GitLab

icon.name() is too new

CCBUG:327156

BUG:32715

10:20:21 < dmitryK> boud: yes, without it the lines may become dizzy in somecircumstances
10:23:29 < dmitryK> boud: you can comment it out temporarily
10:23:30 < dmitryK> boud: but we must not forget about it, because the results will be not pixel-correct
10:23:30 < dmitryK> boud: i mean subpixel precision will be broken for the presets which cause this assert
10:23:45 < dmitryK> boud: you can comment it, but add a note to the bugreport so that we would not close it ;)

The GTK file dialog integration is broken in so many interesting ways
that we'd better disable it completely.

This reverts commit d441a1e4748d799afc6174860b77a29d0fa5b553.

Okay, disabling the glib eventloop is still _very_ desirable,
Animtim reports a huge performance loss following the patch
enabling it everywhere.

BUG:327128

It tested the Qt tablet handling, so isn't relevant for us anymore

Conflicts:
	krita/ui/opengl/kis_opengl_canvas2.cpp

This reverts commit 2f1a79296ec058690ad0ff1796098c7397a169a3.

This reverts commit 806df3c07e80cd99236c496d51cf3650b722b841.

The time measurements of the tablet events are really unstable,
especially when using openGL canvas, which causes the Weighted Smoothing
processed lines bended and waggy. This patch makes the algorithm use
distance parameter instead of the velocity parameter, which does not
depend on time, so does not fluctuate so drastically.

BUG:326512

REVIEW: 113261

thanks boemann & boud for review

Create plugin system for semantic items, turn hardcoded contact, event & location items into plugins

REVIEW: 113588

thanks boemann & boud for review

This patch fixes veeery slow painting with the tiny brushes over
huge canvas (7px brush over 7000px canvas).

This happened due to too many mipmap regenerations happening, so now this
is done in the paintGL function instead.

We don't have 3D objects so we don't need GL_MULTISAMPLE

This causes bended lines when the events are coming not-gradually.

The compression is done in two stages:
1) The signals are compressed using KisSignalCompressor with a period
   of 10 ms. This effectively limits the frame rate by 100fps.
2) On some (slow) systems 100 fps might be too much, so before delivering
   the signal to the openGL canvas it is now checked whether the pipeline
   has already finished processing our previous update requests using
   KisOpenGLCanvas2::isBusy(). If the pipeline is still congested, we just
   resend the signal back to KisSignalCompressor.

Screenshots are to follow.
CCMAIL:hughsient@gmail.com

This reverts commit 69f3b5fd0bea83ea634b413887512c0f46bb2437.

The throttling will be reimplemented using KisSignalCompressor object

"Dummy profile" is very descriptive, but if we map a request for a
colorspace with an empty profile name to a colorspace with a profile
named "Dummy profile" the mapping function gets all confused.

If the dox say that the gradient is owned by KoGradientBackground,
then let's not delete the gradient we just passed it.

This also disables the copy and operator= constructors for all shape
background classes, for safety's sake.

BUG: 779777
BUG: 326902