Details:
- https://sourceforge.net/p/irrlicht/bugs/433/
- https://github.com/minetest/minetest/issues/1687#issuecomment-61368769
- https://forum.minetest.net/viewtopic.php?f=42&t=9190&start=125#p159364
In case when "settings.h" is included from "emerge.cpp" or
"environment.cpp", u64 type has "unsigned long" length because
previously <stdint> was included. When "settings.h" is included from
"settings.cpp", u64 has "unsigned long long" length because no <stdint>
was included previously. This leads to different signatures of "setU64" method
and linker cannot find appropriate symbol.
The best fix of this bug should be done in the Irrlicht, but as hotfix I
think this is OK and better than types changing.
Previously this bug didn't appear because there was no "settings.cpp" file and
implementation of all methods was done in the header file.
Add 'absolute value' option to noise map functions
Extend persistence modulation to 3D noise
Extend 'eased' option to noise2d_perlin* functions
Some noise.cpp formatting fixups
NOTE 1: This does not mean a 2x increase in framerate. Increase in fps may be up to 1-2fps
NOTE 2: This local 'caching' of settings is not optimal and an alternative solution will be worked on after 0.4.11 is released
- Remove blank setting values when setting has a group
- Pair setting values with groups in file when possible
- Preserve user-set whitespace in setting objects
- Delete setting value when setting NoiseParams group
- Delete overwritten groups outside of lock
Without this patch node 0,0,0 is highlighted when enable_node_highligting is false
There is a minor lighting issue remaining, however it seems to be related to a different bug (https://github.com/minetest/minetest/issues/1887)
The ratios between the sizes of form elements, including text, is now
fixed, aside from variations caused by rounding. This makes form layout
almost fully predictable, and particularly independent of player's
screen size. The proportions of non-text elements are the traditional
proportions.
For compatibility, the way in which element positions and sizes are
specified remains unchanged, in all its baroqueness, with one exception.
The exception is that the position of a label[] element is now defined
in terms of the vertically center of the first line of the label,
rather than the bottom of the first line of the label. This change
allows a label to be precisely aligned with button text or an edit box,
which are positioned in a centering manner. Label positioning remains
consistent with the previous system, just more precisely defined.
Make multi-line label[] elements work properly. Previously the code set
a bounding rectangle assuming that there would be only a single line,
and as a result a multi-line label would be cut somewhere in the middle
of the second line. Now multi-line labels not only work, but have
guaranteed line spacing relative to inventory slots, to aid alignment.
Incidentally fix tabheader[] elements which were being constrained to
the wrong width.
Given an unusually large form, in variable-size mode, the form rendering
system now chooses a scale that will fit the entire form on the screen,
if that doesn't make elements too small. Fixed-size forms, including the
main menu, are have their sizes fixed in inch terms. The fixed size for
fixed-size forms and the preferred and minimum sizes for variable-size
forms all scale according to the gui_scaling parameter.
Fix typo in Android Makefile ndk path.
Fix touchscreen parts of game.cpp to work after Zeno's refactor.
Fix isdigit and isspace overload conflict with Android Irrlicht in string.h
Enable sensor landscape rotation in Android Manifiest.
Add mapgen v5 to Android build.
Fix Makefile not checking leveldb.
Signed-off-by: Craig Robbins <kde.psych@gmail.com>
f7d6509 introduces error when no endian.h found in the system.
Since "CHECK_INCLUDE_FILE" returns empty string instead of "0", when
"cmake_config.h" is generated it has "#define CMAKE_HAVE_ENDIAN_H " line.
Later we have "#define HAVE_ENDIAN_H CMAKE_HAVE_ENDIAN_H" in the
"config.h", an thus "HAVE_ENDIAN_H" is also empty sting. Because of
this, "#if HAVE_ENDIAN_H" is incorrect preprocessor directive.
Signed-off-by: Craig Robbins <kde.psych@gmail.com>
nodemetadata.cpp, nodetimer.cpp
optimzation: simpler deserialize node position method
staticobject.cpp:
cleanup: use util/serialize.h inlines instead of its own de/serialization
serialize.cpp:
minor optimization/cleanup: avoid generation of unneeded string temporary
CMakeLists.txt, cmake_config.h.in: detection of endian.h
config.h: added HAVE_ENDIAN_H
Commits due to feedback squashed
Signed-off-by: Craig Robbins <kde.psych@gmail.com>
- Don't create and cache an extruded mesh for every (non-node) item.
Instead use a single one per image resolution.
- For cubic nodes reuse a single wield mesh too
- Improve lighting of the wielded item
- Increase far value of wield mesh scene camera, fixes#1770
- Also includes some minor refactorings of Camera and GenericCAO.
a) Fix double tap for jump and show_debug
b) Revert changes to limitFps()... there is no verification that the new method was an issue, but going back to old method just in case
b.2) limitFps() no longer calls device->run() and also no longer has to be called immediately after a call to device->run()
The following is a record of 31 commits before squashing:
Revert "Remove m_ext_ptr in GUIFormSpecMenu, replaced by refcount mechanism"
This reverts commit b49e5cfc7013cef7e9af79d17e04f7e7e4c377d4.
Basic reformatting with astyle
-- additional formatting will be modified, manually, as the need for it is encountered
Start "outlining" what a MinetestApp class might look like
Add MinetestApp::shutdown()
Converted class member functions to camelCase and created protos for new functions
First stage of connect to server done
Add get itemdefs/nodedefs/media code
Init clouds, camera, sky, init GUI, HUD
Input handling
Client events, camera, sound, draw
Fix wield hand getting stuck digging and add debug text back
Fix FPS
Added profiler graph back
Fix FPS issue
Need to work out what went wrong and clean up the copy/paste stuff
Annotate
Various:
Rewrote limitFps()
Limited scope of some variables
Jitter calcs
Reduce scope of objects
Move some stuff out of ::run and minor formatting cleanup
Scope reduction
Function splits
Removed old (broken) limitFps()
Added exception handling back
Fixed some formatting
Reverted commented out unit tests (uncommented them)
Slow clouds down on loading and media screens so the behaviour is like the original the_game()
Formatting/style (no functional changes)
Manually reapply upstream b49e5cf: Remove m_ext_ptr in GUIFormSpecMenu, replaced by refcount mechanism
Fixed silly errors on my part
Minor formatting cleanups
Removed strange differentiation in FPS limiting when loading
FPS limiting was done differently if cloud_menu_background was true, which does not make sense
Cleaning up
Add some comments
Since minetest has no mobs within the core anymore, I suppose these
settings and code should go. Any mod that uses `minetest.setting_getbool`
will work with no problem since the default return value is `false`.
