Adds baseRange parameter to parabolic queries for combined
2D + parabolic detection. StandGround and Chase stances now
use attack range (parabolic) with vision as baseRange, allowing
units to attack enemies visible through friendly vision.
Buildings benefit as well via BuildingAI.
Introduce CanEverReachTarget in the Attack component to check whether
a target is geometrically reachable at all, accounting for height
offsets and turreted units inside buildings.
For non-parabolic attacks (e.g. melee), a simple 3D distance check
from the closest approach point is used.
For parabolic attacks (e.g. ranged), the parabolic range formula
is used to determine if the height difference is surmountable
from the closest horizontal distance to the target.
Add GetClosestApproachDistanceToTurretPoint to TurretHolder
to estimate the minimum horizontal distance to a turret point,
using its local offset and the holder's obstruction size.
Passable buildings are not considered obstacles.
Fixes units on the ground trying to attack unreachable units on walls
or towers when the projectile's arc cannot reach the required height.
The parabolic range formula in GetEffectiveParabolicRange was only
computating height differences from manual HeightOffset values,
completely ignoring actual terrain elevation.
This meant units on hills received no tactical advantage
despite the UI stat tooltip correctly showing extended ranges.
Fixes#8889
CSimulation2 used , CSimulation2Impl mixed /, and
CTurnManager used for the same turn-count concept, forcing a
static_cast<int64_t> at the one place these types already met
(rejoin-test comparison).
Add (alias for std::int32_t) in SimulationCommand.h and use
it consistently for every turn counter in CSimulation2/Impl,
CTurnManager and its subclasses (CLocalTurnManager,
CReplayTurnManager), CNetServerTurnManager/CNetClientTurnManager, and
the network wire format (m_Turn in CEndCommandBatchMessage,
CSimulationMessage, CSyncCheckMessage, CSyncErrorMessage, and
m_CurrentTurn in CLoadedGameMessage).
Turn *duration* fields (m_TurnLength, m_CommandDelay,
DEFAULT_TURN_LENGTH, COMMAND_DELAY_SP/MP, SetTurnLength,
GetSavedTurnLength return type) are left as u32 — they're milliseconds,
not a counter, and out of scope here.
Remaining turn_id_t vs size_t comparisons use std::cmp_equal or
explicit casts, matching the existing pattern in Simulation2.cpp.
Fixes#8718
To follow coding convention rename the vector Node to m_Nodes,
MaxDistance to m_MaxDistance and NodeCount to m_NodeCount.
Signed-off-by: Ralph Sennhauser <ralph.sennhauser@gmail.com>
966727b52e made the player state an enum and introduced more descriptive
functions to achieve the same thing, but it seems they were forgotten to
be replaced in a few places, which this patch fixes.
Instead query g_VideoMode where needed. As both g_xres and g_yres aren't
global anymore remove static from UpdateRenderer. While at it use more
desicriptive names: GetXRes -> GetWindowWidth, GetYRes ->
GetWindowHeight.
This reduces the amount of variables tracking width and height by one
each.
Signed-off-by: Ralph Sennhauser <ralph.sennhauser@gmail.com>
Adjust function arguments to prevent errors when loading/rejoining a game with AI. These occurred due to 2f2cbb96bf not adding arguments for the state/playerID
Fixes: #9001
A regex-based
search across the codebase confirms there are no other instances
of "Man" that should be renamed to "Manager".
Variables were not renamed.
This commit performs code cleanup to improve
code clarity and consistency:
Remove redundant 'virtual' keywords from methods that are already marked
with 'final' or 'override', as well as reducing redundant 'override
final's to 'final'.
The disabled test are all broken one way or another. The maps used in
the tests got renamed. The pathfinder component got some performance
improvements which now require to call UpdateGrid() once before use.
Finally for the InitAttributes to be deep frozen they need to be an
object, so initialize them.
Signed-off-by: Ralph Sennhauser <ralph.sennhauser@gmail.com>
test_hash depends on modmod for xml schemas. InitAttributes need to be
set as deep freezing expects them to be a valid object. The map used got
renamed at some point. Finally prior to running the sim PreInitGame()
and InitGame() need to be called.
Signed-off-by: Ralph Sennhauser <ralph.sennhauser@gmail.com>
Previously, edges where bundled collected first then sorted in 4 AA and 1 unaligned bucket. We can separate the unaligned edges right away, which is a little faster.
Also make sure Vertex::pred is initialized.
This patch splits "cutscene mode" (disabling silhouttes, territory
borders and other visual overlays) from the cinema manager component
and moves it to the renderer, since it doesn't influence the simulation
anyway. The mode can now be independently controlled by the GUI. This
is done so it can also be used for other narrative elements like speech
or dialogue in the future. Cutscene mode is still always enabled while
cinema paths are playing, though.
By design, this also fixes the issue that range overlays weren't hidden
during cutscene mode.
It was previously stored at the end of the array of per-player flags and
set by calling SetLosRevealWholeMap with the player ID -1. However, ever
since the introduction of observer mode in 883f307b40, -1 is the player
ID of observers everywhere else including for GetLosRevealWhole:
GetLosRevealWhole(-1) always returns true in order to reveal the map to
observers, however, the cinema manager, for example, called it meaning
to find out whether the whole map is revealed to all players.
To fix this and avoid confusion about this in the future, this patch
splits this flag from the per-player flag array and introduces new
functions to set and retrieve it.
This name is more descriptive.
And the plan is to split off the extra player value of the vector into
an own flag in the future, and LosRevealAllForAll would have been a poor
name for that.
It's better to construct a js-array from a `JS::RootedValueVector`.
Because it is more strongly typed and the index doesn't has to be
specified when appending an element.
Some usages are replaced with `JS::RootedValueArray`.
Fixes: #8702
When entities in formations were renamed (e.g., during promotion),
the formation would immediately recalculate all member positions,
and queue movement orders causing visible shuffling.
Changes:
1. Transfer existing offsets movement to the renamed entity
to maintain current formation structure
2. Schedule offset recalculation for the next tick to allow proper
reordering after all systems have updated
This preserves formation integrity during renames while allowing
eventual optimal position recalculation.
Fixes#8656
The 0 tolerance to prevent the "waltzing" that was set before this
cannot happen anymore as we now since check for being at destination
before sending a move request in UnitAI. Adding a new small tolerance
now prevents some small movement adjustments of formation members
near their destination.
Fixes#8592
Requested in #8244 for scripting purposes and automated testing. Extend
the original design by adding a means to pass an exit status. This also
comes in handy in case one wants to cleanly error out from JS on parsing
errors of command line arguments as reported in #7967.
Fixes: #8244
Signed-off-by: Ralph Sennhauser <ralph.sennhauser@gmail.com>
`LoadScripts` and `LoadDefaultScripts` have to be called imediately
after the `CSimulation2` constructor.
By doing the necesarry work in the constructor and removing
`LoadScripts` and `LoadDefaultScripts` the interface of `CSimulation2`
is simplified.
Formation controllers now display their movement paths when enabling
UnitMotion debug overlay.
Key changes:
- Formation controllers show long paths in blue and short paths in green
- Formation controllers path are rendered if selecting
some of the formation's members
- Prevent the gui from making redundant calls to SetMotionDebugOverlay
Existing limitations (not regressions):
- Paths are often cleared before entities complete them
- Formation members have short paths that rarely get rendered
Some tasks are invoked multiple times. Normally those tasks are broken
up inside a loop and had to be continued there. With coroutines that is
easier as it's possible to suspend inside a loop.
Coroutines which are lambdas should not capture anythig as the lifetime
of the captured values might end before the coroutine completes. For
that purpose `std::bind_front` is used.
All functions had a `LDR_` prefix. The prefix is removed.
Functions and globals which are only used in Loader.cpp are now
contained in an anonymous namespace.
Use futures to spawn several threads that concurrently work through the
active range queries each turn instead of doing everything serially on
the main thread -- similar to how the pathfinder computes pathfinding
requests asynchronously. This significantly increases performance.
Note: It still can't run in parallel to the rest of the simulation update since
the range computations depend on the state of the simulation (like the
position of units).
When a script in "simulation/helpers/" contained an error. Files in
"simulation/components" aren't loaded. The return value of
`LoadDefaultScripts` indicated an error but was ignored. The simulation
still tried to start.
Now instead of returning a ignoreable error code the error is thrown. In
the common path the error is implicitly rethrown to the JS-function
which tried to start the game.
fixes: #8133
Commit 3eee3a444d added final keyword to
simulation classes but left virtual keywords in place.
Signed-off-by: Ralph Sennhauser <ralph.sennhauser@gmail.com>
The functions to set them explicitly are removed. This makes the
interface of `CSimulation2` smaller.
Also serializationtest and rejointest can't be active at the same time.
Add a warning about that and use a `std::variant`.
