This moves the "async" pathfinding computations to a worker, preparing
the architecture for threading.
Tested By: Kuba386, Stan`
Differential Revision: https://code.wildfiregames.com/D1918
This was SVN commit r22902.
Make that function static, so that it can be used for these functions
without slowly having to obtain the ScriptInterface instance using
GetScriptInterfaceAndCBData just to get the JSContext again.
Remove few redundant conversions for CreateObject arguments.
Differential Revision: https://code.wildfiregames.com/D2128
Tested on: gcc 9.1.0, clang 8.0.1, Jenkins
Tedious performance testing in: D2128, D2127
This was SVN commit r22894.
D1987/99a341f379 introduced logic to predict the target movement, which
fixed unit chasing. However, sometimes fleeing units would then predict
that their target will end up in front of them, so they turned around
towards the attacker.
This is fixed by not anticipating the position when it would cause the
vector towards to target to change direction.
Reported By: Freagarach
Tested By: Freagarach
Fixes#5541
Differential Revision: https://code.wildfiregames.com/D2275
This was SVN commit r22885.
The function was put in an anonymous namespace to avoid linking errors
should another one be defined elsewhere.
Reviewed By: elexis
Differential Revision: https://code.wildfiregames.com/D2243
This was SVN commit r22825.
By leveraging custom ordering in std::set, MakeGoalReachable and
FindNearestNavcellInRegions can be optimised and the code simplified.
Differential Revision: https://code.wildfiregames.com/D1882
This was SVN commit r22817.
Units can be stuck in gather states when the target ought to be
collectable from, but it's actually unreachable.
A new FINDINGNEWTARGET intermediate state for GATHER will try and find a
new, different target to collect from. This should generally improve the
behaviour (perfect reachability checks would be required to completely
fix it).
Use this also when the target is knowably unreachable / uncollectible.
Fixes#5529
Differential Revision: https://code.wildfiregames.com/D2120
This was SVN commit r22816.
In some rare cases, units could be stuck in the special state of 3
failed path computations, making them always compute long paths instead
of trying short paths again. This can happen when they compute a long
path successfully, but the unit cannot actually move as it gets
obstructed right away.
Make sure this state is never kept for more than one turn to fix this
problem.
Refs #5569 (probable fix but kept open for further investigating).
Differential Revision: https://code.wildfiregames.com/D2239
This was SVN commit r22815.
The ModifiersManager system component provides an interface to add and
remove modifiers, and get modified stats.
The goal is to merge all the different stat-modifying systems 0 A.D. has
implemented over the years.
This commit makes technologies and auras use ModifiersManager. Some
cheats and AI bonuses also have a similar stat-modifying effect that
have not yet been updated.
Further, this system component makes it possible for e.g. triggers to
easily add modifiers, enabling the writing of Castle Blood Automatic,
RPG or Tower Defense maps without the need for mods or hacks.
The 'Modifier' name was preferred over 'Modification' as it is shorter
and more readable, along with the logic that 'modifiers' store
'modifications' and this stores modifiers. Renaming of other functions
and classes has been left for future work for now.
Internally, this uses a JS data structure. If performance issues arise
with it in the future, this data structure or the whole component could
be moved to C++.
The performance has been tested to be about as fast as the current
implementations (and specifically much faster for global auras with no
icons). Testing showed that sending value modification messages was by
far the slowest part.
Comments by: leper, Stan, elexis
Differential Revision: https://code.wildfiregames.com/D274
This was SVN commit r22767.
Optimise MakeGoalReachable using global regions, by leveraging the fact
that we can easily know which goal regions we can reach. Particularly
for point-goals, it becomes almost instant.
This does not change hashes.
Differential Revision: https://code.wildfiregames.com/D1835
This was SVN commit r22653.
A call to SelectAnimation("idle") was left over.
Reported by: elexis
Fixes#5552
Differential Revision: https://code.wildfiregames.com/D2169
This was SVN commit r22648.
This moves the renderer options into their own class to:@
- allow one to only include the rendering options, not the whole
renderer header, when one wants access to rendering options.
- centralise rendering changes and their side-effects.
- clean up code.
Tested by: historic_bruno, Freagarach
Commented by: historic_bruno
Differential Revision: https://code.wildfiregames.com/D1929
This was SVN commit r22610.
Use the "pretend correct path" behaviour for short goals too. This is a
fix for #5545, since the position check now works correctly without
needing to add a tricky check for path-vs-path distance.
Simplify HandleObstructedMove to use ComputePathToGoal more. This means
we occasionally compute a long path when stuck, which fixes two cases of
stuck units reported in #5547.
Move some common calls into functions for convenience.
Make sure the short-path-waypoint-range-relaxing introduced in
32e8ed51aa doesn't happen for the last waypoint, which caused units to
occasionally never get in range of the last waypoint.
Clear short path waypoints when computing a long path.
This won't fix all instances of unit dancing, but it should improve most
of them.
Fixes#5545, Fixes#5547
Differential Revision: https://code.wildfiregames.com/D2135
This was SVN commit r22609.
A random phenotype can be chosen by giving multiple tokens to the
template.
This allows giving different looks to the same template.
Comments By: stan, vladislav, elexis
Patch By: Freagarach
Reviewed By: wraitii
Differential Revision: https://code.wildfiregames.com/D1955
This was SVN commit r22586.
BeginPathing renamed to ComputePathToGoal, as that is what this function
does.
IsMoving renamed to IsMoveRequested, as the function returns true when
the unit has a move request going on, not when it is actually moving
across the map, which was misleading.
UpdateMovementState's implementation moved closer to where it is used.
PathIsShort and WAYPOINT_ADVANCE_MAX are currently unused, thus deleted.
Differential Revision: https://code.wildfiregames.com/D2067
This was SVN commit r22568.
This helps with #3144 and units not going idle when ordered to clump
together.
By rejecting paths that would not take the unit closer to the goal than
it is, we can avoid the case where a unit at A finds a short path to B,
goes there, gets stuck, finds a new short path to A, etc. ad infinitum.
It doesn't completely fix the problem since two units moving might still
occasionally become stuck against one another, but it makes it rarer
(unit pushing would probably finish solving this).
This assumes that being as close as possible to the goal is the best
behaviour when trying to move somewhere, even when it is unreachable.
Refs #3144
Differential Revision: https://code.wildfiregames.com/D2075
This was SVN commit r22566.
This fixes a regression introduced by 055c848c1a: when an entity is
ordered to move to a target within short path distance (or direct path
distance), it no longer computed at least one long path, which meant it
could be stuck forever if the target was not actually reachable (such as
behind a wall).
To fix this, compute a long path after 3 failed computations, which
should result in a delay of 1-3 turns. The previous code did this after
1 failed try - the decision to make it 3 is mostly based on the idea
that in most cases, being stuck means we ran into units, not that we
were ordered somewhere close. Should there be complaints, it could be
lowered to 2 or 1.
This fixes a second issue, reported in #4473: units sometimes get stuck,
particularly when trying to garrison a turret from 'inside' the walls.
The issue is that the turret is not accessible via the inside as its
obstruction + garrison range is blocked by the surrounding walls.
However, as introduced by 6e05a00929, TryGoingStraightToTargetEntity
ignores all entities with the obstruction group of the target (the
reason for this being that otherwise it would never succeed, since the
line towards the target would likely go through the target).
For walls and formations, this means ignoring possibly too many
entities, and in the case of #4473, ignoring wall pieces. The unit thus
mistakenly thought it could direct-path to the turret, and got stuck.
To fix this, we can ignore specifically the targeted entity's
obstruction tag. This can be considered a fix to 6e05a00929.
temple accepted an earlier version of this patch (specifically elexis'
version).
Fixes#4473
Based on a patch by: elexis
Differential Revision: https://code.wildfiregames.com/D1424
This was SVN commit r22533.
As reported by #5521, Ordering units to walk to a point in a forest can
lag terribly, as units will end up computing long short paths in the
forest, which can result in `ComputeShortPath` calls that individually
take upwards of 80ms.
This can be alleviated by allowing units to stop a bit earlier. A23
handled this in UnitMotion directly, but it's better to handle this in
UnitAI to avoid surprises and to make it customisable on a per-state
level.
This diff further speeds up using the short pathfinder by starting with
a smaller search range, limiting the max-range more and moving the range
slightly towards the goal.
This also refactors UM sending messages to UnitAI so that we may in the
future push more information (in particular, the entity_id that a unit
was obstructed by could be interesting for COMBAT).
This doesn't fix the possibility of lag, but it reduces its occurrence
to levels that should be similar to A23 and thus acceptable enough.
Tested By: Freagarach
Fixes#5521
Differential Revision: https://code.wildfiregames.com/D2105
This was SVN commit r22526.
Units right now try going to their next long waypoint using the
short-range pathfinder. This works, but it tends to leads to units
clumping together when shuttling for example.
By switching to short paths earlier, and by scrapping the long waypoints
when doing so, we can make movement more natural until we have unit
pushing.
Some cleanup in how the short-path domain range gets handled, and
increase the max-range by one tile to improve rare cases.
Differential Revision: https://code.wildfiregames.com/D2095
This was SVN commit r22507.
As reported by @bb in #5512, catapults and other big-range units might
not approach targets correctly when trying to get in range.
To be sure that we will be in range of a square obstruction approximated
as a circle, we need to consider the inscribed circle and not the
circumscribed circle.
Reported by: bb
Fixes#5512
Differential Revision: https://code.wildfiregames.com/D2086
This was SVN commit r22495.
Three changes:
- Assume a certain incertain based on distance to the target, to avoid
recompute paths every turn when the target is far way and moving.
- Handle cases where the target is unreachable to the long-range
pathfinder and we would be recomputing every turn.
- If we went straight, assume we don't need to recompute a path.
These together make moving entities recompute paths far less often,
speeding up the game.
Differential Revision: https://code.wildfiregames.com/D2066
This was SVN commit r22474.
Compute the outward quadrants once and for all instead of setting them
dynamically, because there is no reason why we should always arrive from
the same quadrant as the first time we see a vertex.
Don't consider quadrants for the start-vertex, because of the edge
expansion (which can put us in illegal quadrants)
These result in (much) better paths, the tradeoff being that we now look
at some more vertices.
Fixes#5476
Differential Revision: https://code.wildfiregames.com/D1908
This was SVN commit r22473.
Since m_ExpectedPathTicket.m_Type is uninitialised before being used,
serialisation could fail with an out of bouds error. This fixes it by
giving it an (arbitrary) default value.
Reported by: gameboy
Confirmed and debugged by: Angen
Reviewed By: Angen
Differential Revision: https://code.wildfiregames.com/D2074
This was SVN commit r22470.
Reorder code flow, handle long paths and short paths in a more explicit
manner, and only fail after a certain number of failed path computations
to avoid going idle too easily.
Make sure WALKING orders in UnitAI stop when the move fails to avoid
units being stuck.
Differential Revision: https://code.wildfiregames.com/D1907
This was SVN commit r22464.
As reported by @Freagarach on #5496, there can be broken behaviour as
UnitMotion::PathResult may call RequestLongPath with an uninitialised
path goal when ComputeGoal fails.
To fix this, check the return value everywhere and react accordingly in
case of failure.
Fixes#5496
Differential Revision: https://code.wildfiregames.com/D2063
This was SVN commit r22458.
m_PathState kept 3 pieces of information that can be deduced using other
variables:
- whether the entity was following a path. This can be better deduced
by checking if the entity has short/long waypoints.
- whether a path is currently being requested. This can be better
deduced by checking the value of m_ExpectedPathTicket
- whether the requested path is long or short. This can be stored
directly alonside the path request ticket number, making it more obvious
why it exists.
With these changes, m_PathState can be removed.
Differential Revision: https://code.wildfiregames.com/D1903
This was SVN commit r22456.
It is generally better to use the movement request data as that is our
actual target and goals are just a representation of that at any given
time. Where a PathGoal is required, compute this lazily.
This allow failing easily in MoveToX() functions if the goal cannot be
computed.
Differential Revision: https://code.wildfiregames.com/D1902
This was SVN commit r22452.
Introduced by 70e22e2923. PositionChanged messages sent between
MT_Destroy handling in CmpObstruction and the actual destruction of the
component could re-recreate an obstruction, which would then live
forever in the obstruction manager.
This prevents that by de-activating the obstruction when it is
destroyed, so that further PositionChanged messages do nothing.
Differential Revision: https://code.wildfiregames.com/D2060
This was SVN commit r22451.
MoveToPoint and MoveToTarget both compute a goal from a move request.
They can be combined to reduce duplication and streamline the code.
Differential Revision: https://code.wildfiregames.com/D1984
This was SVN commit r22450.
HandleObstructedMove contains logic to drop long waypoints, should they
be occupied by units (because that can get the pathfinder stuck, since
the long-range pathfinder can find a path, but not the vertex one).
However PathResult already drops waypoints this way, so this code can be
removed for simplification.
Differential Revision: https://code.wildfiregames.com/D1981
This was SVN commit r22448.
This fixes a number of issues with formations:
- Gives the controller an obstruction. This fixes the bug where units in
formations can't gather from a tree.
- Stop special-casing formation members in PathResult. This fixes
formation members being stuck in-place when they run in an obstruction.
- Makes sure units stay in Formation.IDle when they are idle so that
D1337/D1901 can work correctly in the IDLE state.
- Warn if animals enter this state.
- Make sure that formation members that end up in INDIVIDUAL.IDLE go
back to FORMATION.IDLE for sanity and for better housekeeping (refs
#3144 - fixed completely upstream).
Differential Revision: https://code.wildfiregames.com/D2048
This was SVN commit r22447.
Use UpdateMovementState to inform the visual actor of the unit's speed,
which cill update the movement animation accordingly.
The removes the need for UnitAI to handle movement animation using the
special "move" state.
Differential Revision: https://code.wildfiregames.com/D1901
This was SVN commit r22446.
The animation overrides are a little awkward. Instead, use animation
variants, which allow defining different animations (and props and such)
easily for different units in a formation.
This lets us clean up the special C++ code to override the walk
animations.
It continues the work started by 8446b92f1b.
Original Patch By: temple
Differential Revision: https://code.wildfiregames.com/D1337
This was SVN commit r22442.
Following D1899/98f609df1d, we have logic that can handle calling
MoveSucceeded() even if we don't go through the special STATE_STOPPING
state, and following D1898/8ac104b07a we no longer need STATE_STOPPING
for animation sync either.
m_State can therefore be entirely removed as it is redundant with other
information.
Differential Revision: https://code.wildfiregames.com/D1900
This was SVN commit r22441.
After the recent UM changes, units sometimes chase/flee forever as they
can never actually get in range.
This is because moving to the current target's position is not enough
when the target is moving.
By accounting for the target movement's in ComputeTargetPosition, the
behaviour is much improved.
Differential Revision: https://code.wildfiregames.com/D1987
This was SVN commit r22431.
This new version compares the final waypoint with the target's
obtruction shape and uses reachability checks to know if we will be in
range or not.
Differential Revision: https://code.wildfiregames.com/D1983
This was SVN commit r22430.
This is a partial revert of 4fda917f46, which skipped the
"MoveToTargetAttackingRange" in APPROACHING. I (incorrectly) assumed
that the original order was still perfectly fine, but in fact the
attacker's max range may have changed as that depends on the relative
elevation between attack and target - and so the original order might
never get us in range!
This was introduced originally in 8c74df2acd.
Add a comment to clarify this.
Further, this makes sure UnitMotion still is aware that it has a target
even if it is in range from the beginning, as that could lead
to stuckiness (and did when chasing sometimes). This was done in D1984
anyways.
Fixes#5478.
Differential Revision: https://code.wildfiregames.com/D2035
This was SVN commit r22429.
This makes it possible to merge the TryGoingStraightTo family of
function.
This means it is now called even for point goals and decreases the
likelihood of stuck units (as we take over from the long-range
pathfinder which brings units to navcell centers, which may not be in
the goal).
Differential Revision: https://code.wildfiregames.com/D1982
This was SVN commit r22426.
D981/c219ee54b2 changed IsInXChecks to use edge-to-edge distance instead
of centre-to-edge, which broke UnitMotion's min-range movement, which
assumed distance to the center.
Since units are represented as squares, the diagonal point may be closer
to the target than the "real" clearance by a factor √2, so the delta
between minimum range and maximum range should be at least `(√2 - 1) *
clearance` to be safe in all situations (this is generally not a problem
for regular units which have a clearance of 0.8, but could be one for
catapults or elephants).
Differential Revision: https://code.wildfiregames.com/D1969
This was SVN commit r22422.
Previously, unitMotion had no code that checked particularly if the
target was still in the world.
When the target moved out of the world, unitMotion would follow the path
to its last known position, then send a "MoveSucceeded" message once
there.
Following 98f609df1d, this message was no longer sent. Thus unit would
follow their path to its last waypoint and stay there, unable to carry
on or finish the order. UnitMotion now explcitly sends a "MoveFailed"
message.
This still changes behaviour from A23, requiring further revisions to
UnitAI (see D1992 for one such case).
Minohaka tested an earlier version of this change (which incorporated
D1992) and accepted it.
Differential Revision: https://code.wildfiregames.com/D1979
This was SVN commit r22415.
