[Kde-bindings] QtRuby: Qt::Variant#value fails for ushort and QChar variants

[Yaohan Chen]

Thanks for the quick response. I wonder if the workarounds fail because of
endian/size difference on your machine. I did try with "native endian" (_
instead of > at the end of unpack argument), but that makes them fail for me.
Anyway I wrote them only based on experiment and I haven't found more detailed
specification of QDataStream format. It's great that the workarounds will
be no longer necessary!

Yaohan Chen


On Wednesday, April 11, 2012 01:40:48 PM Richard Dale wrote:
> On Wednesday, April 11, 2012 03:11:33 AM Yaohan Chen wrote:
> > Hello,
> > 
> > I'm getting a segfault when calling Qt::Variant#value for QVariant<ushort>,
> > and NoMethodError for QVariant<QChar>. I've included test cases that show
> > the failures as well as workaround using String#unpack below. They require
> > RSpec, RubyInline and of course QtRuby to run. Could anyone confirm if you
> > have the same problem?
> > 
> > Thanks,
> > 
> > Yaohan Chen
> > 
> > 
> > #!/usr/bin/env rspec
> > 
> > require 'Qt'
> > require 'inline'
> > 
> > # These tests demonstrate that when you get a Qt::Variant by unserializing a
> > serialized # QVariant via QDataStream, the value method fails if the
> > variant's type is QChar or ushort. describe 'Qt::Variant#value' do
> >   it 'returns the value of an int variant' do
> >     variant = unserialize_variant SerializedVariant.of_int(15)
> >     # This one works for me
> >     variant.value.should == 15
> >   end
> > 
> >   it 'returns the value of a QChar variant' do
> >     variant = unserialize_variant SerializedVariant.of_QChar(10)
> >     # This fails for me saying the QVariant does not have method toChar
> > defined variant.value.should == 10
> >   end
> > 
> >   it 'returns the value of a ushort variant' do
> >     variant = unserialize_variant SerializedVariant.of_ushort(5)
> > 
> >     # pending 'causes segfault'
> > 
> >     # This causes Ruby to segfault for me. If it happens for you too,
> > uncomment the pending # line above so RSpec can report the results of other
> > examples. variant.value.should == 5
> >   end
> > end
> > 
> > # These extract Ruby values directly from serialized variants without Qt, to
> > show that the # correct data is passed from C++ to Ruby. They use
> > String#unpack and skip the first # 5 bytes ('@5') which are QVariant's type
> > information and "null flag". See #
> > https://qt-project.org/doc/qt-4.8/datastreamformat.html
> > describe 'Workaround using String#unpack' do
> >   it 'returns the value of an int variant' do
> >     string = SerializedVariant.of_int(15);
> >     # l> for big-endian 32bit signed integer
> >     string.unpack('@5l>').first.should == 15
> >   end
> > 
> >   it 'returns the value of a QChar variant' do
> >     string = SerializedVariant.of_QChar(10);
> >     # s> for big-endian 16bit signed integer (QChar stores a UTF-16
> > character) string.unpack('@5s>').first.should == 10
> >   end
> > 
> >   it 'returns the value of a ushort variant' do
> >     string = SerializedVariant.of_ushort(5);
> >     # S> for big-endian 16bit unsigned integer
> >     string.unpack('@5S>').first.should == 5
> >   end
> > end
> > 
> > # Given string containing serialization of a variant, return the variant
> > def unserialize_variant(string)
> >   byte_array = Qt::ByteArray.new string
> >   data_stream = Qt::DataStream.new byte_array
> >   variant = Qt::Variant.new
> >   data_stream >> variant
> >   variant
> > end
> > 
> > # This module contains functions that return serialization of several
> > variant types module SerializedVariant
> >   inline :C do |builder|
> >     builder.add_compile_flags %w[-x c++]
> >     builder.include '<QtCore/QByteArray>'
> >     builder.include '<QtCore/QDataStream>'
> >     builder.include '<QtCore/QVariant>'
> >     builder.add_link_flags %w[-lQtCore]
> > 
> >     builder.prefix <<-CODE
> >       // Return serialization of the given QVariant
> >       VALUE serialize(QVariant variant) {
> >         QByteArray byte_array;
> >         QDataStream data_stream(&byte_array, QIODevice::WriteOnly);
> >         data_stream << variant;
> >         return rb_str_new(byte_array.data(), byte_array.size());
> >       }
> >     CODE
> > 
> >     # :expand_types => true makes RubyInline insert VALUE self and figure
> > arity automatically builder.c_raw_singleton <<-CODE, :expand_types => true
> >       // Return serialization of a QVariant containing the given int
> >       VALUE of_int(VALUE v) {
> >         QVariant variant = QVariant::fromValue<int>(FIX2INT(v));
> >         return serialize(variant);
> >       }
> >     CODE
> > 
> >     builder.c_raw_singleton <<-CODE, :expand_types => true
> >       // Return serialization of a QVariant containing the given QChar
> >       VALUE of_QChar(VALUE v) {
> >         QVariant variant = QVariant::fromValue<QChar>(FIX2INT(v));
> >         return serialize(variant);
> >       }
> >     CODE
> > 
> >     builder.c_raw_singleton <<-CODE, :expand_types => true
> >       // Return serialization of a QVariant containing the given ushort
> >       VALUE of_ushort(VALUE v) {
> >         QVariant variant = QVariant::fromValue<ushort>(FIX2INT(v));
> >         return serialize(variant);
> >       }
> >     CODE
> >   end
> > end
> 
> Thanks for the bug report and the very interesting test program. I installed 
> RubyInline to get it working and I'm really impressed how you can combine C 
> code and ruby in the same source. QtRuby has never really had any test code 
> and I think for the next version 3.0 it would be nice to do some tests using 
> RSpec/RubyInline like you've done.
> 
> I've fixed the bug in Qt::Variant#value in the master branch. I needed to 
> change the QChar comparison to:
> 
>   it 'returns the value of a QChar variant' do
>     variant = unserialize_variant SerializedVariant.of_QChar(10)
>     # This fails for me saying the QVariant does not have method toChar defined
>     variant.value.should == Qt::Char.new(10)
>   end
> 
> The three workround tests don't work for me though and they still fail, and I 
> don't know if that is a problem with QtRuby or the test code.
> 
> -- Richard
> -- Richard
> 
> 
> 
> _______________________________________________
> Kde-bindings mailing list
> Kde-bindings at kde.org
> https://mail.kde.org/mailman/listinfo/kde-bindings