Hopefully fix "invalid state when picking" warnings. #Hackfest2016

fixes a few of jenkins warnings

More GObject boilerplate reduction

Fix build (not pretty).

Fix bounding box cache issues in general, and prevent the selector tool from updating anything in case of identity affines (which prevents the bounding box from being invalidated)

DrawingContext: change variable names ct to dc (bug #1272073)

UI. Fix for Bug #1072007 (Mouse scroll zoom depends on if the cursor is over an object).

Fix for 657463 : Mousewheel zooming by two steps, not one

Stop deriving SPCanvasItem from GtkObject

Revert r11370. Fix crash with grid in default template

Drop some GtkObject usage

Replace deprecated GTK_TYPE macros

Initial C++ification of SPCanvas.

More GSEAL issues

Cleaned up display-forward.h, including many redundant usages.

Update 2Geom. Remove all use of NRRectL.

Reduce default rendering cache size to 64 MiB

Add user preference for rendering cache size

Initial autocache work

Rewrite NRArena -> Inkscape::Drawing. Call render and update methods on the Drawing rather than on the root DrawingItem.

Rewrite NRArenaItem hierarchy into C++

Per-item render cache. Cache some offscreen data to facilitate smoother navigation.

Replace direct use of Cairo contexts and surfaces in the rendering tree with wrappers which keep some extra information about the surface, amd NRRect and NRRectL use with Geom::Rect and Geom::IntRect. Should simplify implementing filter primitive subregions.

Clean up some commented-out code

Add SPCanvasArena caching layer. Currently breaks for clipped groups that contain filtered objects (Cairo clipping bug?)

GTK+ cleanup: gtk_type_class

Switch to GObject

Replace deprecated GtkSignal

update to latest 2geom !

Split SPCanvasItem and SPCanvasGroup to individual .h files. Removed forward header.

Super duper mega (fun!) commit: replaced encoding=utf-8 with fileencoding=utf-8 in all 1074 Vim modelines. The reason for this is that (a) setting the encoding isn't nice, and (b) Vim 7.3 (with modeline enabled) disallows it and pops up an error whenever you open any file with it ("invalid modeline"). Also corrected five deviant modestrings: * src/ui/widget/dock.cpp and src/ui/widget/dock.h: missing colon at the end * src/ui/dialog/tile.cpp: removed gratuitous second colon at the end * src/helper/units-test.h: removed gratuitous space before a colon * share/extensions/export_gimp_palette.py: missing textwidth=99 That's my geekiest commit yet.

Make nr_arena_invoke_render expect an already transformed context, to remove a limitation to integer translations imposed by NRRectL

Completely remove NRPixBlock

Wholesale cruft removal part 1

Remove some cruft

Consolidate Cairo utils in display/cairo-utils.h. Fix icons harder.

Initial Cairo rendering commit: solid shapes, gradients, opacity and patterns

Improved version reporting. Add SVN revision and custom status to version string on Linux.

Cmake: Corrections for mistakes

Next roud of NR ==> Geom conversion

whitespace

no more creating and looping over extra buffers for rendering canvas arena, draw to the canvas buf directly now that it's also 32bpp

Remove warnings

lay groundwork for non-filter view mode

Warning cleanup (including OFFSET_OF warning)

Warning cleanup

noop: CodingStyle: re-indent a few files that had mixtures of spaces & tabs for indentation.

make arena remember its parent canvasarena, if any (so that an arenaitem can get access to the canvas viewbox for picking optimization)

blind fix for endianness, needs testing by someone on a big-endian machine

create/destroy the cairo_t for all items to render into (currently used only in outline mode)

API change: render methods now take a cairo_t (not yet used)

more comments

more comments on cairo's and our own woes

a bunch of cairo fixmes in anticipation of a more direct rendering with less intermediate buffers and unpacked 32bit RGBA throughout

purge fred's arena item tile cache (has been disabled for quite some time anyway, bitrotting; enabling it gives no measureable advantage)

deal with the possible failure to create pixblock due to insufficient memory

peeled back the gboolean code as it hit on some complexity theory principles... need to rethink and incrementally change gbooleans to bools

Ok, committed msgloan's patch to convert gbooleans to bools thus completing one major janitorial task we identified....

Rendering optimisation, which gives best results for zoomed in drawings with complex (multi-node) paths. Optimisation focus - eliminating cubicTo and LineTo nodes outside of visible area, so no unneeded calculations is needed and less memory is consumed (this avoids crashes which were occurring previously if zooming into a complex drawing)

moving trunk for module inkscape