Yes as i recall that code was made to handle masks being exported correctly for OpenOffice or LO - I'd strongly prefer it not being thrown away like that

When i investigate to find out why it's saved square not rectangle i suspect that code, but i'm still not absolutely sure that it's wrong. Maybe @boemann knows better. If so can we remove m_viewBound as well?

Looks good to me +1

• If no textshape is editied, select the first found text shape

Arghh, accidentally propt a new diff on top of this, howto undo this?
(arc is fine, but do not do anything wrong)

• DefaultTool: Do not considder parent flags when checking Move-/ResizeAllowd
• DefaultTool: MoveStrategy: Check if shape is movable
• DefaultTool: ResizeStrategy: Check if shape is resizeable
• DefaultTool: ShareStrategy: Check if shape is shearable
• DefaultTool: RotateStrategy: Check if shape rotation is allowed
• Fix undo/redo of anchored shapes

Let's make some fix about that.

No, it should not be possible to activate a tool without a shape.

No, more sinister things are going on...
Adding a comment to an annotation shape should not be possible, so changes to the text tool is required.
Patch on it's way.

Thank you all for the review.

Fine with me.

I'm OK with it +1, @danders or @boemann should accept it.

@dcaliste @anthonyfieroni Any updates here, pro/con? Would be a terrible shame to let this fall through :)

Also can you add parentheses over single line code and if you know better way to get impl() without duplicate calls will be great :)

Thank you @leinir for your review. I've not disappeared, but I'm still thinking about a way not to expose the impl object from Document class, while not cluttering the Document API with methods like setBackgroundColor(), setPageCache() or any other future canvas methods that I may think about.

Looks pretty good to me. It'd be nice if we could avoid exposing DocumentImpl (it's supposed to be sort of hidden), if you could have a think on a way to avoid that, i'd appreciate it :)

Ahh well in that case I don't mind and the words code looks clean enough - I'll let someone else review the components

Sorry, I was not clear in my description, I'm not setting the paper background color, but the item background color, i.e. the small area around and between the pages. In most desktop application it is light grey. But when using the KWCanvasItem in another context, it would be nice to be able to set this color. It's kind of desktop color if you prefer.

'm sorry but the background should not be settable - it's a document property at best and in fact paper is mostly white, so shouldn't even be settable

Yes please, categorification and less dependency on the support library is definitely a good thing :)

Maybe it would be nice to replace the Q_ASSERT with an if (factory) {} and put a warning in the else part in case of a wrong installation and plugins are missing. What do you think ?

This is looking pretty good to me, really... i know there was talk about deprecating qreal at some point, but turns out that never happened?

Update the patch, using categories for logging.

@anthonyfieroni I see, like in filters/sheets/html for instance. I'm going to update the diff as soon as ready. Thanks for the comment.

This is old practice, please make a logging category -> https://doc.qt.io/qt-5/qloggingcategory.html

It was merged.

Any other ideas for special handling of headerRows == 0?

Yes please, thank you! That has frankly been bugging me for a while... We're sure there's no side effects here? But, yeah, having two menus pop up underneath each other is sort of annoying ;)

After reading the fine print, see: http://docs.oasis-open.org/office/v1.2/os/OpenDocument-v1.2-os-part1.html#__RefHeading__1420028_253892949
it seems to me a shape can be anchored to a page in different ways. Note this:
"If the attribute value is not from-left and not from-inside, the svg:x attribute associated with the frame element is ignored for text documents."
This would explain that LO seems to discard x,y when anchor info is not saved.
Afaics there is no defined default values for anchor values.

Bah I can't remember this, but page anchored shapes don't have anchors as such, so I'm wondering what exactly is this writing to file?

Another try:
Don't mark table as misfit when it does not have header rows

Good job!

It can be allocated only in read_tbl but as you can see in bug report, use-after-free is very common to happen.

Thanks, lots of other people also touched this code but I tried this time to review :) I propose to separate the string literal changes.

No idea either - jaroslaw was the original author so maybe he can take a look

Sorry, don't have any qualified opinons, I don't quite understand the original code either, so...
Imo if you think it is good, land it.

Ping on this, did you try it?

Something as

if (d->headerRows) {
    cursor->row = 0;
}

It does not work, the problem is that d->headerRows is 0

I am more thinking of this place:

In D15428#324451, @boemann wrote:

One reason why this might fail is if virginpage becomes false when we add the table headerrow (if we can fix this and that doesn't have other ill effect then I would prefer such a solution).
So find out where we set virginpage to false, and make sure it doesn't go from true to false when adding the header row

I hope this makes sense

The idea is that if we are at the beginning of a new page we don't get in here (virginpage is true)
so we only reset if we are somewhere down on a page meaning we wil have more space to try on next time around
But we should reset if the only thing we fitted was the table header as a table headerrow should never be the only thing on a page

Initialize labelYOffset even layout line is not valid

Hmm, imo styling odf apps content area should be considdered a bug :=)
Because: All info pertaining to presentation of content is stored in odf file and is controllable by user, so styling it would often create strange resullts. (And it would not be much wysiwyg about it)
Of course everything else like tools, dialogs, window decorations can be styled.

Aggressive refactor, it makes m_table a shared pointer, presetTables is one time initialized.

It's inspired in forum discussion, it has predominately advice that other Calligra apps has a white background despite palette but sheets.

No Idea about the code, but in general I have no problem with sheets respecting the palette