt_gui/plqt.cpp in the source tree and set it to 0.)

Note, to obtain complete tests of qt results, try the "test_all_qt"
target, e.g., run

make -j4 test_all_qt

right after you run the cmake command (with -DBUILD_TEST=ON).  You
should expect interactive test results to flash momentarily on your
screen as well as a large accumulation (something like 2GB!) of plot
file results generated by all the qt devices collected in
examples/test_examples_output_dir.  Then to see what the results are
like in a wholesale way you would want to do something like this

cd examples/test_examples_output_dir
for FILE in x*.pdfqt; do echo $FILE; display $FILE; done

where display in the imagemagick command that allows you to view most
file formats.  And similarly for x*bmpqt and the rest of the qt file
device results.  Also, if you want to see the interactive results a
lot slower (to allow you to carefully evaluate the rendering), then
run

examples/c/x00c -dev qtwidget

etc., for all our standard examples after the above test_all_qt target is
run (which should build all required dependencies for the suggested
-dev qtwidget tests).

If zeroing empirical_yOffset produces perfect pdf results (as
compared, say, to the perfect cairo results for all our examples
displayed on our website) for you, then how many other qt devices also
produce exact rendering agreement with those pdf results and how many
qt devices (as you have already shown for -dev pngqt) still have
alignment or font issues?

Note, we found lots of Qt4 rendering and font bugs with all but the
later (i.e., x >= 5) versions of Qt4.x. So I expect it is similarly
going to take several more "minor" releases of Qt5 before all the
character alignment and font issues that affect the PLplot qt devices
are fixed.  So for now, our qt device users who want absolute
reliability should probably stick with Qt4, but it would still be good
if you are willing to do the testing requested above just to see where
we stand with the rendering quality of Qt5. Note also that I want to
be quite selective of what Qt5 versions I build since that is a long
process, but if you report rendering perfection for at least one qt
device (e.g., -dev pdfqt) for all our standard examples with the
opensuse 5.3.1 version of Qt, than I would be willing to try to build
the upstream 5.3.1 to see whether I get similar results for upstream
or whether the rendering fixes relative to 5.2.1 are only in the
opensuse version.

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); the Time
Ephemerides project (timeephem.sf.net); PLplot scientific plotting
software package (plplot.sf.net); 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
__________________________