| Commit message (Collapse) | Author | Age |
|
|
|
| |
more robust
|
|
|
|
|
|
|
|
|
|
| |
Previously, we assumed that reply->str was NULL
terminated. However, this turned out to be not true,
as users reported crashes in strlen connected to
where reply->str was appended to an std::string.
Use the method recomended by the docs, to read the
length separately.
|
|
|
|
|
| |
Add new node property 'floodable', default false
Define "air" as floodable = true in C++ and lua
|
|
|
|
| |
cases. This change covers all use cases.
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
| |
Removing the target file on other platforms was enabled likely unintentionally
by commit 5f1f1151d3a9c113902630adc16cc3f4845da7ba.
This may be the reason why there has been corruption of files on Linux on hard
shutdowns.
Previously I described the problem and this fix in issue #3084.
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
My commit
e2d54c9f9275e4f77ec33be8054621d42945f7a4 "shutdown when requested from lua in singleplayer too"
broke minetest's feature to connect to servers. The client crashed
after the connection init was complete.
Thanks to @sofar for reporting the bug.
Fixes #3498.
|
|
|
|
|
|
|
| |
If an `XDG_CACHE_HOME` can't be found or `RUN_IN_PLACE` is enabled,
`path_cache` is left at its default of `$PATH_USER/cache`
(at a time when `PATH_USER` is `..`), rather than being reset to
`$PATH_USER/cache` after `PATH_USER` has been properly set.
|
|
|
|
| |
And replace manual tests for error with SQLOK() where possible.
|
|
|
|
|
|
| |
This allows other applications (e.g. minetestmapper) to interrogate
the database while minetest is running, without causing an almost
certain minetest crash.
|
| |
|
|
|
|
|
|
|
|
| |
items"
This reverts commit f14e7bac54af65e3d3d99f89f23f114b17058e49.
Reverted due to missinterpretation of agreement, obvious dislike and me not interested in doing fights for feature I don't actually need
|
|
|
|
|
|
| |
This reverts commit 91bafceee6606fab79db1bde4cba01b84fed65c7.
Reverted due to missinterpretation of agreement, obvious dislike and me not interested in doing fights for feature I don't actually need
|
|
|
|
| |
once per step
|
| |
|
| |
|
|
|
|
|
|
|
| |
Before, minetest.request_shutdown didn't shut down
singleplayer instances or server instances from the server tab.
This commit fixes this. Fixes #3489.
|
|
|
|
|
| |
-> Redis backend: break from switch to fix bug
-> Dummy and redis backends: reserve the count so that creating the list is faster
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
Previous commit
70ea5d552e283bd5bce3278cbf3819bd87ba2602 "Add support for limiting rotation of automatic face movement dir entitys"
by sapier has broken minetest's feature to open worlds. This was due to a
missing stack pop operation.
Thanks to @oleastre for reporting this bug and suggesting the fix.
|
| |
|
| |
|
| |
|
|
|
|
| |
or change the nametag text of players
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
3D Mandelbrot/Mandelbar
3D Christmas Tree
3D Mandelbulb
3D Cosine Mandelbulb
4D Mandelbulb
Plus corresponding julia set for each
Add credits for formulas
Rename parameter 'formula' to 'fractal'
Speed optimisations
|
| |
|
|
|
|
|
|
|
| |
* Horizontal and vertical are used wrongly. Use height and width because
horizontal/vertical describes different things. Thanks @kilbith for pointing out.
* Update minetest.conf.example and settings_translation_file.cpp
* Correct maximum/minimum copy paste mistake.
|
|
|
|
|
|
| |
On non-windows platforms this just used a semaphore,
which meant that multiple calls to signal() would
result in wait() returning multiple times.
|
|
|
|
|
| |
This is set to the XDG cache path where possible.
It's set to the app's cache path on Android.
|
|
|
|
|
|
|
|
|
|
|
| |
Fix the type of the threshold value for mapgen.
The commit
a78dd7f2b6b0e1fefdbaa1ae21b722dd4459e4f4 "Fix spelling of noise_threshold"
has changed it to be read as int, but it can have non-integral values too.
Thanks to @kwolekr for pointing this out.
|
|
|
|
|
|
|
|
|
|
| |
To terminate unwanted shadows from floatlands or realms above
Also add to LuaVoxelManip calc_lighting for use in mapgen mods
Remove the 2 argument calcLighting, mapgens now use the 5
argument form to specify the volumes for propagateSunlight and
spreadLight
In mgsinglenode replace calcLighting with setLighting and
clean-up use of tabs and spaces
|
| |
|
|
|
|
|
|
|
| |
On openSUSE luajit is not detected correctly.
This is because openSUSE is using a lua version suffix, like other Linux distributions do it also.
So the include directory is:
include/luajit-5_1-2.0
|
|
|
|
|
|
| |
openSUSE sets that option to ON.
Fixes #3420.
|
|
|
|
| |
Mgv7/flat/fractal: Reduce tunnel noise spreads to 96
|
| |
|
|
|
|
|
|
|
| |
A segfault exposed a missing 'vm->m_area.contains()'
check in makeCorridor that allowed the calculation
of vm index for a node outside the vm area. The huge
and invalid index number caused getContent to fail
|
| |
|
|
|
|
|
|
| |
For future river mapgens
Dungeons will not generate in river water, to
avoid dungeons filling and blocking river channels
|
| |
|
|
|
|
|
|
| |
Improve default parameters
Update and improve documentation
Unhide mapgen, but is still unstable
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Fixes bug for which commit
6c37e89f08f962eaba788a31f5d3c798ceaa65e6 "Fix old client showing duplicated health bar on new server"
by @sapier
laid the groundwork, where the server has
updated its copy of the hud flags without
respecting the mask.
Fixes #3395.
|
|
|
|
|
|
|
|
|
|
| |
Flag is set by default in MapgenParams
The global 'trees' flag remains but is now
undocumented and unset by default in MapgenParams
Add mgv6_spflag 'trees' set by default in
defaultsettings.cpp to affect new worlds only
This is automatically backwards
compatible for existing worlds
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
* Remove trailing lines from src/sound_openal.cpp
* Don't do a horribly ugly copy of the file's path, allocating and deallocating a
10 kb buffer in the process. This copy was needed for backwards compatibility
with libvorbis 1.3.1 and earlier, as the removed comment explains.
However, even Ubuntu precise has 1.3.2 already. Dropping support and sparing
the ugly copy can therefore be considered safe.
* Actually load sounds from the memory, not caching them at the disk first,
removing the old hack. This is the main motivation for the commit.
|