crash bug Bug 342185

[Boudewijn Rempt]

Hi Alex,

I've cc'ed calligra-devel, since this problem is deep in the shared kotext 
code, and I think that only the words development lead, Camillab knows 
anything about it! I wouldn't be able to guess the correct solution...

On Fri, 30 Jan 2015, Alex Demko wrote:

> Although the submitter reported this bug resolved, I think it's still
> open, as I'm able to reliably reproduce it in the most recent
> calligra/2.9 branch. If I increase Drop Caps->Number It Covers to say
> 4, and then keep decrementing Indents/Spacing->Line Spaces->Fixes
> lower and lower, around 9 or so I can get it to crash. I assume it's
> all dependant on the font any ways.
>
> The bug triggers the assert in:
>
> void KoTextLayoutArea::setBottom(qreal bottom)
> {
>    d->boundingRect.setBottom(bottom + qMax(qreal(0.0),
> d->verticalAlignOffset));
>    Q_ASSERT_X(d->boundingRect.top() <= d->boundingRect.bottom(),
> __FUNCTION__, "Bounding-rect is not normalized");
>    d->bottom = bottom;
> }
>
> Which is understandable because bottom in my case is -9.57456e-13, aka
> -tiny. With top()==0 and bottom()==-tiny, it's not normalized. How did
> it become -tiny and not real 0? It seems in this bit of code:
>
> libs/textlayout/KoTextLayoutArea.cpp:1295:
>
>        d->neededWidth = qMax(d->neededWidth, line.naturalTextWidth()
> + d->indent);
>
>        if (!runAroundHelper.stayOnBaseline() &&
> !(block.blockFormat().hasProperty(KoParagraphStyle::HiddenByTable)
>         && block.length() <= 1)) {
> // d-> is -6 and maxLineHeight is 6
>            d->y += maxLineHeight;  // <---- the culprit
> // d->y is now -tiny
>            maxLineHeight = 0;
>            d->indent = 0;
>            d->extraTextIndent = 0;
>            ++numBaselineShifts;
>        }
>
>
> It seems -6 + 6 gives us -tiny, which is understandable as this all in
> real numbers and round off errors are expected.
>
> Now, this is inside KoTextLayoutArea::layoutBlock(), which is quite
> the monster, so it might be easier for devs familiar with this code to
> handle it. Alternatively, I can fix it, but I'd need to know the
> approach... Can d->y be reset to its entry value? Can it be reset to
> 0? etc.
>
> Thanks
> _______________________________________________
> Krita mailing list
> kimageshop at kde.org
> https://mail.kde.org/mailman/listinfo/kimageshop
>