GitLab

This replaces the raw JSON data and shows it nicely structured tree
view, making it easier to navigate.

Model taken from QJsonViewer
(https://git.simply-life.net/simply-life.net/qjsonviewer/) with some
small improvements/changes.

- Split off some functions
- Render upside down, then flip vertically in the end, to get all items
positioned correctly
- Rename some variables/members

- addCacheableShaderFromSourceCode() wasn't added until Qt 5.9
- Flip the rendering vertically directly, so mirrorVertically isn't
required (Qt 5.6+)

- The DGRP decoder now has friendlier enum/flag-based types for some
fields

- The decoder exposes some limited internal structure (frames) through
its API

- The DGRP decoder no longer composes the files to images, and therefore
doesn't derive from the MultiImageChunkDecoder anymore. Instead
rendering is expected to be done by the viewer.

- The view now uses OpenGL to render the draw group

- The renderer could still do with a little cleanup (better naming and
such), but other than that it does seem to work well

This started throwing errors...

- Pixels without alpha channel data should be fully opaque or
transparent (depending on palette color)
- Pixels with alpha data: the alpha channel is scaled from 0-31, so
multiplication by 8 is required to get the proper 8 bit value
- Write directly into data of QImages, using Format_Grayscale8 for Z and
Format_Alpha8 (alpha) formats instead of palettes. This is semantically
more correct.

Also use configuration to determine version of plugin to deploy (*.dll
or *d.dll for debug)

Deploying the correct version should make sure the plugins are always
loadable, regardless of whether it's a debug or release build.

These were not installed, causing the split build to fail

Qt doesn't need QMetaEnum include to access it in recent versions (QIcon
pulls it in, so it seems). However, Qt 5.5 still needs it more
explicitly.

Also make sure dbpfm depends on the correct minimum Qt version, also
during the split build.

- Use the correct (renamed) QRC file
- Correctly map the URL to the QRC filename or local filename

Also keep the default icon's filename consistent, for less confusion

Plain Windows BMP (source: http://wiki.niotso.org/IFF#FBMP)

- New DBPFM plugin type (TypeInfoFactory) to provide the necessary
information (includes necessary CMake stuff)
- The TypeInfoImageProvider uses these plugins to generate pixmaps from
provided icons
- New proxy model to add decorations and manipulate Display of types in
DBPFM
- Initial plugin for Sims 1 type info

Adds icons from KDE's Breeze set as fallbacks, but attempts to use
system defaults first (only on Linux).

All in all this adds a touch of user-friendliness, as it's become much
easier to identify types (now by names, rather than hex codes)

Minor required changes:
- Q_ENUM() for the container enums (requires Qt 5.5+)
- Bump copyright year in Readme

The URL was changed at some point

This solves the problem of having 2 "Images" tabs.

Add a unit test for this too. Basically we weren't returning the proper
parent modelIndexes, but some meaningless stuff (the child row pointing
to the parent item, never an invalid QModelIndex() for root items, etc)

This hopefully fixes build issues on the CI build because the different
types of QString and char[]

This was introduced in Qt since the last time I did anything on this
area. The TableToTreeProxyModel groups items in the base table model by
keys etc.

It's not yet fully functional: adding/removing chunks or changing their
data will not properly propagate through to the proxy yet. But: this is
a start and make the tree a lot more usable.

Additional changes:
- the old table still lives as DbpfTable
- the main view starts up with a DbpfTable as tab
- the base container model now always returns formatted text when
accessing data based on role, similar to the DisplayRole. This is more
consistent. Raw data may be added again in the future if it's necessary.

The chunk's offsets need to be moved to after the chunk prefix. This
solves crashes when using Save As for an IFF file

- Make sure overrideLineHeight is set to < 0, so automatic height is
used properly
- Make sure lineHeight change is emitted when changing mode
- Make sure font changes are processed initially (when cellSize is still
empty)

Main window geometry and file dialog paths

It's outside the (transparent) ScrollView so it provides the correct
backdrop without requiring any updates/repainting/whatsoever.

Mostly affects selection

Better integration into the host platform, i.e. a more portable
solution.

The only exception so far is Ctrl+Up/Down to scroll the view without
moving the cursor; there are no standard sequences for this. Also,
begining/end of "block" (current view) movement does not exist on
anything but Mac, so also uses custom keys too.

Also a minor addition: Cancel (Escape key) now also deselects

(Key) events are accepted by default - so any key press would basically
reset the selection

I.e. much like clicking the very left of the hex view. The first click
doesn't yet select the line, dragging does selection. A simple click
simply moves the cursor to the start of the given line.

This is consistent with at least Kate behavior in KDE; possibly the same
for other editors too.

But only if that or ASCII view doesn't already have the active focus

The Hex->Ascii bindings are now completely done through property
bindings again (they aren't broken through JS assignments anywhere). For
Ascii->Hex, also add Binding for the selection, so clicking the gutter
doesn't break the property binding

I.e. when moving the cursor from after to before the start selection
(skipping the start character itself), the previous behavior was
inconsistent (it would or wouldn't select the starting byte sometimes).

Now the data is in character positions and normalization is done for the
public API. This way the selection can also be properly determined to
extract the selected data (still todo).