MSVC and GCC evaluate parameters in right-to-left order, whereas Clang
evaluates in left-to-right order, and of course, an optimization could
leave the order of evaluation completely indeterminate.
This commit fixes all instances of the error by explicitly assigning the
results of expressions that use PseudoRandom::next() or range() to their
respective vector components.
The right-to-left evaluation behavior is preserved since Clang is much less
commonly used to compile Minetest than GCC and MSVC combined, and would
therefore cause the least harm.
BiomeGen defines an interface that, given a set of BiomeParams, computes biomes
for a given area using the algorithm implemented by that specific BiomeGen.
This abstracts away the old system where each mapgen supplied the noises
required for biome generation.
We can remove the function in MtNativeActivity now
as it serves precisely that purpose: to tell irrlicht
that we handled the esc key.
TODO for later:
* Perhaps try to find a more performant container than KeyList
Fixes a bug where packet reordering made the server give the
client two peer ids instead of one. This in turn confused
reliable packet sending and made connecting to the server fail.
The client usually sends three packets at init: one "dummy"
packet consisting of two 0 bytes, and the init packet as well as
its legacy counterpart. The last one can be turned off since commit
af30183124, but this is of lower
relevance for the bug. The relevant part here is that network
packet reorder (which is a normal occurence) can make the packets
reach the server in different order.
If reorder puts the dummy packet further behind, the following
would happen before the patch:
1. The server will get one of the init packets on channel 1 and
assign the client a peer id, as the packet will have zero as
peer id.
2. The server sends a CONTROLTYPE_SET_PEER_ID packet to inform
the client of the peer id.
3. The next packet from the client will contain the peer id set by
the server.
4. The server sets the m_has_sent_with_id member for the client's
peer structure to true.
5. Now the dummy packet arrives. It has a peer id of zero, therefore
the server searches whether it already has a peer id for the
address the packet was sent from. The search fails because
m_has_sent_with_id was set to true and the server only searched
for peers with m_has_sent_with_id set to false.
6. In a working setup, the server would assign the dummy packet to
the correct peer id. However the server instead now assigns a
second peer id and peer structure to the peer, and assign the
packet to that new peer.
7. In order to inform the peer of its peer id, the server sends a
CONTROLTYPE_SET_PEER_ID command packet, reliably, to the peer.
This packet uses the new peer id.
8. The client sends an ack to that packet, not with the new peer id
but with the peer id sent in 2.
9. This packet reaches the server, but it drops the ACK as the peer
id does not map to any un-ACK-ed packets with that seqnum. The
same time, the server still waits for an ACK with the new peer
id, which of course won't come. This causes the server to
periodically re-try sending that packet, and the client ACKing it
each time.
Steps 7-9 cause annoyances and erroneous output, but don't cause
the connection failure itself.
The actual mistake that causes the connection failure happens in 6:
The server does not assign the dummy packet to the correct peer, but
to a newly created one.
Therefore, all further packets sent by the client on channel 0 are
now buffered by the server as it waits for the dummy packet to reach
the peer, which of course doesn't happen as the server assigned
that packet to the second peer it created for the client.
This makes the connection code indefinitely buffer the
TOSERVER_CLIENT_READY packet, not passing it to higher level code,
which stalls the continuation of the further init process
indefinitely and causes the actual bug.
Maybe this can be caused by reordered init packets as well, the only
studied case was where network has reliably reordered the dummy
packet to get sent after the init packets.
The patch fixes the bug by not ignoring peers where
m_has_sent_with_id has been set anymore. The other changes of the
patch are just cleanups of unused methods and fields and additional
explanatory comments.
One could think of alternate ways to fix the bug:
* The client could simply take the new peer id and continue
communicating with that. This is however worse than the fix as
it requires the peer id set command to be sent reliably (which
currently happens, but it cant be changed anymore). Also, such a
change would require both server and client to be patched in order
for the bug to be fixed, as right now the client ignores peer id
set commands after the peer id is different from
PEER_ID_INEXISTENT and the server requires modification too to
change the peer id internally.
And, most importantly, right now we guarantee higher level server
code that the peer id for a certain peer does not change. This
guarantee would have to be broken, and it would require much
larger changes to the server than this patch means.
* One could stop sending the dummy packet. One may be unsure whether
this is a good idea, as the meaning of the dummy packet is not
known (it might be there for something important), and as it is
possible that the init packets may cause this problem as well
(although it may be possible too that they can't cause this).
Thanks to @auouymous who had originally reported this bug and who
has helped patiently in finding its cause.
Version 1.8.2 of irrlicht changed the way that IGUIStaticText::getTextHeight() works and since that release properly deals with newlines.
From irrlicht changes.txt for 1.8.2, "IGUIStaticText::getTextHeight returns now the correct height for texts with newlines even WordWrap is not set."
* Remove the copy from db::loadBlock by using a pointer to the destination
* cleanup db backend, the child backend doesn't have to set their functions as virtual
The path finding code works fairly well except that it considers
anythin not CONTENT_AIR to be "above the surface". This results in
paths that are unwalkable for entities since e.g. plants are not
walkable. The path would force them to jump on top of grass plants,
etc..
The obvious solution is not to use CONTENT_AIR as a criteria, but
instead distinguish between walkable and non-walkable nodes. This
results in paths that properly walk through grass nodes.
This was extensively tested by a flock of electric sheep.
Note that for underwater purposes this changes the behaviour from
"the surface is walkable" to "ignore water entirely" making the
path go across the water bottom, and pathing fail likely from the
water surface. This is intentional.
Before, the GridNodes were stored in vector<vector<vector<T>>>,
and initialized in advance. Putting three vectors inside each other
puts lots of unneccessary stress onto the allocator, costs more memory,
and has worse cache locality than a flat vector<T>.
For larger search distances, an the array getting initialized means
essentially O(distance^3) complexity in both time and memory,
which makes the current path search a joke. In order to really
profit from the dijkstra/A* algorithms, other data structures
need to be used for larger distances.
For shorter distances, a map based GridNode storage may be slow as
it requires lots of levels of indirection, which is bad for things like
cache locality, and an array based storage may be faster.
This commit does:
1. remove the vector<vector<vector<T>>> based GridNodes storage that
is allocated and initialized in advance and for the whole
possible area.
2. Add a vector<T> based GridNodes storage that is allocated and
initialized in advance for the whole possible area.
3. Add a map<P,T> based GridNodes storage whose elements are
allocated and initialized, when the path search code
demands it.
4. Add code to decide between approach 2 and 3,
based on the length of the path.
5. Remove the unused "surfaces" member of the PathGridnode class.
Setting this isn't as easy anymore for the
map based GridNodes storage.
* Fix naming style for methods and classes:
Use camelCase for methods and PascalCase for classes as
code style demands it. And use sneak_case for methods that
are not member of a class.
* Replace "* " with " *" for Pointers
* Same for references
* Put function body opening braces on new line
* Other misc minor non functional style improvements
There is no need to put them into the header, they are solely used
inside the pathfinder.
Another advantage of this change is that only the pathfinder.cpp has
to be compiled if PATHFINDER_DEBUG gets defined or undefined, not
all files including the .h.
This commit moves the pathfinder classes to the cpp file without
modifications.
Also, the PATHFINDER_DEBUG macro gets moved to the cpp file and
the PATHFINDER_CALC_TIME macro gets moved to a plce where it
actually does work.
Previously, race conditions occurred inside logging, that caused
segfaults because a thread was trying to use an old pointer that
was freed when the string was reallocated. Using a fixed-length buffer
avoids this, at the cost of cutting too long messages over seveal lines.
The shadow bug at y = 63 was caused by dark air being placed as dust,
when the biome dust was unspecified it was falling back to 'air'
In dustTopNodes only dust == 'ignore' will disable dust placement
Previous mountain terrain generation was by necessity placing
stone in air, this was removing air from any overgenerated
structures such as tunnels, dungeons and large caves
Moving it into the base terrain generation loop ensures that
only 'ignore' is replaced
generateRidgeTerrain: only return if node_max.Y < water_level - 16
Previously, if water level was set a few nodes above a mapchunk
border the river channel was only partially excavated
Instead of doing nothing at node_max.Y + 1 use 1-down
overgeneration for tunnel generation and noisemaps
Move some old unused code in mgv7 to end of file
It was caused by player not moving because fall was prevented, but their
velocity still increasing, causing fatal fall damage when world was
finally loaded. This commit fixes it by setting player velocity to zero
when the world around them is not loaded.
* Fix leak like behaviour if you load multiple schematics in a loop.
* Cleanup check in for, fixing theoretical out of bounds read if
Schematic::deserializeFromMts reduced the number of elements
in m_nodenames. A != check may need an overflow of the counter
before it hits, if origsize is larger than m_nodenames.size().
* Fix function name passed to errorstream: it was wrong. Also use
__FUNCTION__ instead of manually using the method name at other
places in the function.
* Don't shadow the name member in the loop.
This fixes#3935, a regression from 0338c2e.
An 'optimization' was performed where an index for the VoxelManip being
operated on was mistakenly used for bounds checking within the incorrect
VoxelArea, namely, the area wherein light should be spread.
This allows the player to more easily target and punch connected
nodeboxes, especially if they have a fixed nodebox that is very
small, like technic cabling, or xpanes. Tried it on fences and
my xpane conversion, and happy with the result.
Changes:
* Fix indentation.
* Pass strings by const reference.
* Merge Strfnd and WStrfnd into one class instead of copying them.
* Remove trailing spaces.
* Fix variable names.
* Move to util.
* Other miscellaneous style fixes.
ABM's are hardcoded to run every 1.0s, NodeTimers are hard coded to
run at every 1.0s. Block mgmt is running every 2.0sec.
However, these timers can be better tuned for both higher and lower
values by server owners. Some server owners want to, and have the
resources to send more packets per second to clients, and so they
may wish to send smaller updates sooner. Right now all ABM's are
coalesced into 1.0 second intervals, resulting in large send queues
to all clients. By reducing the amount of possible timers, one can
get a far better response rate and lower the perception of lag.
On the other side of the camp, some servers may want to increase
these values, which again isn't easily doable.
The global settings abm_interval and nodetimer_interval are set to
current values by default. I've tested with 0.2/0.5 type values
and noticed a greatly improved response and better scattering of
nodetimers, as well as enjoying not faceplanting into doors with
pressure plates anymore.
The legacy init packet (pre v25) sends information about the client's
password that a server could use to log in to other servers if the
username and password are the same. All the other benefits of SRP of
protocol v25 are missed if the legacy init packet is still sent during
connection creation.
This patch adds an option to not send the v25 init packet. Not sending
the v25 packet means breaking compat with pre v25 servers, but as the
option is not enabled by default, no servers are affected unless the
user explicitly flips the switch. More than 90% of the servers on the
serverlist support post v25 protocols.
The patch also fixes a bug with greying out of non compliant servers
being done wrongly, the min and max params were mixed.
Fix two bugs related to the reconnect feature
introduced by commit
3b50b2766a "Optional reconnect functionality"
1. Set the password to the stored one
Before, we have done the reconnect attempt with a
cleared password, so using the feature would only
work if you had an empty password.
Thanks to @orwell96 for reporting the bug.
2. Reset the reconnect_requested flag after its use
the_game only writes to the reconect_requested flag
if it sets it to true. It never sets it to false.
If the flag is not reset after its use, all "reset"s
to the main menu will look like the server had
requested a reconnect.
* No function overloading
* Adhere coding style and with method names following
lowercase_underscore_style
* Use std::string in external API, handling these is
much more fun
Don't excavate the overgenerated stone at node_max.Y + 1,
this creates a 'roof' over the tunnel, preventing light in
tunnels at mapchunk borders when generating mapchunks upwards.
1. Copy-paste error: properly test for back-connection.
In the case of two different connected nodebox types, we want to
assure that if A connects to B, that B also connects to A. This test
was accidentally not implemented correctly.
2. Clear the connects_to_ids before deserializing.
With each new connected node, the deserialization code added more
and more targets to the map, since the map wasn't cleared in between
deserialization steps. This caused e.g. wall blocks to connect to
things in the fence connects_to map.
NDT_CONNECTED attempts to connect to any side of nodes that it can
connect to, which is troublesome for FACEDIR type nodes that generally
may only have one usable face, and can be rotated.
We introduce a node parameter `connect_sides` that is valid for
any node type. If specified, it lists faces of the node (in "top",
"bottom", "front", "left", "back", "right", form, as array) that
connecting nodeboxes can connect to. "front" corresponds to the south
facing side of a node with facedir = 0.
If the node is rotatable using *simple* FACEDIR, then the attached
face is properly rotated before checking. This allows e.g. a chest
to be attached to only from the rear side.
We introduce a new nodebox type "connected", and allow these nodes to
have optional nodeboxes that connect it to other connecting nodeboxes.
This is all done at scenedraw time in the client. The client will
inspect the surrounding nodes and if they are to be connected to,
it will draw the appropriate connecting nodeboxes to make those
connections.
In the node_box definition, we have to specify separate nodeboxes for
each valid connection. This allows us to make nodes that connect only
horizontally (the common case) by providing optional nodeboxes for +x,
-x, +z, -z directions. Or this allows us to make wires that can connect
up and down, by providing nodeboxes that connect it up and down (+y,
-y) as well.
The optional nodeboxes can be arrays. They are named "connect_top,
"connect_bottom", "connect_front", "connect_left", "connect_back" and
"connect_right". Here, "front" means the south facing side of the node
that has facedir = 0.
Additionally, a "fixed" nodebox list present will always be drawn,
so one can make a central post, for instance. This "fixed" nodebox
can be omitted, or it can be an array of nodeboxes.
Collision boxes are also updated in exactly the same fashion, which
allows you to walk over the upper extremities of the individual
node boxes, or stand really close to them. You can also walk up
node noxes that are small in height, all as expected, and unlike the
NDT_FENCELIKE nodes.
I've posted a screenshot demonstrating the flexibility at
http://i.imgur.com/zaJq8jo.png
In the screenshot, all connecting nodes are of this new subtype.
Transparent textures render incorrectly, Which I don't think is
related to this text, as other nodeboxes also have issues with this.
A protocol bump is performed in order to be able to send older clients
a nodeblock that is usable for them. In order to avoid abuse of users
we send older clients a "full-size" node, so that it's impossible for
them to try and walk through a fence or wall that's created in this
fashion. This was tested with a pre-bump client connected against a
server running the new protocol.
These nodes connect to other nodes, and you can select which ones
those are by specifying node names (or group names) in the
connects_to string array:
connects_to = { "group:fence", "default:wood" }
By default, nodes do not connect to anything, allowing you to create
nodes that always have to be paired in order to connect. lua_api.txt
is updated to reflect the extension to the node_box API.
Example lua code needed to generate these nodes can be found here:
https://gist.github.com/sofar/b381c8c192c8e53e6062
The MainMenuManager set the console invisible when a
formspec opened over it, but didn't properly close it,
and the chat console never set itself visible again.
Since commit
3b50b2766a "Optional reconnect functionality"
there is a config option named ask_reconnect_on_crash.
It asks the client to reconnect to the server if the server crashed.
It has been implemeted and works, but due to a function parameter not
being passed it never showed effect. This patch adds the parameter
and fixes the bug.
Also fixes the `reconnect` option of minetest.request_shutdown being ignored.
I've written several experimental bits of code that revolve around the
need for a consistent calendar, but implementing one is extremely hard
in mods due to time changes and mods overriding core.get_timeofday(),
which will conflict.
The second part of the problem is that doing this from a mod requires
constant maintenance of a settings file.
An implementation in core is trivial, however, and solves all of
these problems at virtually no cost: No extra branches in server
steps, and a single branch when minetest.set_time_of_day(), which is
entirely reasonable.
We store the day_count value in env_meta.txt.
The use case is obvious: This change allows mods to create an actual
virtual calendar, or properly account for seasonal changes, etc..
We add a "/days" chatcommand that displays the current day count. No
permissions are needed. It can only retrieve the day count, not
modify it.
Lower water table where base humidity is low.
Alter heat and humidity to compensate for river humidity and altitude
chill.
Correct misuse of surface_max_y in generateTerrain.
Remove sand trails in the water at river mouths.
Remove river water below water_level.
Correct heat/humidity calculations where noises are less than zero.
Correct heightmap errors as much as possible.
Make humidity calculations more readable.
In the calculation of y_start,
when 'column height max' was large it caused
nmin.Y + max_height > nmax.Y - max_height
Now, in this situation y_start is set to the
midpoint between nmin.Y and nmax.Y
Limit y0 and y1 to between nmin.Y and nmax.Y,
otherwise index calculation, which has no checks for limits,
places them at unwanted locations in the voxelmanip
As stated in this forum thread [1], I noticed that there is a
2 second interval in which inventory changes are shown on the
client. @yyt16384 found the source of these 2 seconds:
m_nothing_to_send_pause_timer is set to 2.0 every time there
are no changes to make, but this timer is not reset when
SetBlockNotSent or setBlocksNotSent are invoked. So in worst
case, the changed block will be sent over 2 seconds too late.
With this change, changed inventories are updated almost
immediately, but it causes additional connection load.
This allows mods to perform both asynchronous and synchronous HTTP
requests. Mods are only granted access to HTTP APIs if either mod
security is disabled or if they are whitelisted in any of the
the secure.http_mods and secure.trusted_mods settings.
Adds httpfetch_caller_alloc_secure to generate random, non-predictable
caller IDs so that lua mods cannot spy on each others HTTP queries.
Previously you could steal a secure environment from a trusted mod by wrapping
request_insecure_environment with some code like this:
local rie_cp = minetest.request_insecure_environment
local stolen_ie
function minetest.request_insecure_environment()
local ie = rie_cp()
stolen_ie = stolen_ie or ie
return ie
end
'cam_pos_nodes -/+ box_nodes_d' can exceed the range of v3s16
when a player is near the world edge using a large view range
This previously caused world to disappear
Create new function getBlocksInViewRange() called from
updateDrawList() and renderMap()
Correct code style throughout updateDrawList() and renderMap()
