GitLab

Instead of adding a rect to a region for every dab, we coalesce the
footprints into a big rect. This makes painting faster, but at the end
of the stroke we do much more work since we bitBlt the whole united
rect to the target layer, instead of just the rects the stroke touched.

Note: we might be able to remove the mutex, too.

The reason is that add_definition inteprets escape sequecences differently in newer version.

It was the tests that were wrong, they depended on a faulty behaviour.

One day we'll be able to use a document without its attendant part,
but that day is not now. This should fix quite a number of tests and
bugs.

BUG:305371

Note: we probably are leaking the part object, so I need to check that.

GIT_SILENT

Boud, can you make sure it's correct?

CCMAIL: boud@valdyas.org

got hint from http://www.thelins.se/johan/blog/2011/12/qbound-on-arm/

The left/right/top/bottom variants should override the
fo:margin/fo:padding shorthands.

This fix does that the old code tried to do but failed: Check if the
given path is a directory. But it does so using QDir so it's actually
portable.

BUG: 305794
FIXED-IN: 2.5.1
REVIEW: 106200

This reverts commit 20f3a3201ca2c68819bb2cd7d7365485013d2685.

This reverts commit abba945609f2e1d0138bd8b1a642c03df5bf231f.

==31293== 1,708 (136 direct, 1,572 indirect) bytes in 1 blocks are definitely lost in loss record 2,039 of 2,111
==31293==    at 0x4C2B1C7: operator new(unsigned long) (in /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
==31293==    by 0x6B512A1: QObject::QObject(QObject*) (qobject.cpp:721)
==31293==    by 0x8767D6D: KoCharacterStyle::KoCharacterStyle(QObject*) (KoCharacterStyle.cpp:343)
==31293==    by 0x8773CDF: KoParagraphStyle::KoParagraphStyle(QObject*) (KoParagraphStyle.cpp:97)
==31293==    by 0x7F6E40C: KoTextShapeData::loadStyle(KoXmlElement const&, KoShapeLoadingContext&) (KoTextShapeData.cpp:331)
==31293==    by 0x8B8EDDD: KoTosContainer::loadText(KoXmlElement const&, KoShapeLoadingContext&) (KoTosContainer.cpp:88)
==31293==    by 0x22E762B6: EnhancedPathShape::loadOdf(KoXmlElement const&, KoShapeLoadingContext&) (EnhancedPathShape.cpp:585)
==31293==    by 0x8B3C5D0: KoShapeFactoryBase::createShapeFromOdf(KoXmlElement const&, KoShapeLoadingContext&) (KoShapeFactoryBase.cpp:228)
==31293==    by 0x8B38735: KoShapeRegistry::Private::createShapeInternal(KoXmlElement const&, KoShapeLoadingContext&, KoXmlElement const&) const (KoShapeRegistry.cpp:313)
==31293==    by 0x8B382BB: KoShapeRegistry::createShapeFromOdf(KoXmlElement const&, KoShapeLoadingContext&) const (KoShapeRegistry.cpp:268)
==31293==    by 0x1D047D12: KoPAPageBase::loadOdf(KoXmlElement const&, KoShapeLoadingContext&) (KoPAPageBase.cpp:212)
==31293==    by 0x1CD5DBEA: KPrPage::loadOdf(KoXmlElement const&, KoShapeLoadingContext&) (KPrPage.cpp:126)

here the default style is not freed as the paragraph styles.

REVIEW: 106154

REVIEW: 106168

thanks boemann for review

i) Axis commands
ii) Legend commands
iii) Dataset commands
iv) Chart type command
v) Textshape command

as discussed on ml calligra-devel,
see thread "[REVIEW REQUEST] Cleanup of unused or duplicated icons"
http://mail.kde.org/pipermail/calligra-devel/2012-August/006442.html

Okayed by Boemann and Thomas Pfeiffer

If the engine can handle HDR internally, then we need to mark the
conversion vertex as conserving the dynamic range, otherwise we'll
never find a non-lossy way between, say F16 and F16.

This reverts commit 5bd55414a8a1ebd5a1a9e2da10868c90bf32bc33.

But only if there aren't any compositions. Krita 2.4 can only load
documents where there is nothing else but layers in the image; 2.5
adds compositions and master adds assistants.

So, if there are no compositions, don't save an empty composition tag,
then 2.4 can still load the document. If there are compositions or
assistants, it's tant-pis...