A garrisoned vessel circles its post, engages what comes into range, and
goes back to circling. It steers by writing move_target and lets Ship's own
steering fly it, so it accelerates, banks and brakes exactly like a vessel
in the formation.
It is not in Fleet.ships. It does not follow the formation, does not count
against max_fleet_size(), and does not move when the player flies away -
which is the whole point of stationing it. A garrison that ate a fleet slot
would mean stationing a vessel cost the player a place in their own wedge.
The leash is the load-bearing part. Enemies are judged from the POST rather
than from the vessel, so a garrison that has drifted out to its leash cannot
see further still and be walked away from its outpost one enemy at a time,
and an engagement is clamped to a circle around the post rather than to a
straight line, so it still manoeuvres while pinned.
The orbit starts from where the vessel actually is rather than a random
bearing. A random one makes the first move_target a point on the far side of
the circle, and the vessel flies straight across the middle - through the
outpost it is guarding.
EnemyTargets.nearest_ship now searches the player_ships GROUP instead of
Fleet.ships: an enemy that ignored the guns shooting at it, because they
were not in the player's formation, would let a garrison hold a position
without ever being fought back.
The laser vessel is deliberately not garrisonable. Firing the laser lives in
LaserController, a fleet ability driven by the laser button, rather than in
VesselLaserComponent - so a garrisoned laser would hold its position and
never shoot, and offering a vessel that cannot fight is worse than a short
list.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Vessels come from the HANGAR and never from the fleet. The fleet is what the
player is flying, and taking a vessel out of it to crew a building would be
the game undoing the player's own arrangement. "Whenever available" means
"when one is sitting at the mothership", and an outpost asking for a hull
nobody has built simply waits.
They travel, down the same vessel_deliveries path a finished crafter order
already uses to reach the mothership, rather than teleporting.
What is already in the air counts against the shortfall. Without that the
dispatcher sends the whole hangar to fill one berth, because the crew does
not grow until the first arrival lands.
Two directions, not one: lowering the order flies the surplus home rather
than stranding it, and a vessel whose outpost is levelled under it lands in
the hangar instead of being destroyed - the same rule a transporter
mid-delivery already follows with its cargo.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Landed together because an outpost you cannot crew is decoration, and a
garrison menu with no building to hang it off is nothing at all.
#102, the building. Researched at "Forward Bases", two prerequisites deep
and cheaper than any mid-game structure - the ticket asks for early, and the
answer to waves attacking your buildings should arrive at about the time the
waves can hurt them. It is drawn as a watch mast on a tripod with two open
berths, deliberately the least machinery of any structure here: it makes
nothing and refines nothing, it holds a position. 120hp, more than a
crafter, because being shot at is its job.
#103, the garrison. Two separate things live on its record, and conflating
them was the first thing to get wrong: the ORDER, what the player asked for,
and the CREW, who is actually there. This menu edits the order only; #104
fills the crew. Keeping the order as a wish is what lets an outpost ask for
a hull nobody has built yet and wait for it, rather than the request
evaporating - and it is why the panel shows "here / ordered" instead of one
number.
"Only ships that attack can be assigned" is expressed as capabilities rather
than a list of types, so a new armed vessel is eligible the day it is added
and an unarmed one never is by accident. Today that means the interceptor
and the laser vessel; a shield, a hauler and an engine block cannot hold a
position however much one might want them to.
Two berths by default, and Garrison Bays raises it at outposts ALREADY
standing - the same shape as freight_yards for hubs, because a bonus that
only reached new buildings would mean demolishing and rebuilding to collect
it.
The panel builds its own rows from the catalog rather than from nodes placed
in the scene, so adding an armed vessel later needs no scene surgery.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Every enemy component used to ask the same question - which player SHIP is
nearest - so a structure was scenery they flew straight through. A Building
is also a plain Node2D, so there was nothing to collide with even if one had
tried.
EnemyTargets answers the question once, for pursuit, standoff, the lancer's
lunge and the ranged weapon together, so they cannot drift into disagreeing
about it. It keeps one thumb on the scale: a vessel inside 520px always
wins, even against a nearer building, or the fleet could not defend anything
by interposing itself and an enemy would fly past the ships shooting it.
Buildings get a hitbox - an Area2D in a group, the same shape as a Ship or
an asteroid, so EnemyProjectile needed one extra branch rather than a new
concept. It applies damage on a tick rather than on area_entered: entry
fires once, which is right for a rammer that dies on the first strike, but
would let a sniper that drifted inside sit there scratch-free for ever. On a
tick, an enemy parked on a structure is besieging it.
What one strike costs is data on the enemy definition, for the same reason
dies_on_contact is - a lancer arriving at 950px/s is not a sniper drifting
into a wall.
Also clears the asteroids in test_enemy_projectile_damage. set_process(false)
only stops NEW asteroids; the ones already placed keep drifting, and one
wandering into the shield vessel parked at (2000, 0) spent its single charge
so the projectile fired next killed it. That failed about one parallel suite
run in thirty and never in isolation. Clearing them is what the other combat
tests already do.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The old spawner put one enemy on the map every 3-6 seconds for ever. There
was no rhythm to the danger and nothing to prepare for. Enemies now arrive
in waves: a long quiet, a warning, an attack, and quiet again once the wave
is dead.
The clock lives in GameState, because Main.tscn is destroyed on every trip
to a planet surface and a wave must not be cancelled, nor restarted from
zero, by a landing. The spawning does not: GameState cannot see the world,
so EnemySpawner reads the phase and puts the enemies in it. That also
settles what happens when a wave turns while the player is on a surface -
the wave is waiting in space when they come back.
EnemySpawner keeps its name deliberately. It is still the thing that puts
enemies in the world, and 88 test files already say
`get_node("EnemySpawner").set_process(false)` to ask for a quiet map.
Strength is a function of the FLEET, per the ticket, with the wave number as
a slower second term so a player who never grows still meets escalation. The
budget is spent most-expensive-first, so a large wave is heavier rather than
merely longer, and min_strength keeps the first waves to chasers whatever
the arithmetic says.
"No timer should be visible" is enforced rather than merely observed:
GameState offers wave_warning_pulse(), a 0..1 throb, and no
remaining-seconds accessor at all, so the alert cannot drift into being a
countdown. The test pins both - that no such accessor exists, and that the
alert's own text contains no digits.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Nothing can attack a structure that cannot be hurt, so this is the floor the
rest of #97 stands on.
max_hp is per definition - an extractor bolted to a rock is not the
mothership - and the current value lives on the RECORD in GameState, not on
the Building node. Main.tscn is torn down and rebuilt on every trip to a
planet surface, so damage kept on a node would heal itself the moment the
player landed.
A record written before this carries no hp field at all. That reads as
undamaged rather than as zero, or an existing save would open on a field of
wrecks.
The mothership cannot be destroyed - it floors at 1hp. The bank, the hangar
and research all live there and a wiped fleet rebuilds there, so losing it
would be a soft-lock with no route back. Keyed on the "bank" capability
rather than on the type: whatever building holds the bank is the one that
cannot be lost.
Destroying a hub retires its transporters. Everything else keyed on a
building id already copes with the building being gone - a transporter
mid-delivery puts its cargo back rather than destroying it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The same interop launch flake that produced exit 127 with an empty log has a
second face: a non-zero exit after the engine banner and not one line more.
test_fleet_growth went that way once in six full-suite runs and passed alone
every time.
The line between a flake and a failure is output, not the exit code and not
the banner - every test says something before it stops, its own FAIL
verdict, a SCRIPT ERROR, an engine complaint. A log holding nothing but the
banner is a run that did not happen, so it is retried and, if it stays
silent, reported as NOSTART rather than as a test that failed. A
deterministic silent crash still fails all three attempts.
The check reads grep's output rather than its exit status: `grep -q -v` does
not agree across implementations - ugrep, which is what `grep` resolves to
on the WSL box, reports success for an empty file and failure for a file
full of matching lines.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
A second suicide vessel alongside the chaser, but where the chaser simply
beelines, this one is a rhythm the player can read. It drifts in from beyond
lock_range at half speed, LOCKS a point, rocks 30px back and forth along the
charge axis for 1.2s, draws a thin line down its trajectory for the last
300ms, and only then crosses the gap in one straight 950px/s run before
coasting to a stop and starting over.
The lock is what makes the wind-up worth anything: nothing re-reads the
fleet once it begins, so the charge goes where the target WAS and stepping
off the line works. Same bargain as EnemyBigLaserBeam.
EnemyLungeComponent is a Node2D rather than a plain Node because the
trajectory line has to be drawn somewhere, and top_level keeps that drawing
in world space while the hull it hangs off spins to face the lock.
The roaming spawner gets a 25% lancer share. It is the only place the player
meets an enemy off an authored map, so without that the lancer would exist
only in the setup editor.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The research table's LOCAL column was not merely equal to BANKED, it was the
same store: research lives on the mothership, the mothership is also the
bank, and building_inventory() redirects any bank-capable building to
bank_inventory itself. ResourceBar now drops LOCAL wherever it resolves to
the bank and lets the remaining two columns share the width. Anywhere else -
a crafter, a lab, the transfer panel - the column is untouched.
The "Research" heading went with it. It named a panel you can only reach by
pressing the research button, and it cost the tree canvas 128px, which the
table and the tree have now taken back.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Two sources of random red on WSL, neither of them a test.
A launch that never happened. The Godot binary is a Windows executable
started across the interop boundary, and with six going at once that start
occasionally just fails: exit 127, an empty log, no engine banner, nothing
run. It was indistinguishable from a failure and moved between tests from
run to run. Anything that actually ran got as far as printing the banner, so
the two are easy to tell apart; a start that produced no banner is retried
up to three times and reported as NOSTART if it never comes up. Retries are
printed, never hidden - a test that needed one is saying the machine is
struggling even where it went on to pass.
The working tree as a run directory. WSL ran the suite in place, so six
Godot processes wrote .godot and created and deleted files under
PlanetSetups/ inside the live repo, alongside git and whatever else was
using it. Single tests failed at random under that - a delete that did not
take, a scene that would not instantiate - never the same one twice. WSL now
copies to %LOCALAPPDATA% first, the same isolation Termux already gets from
its own copy, on a native NTFS path so the Windows Godot still reads it at
full speed. Costs about 10s a run and leaves the repo untouched.
Four consecutive clean 124/124 runs since.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Every cooling vessel logged "Invalid polygon data, triangulation failed"
once per frame, for as long as its cooldown ran - 16 of them in a single
test_laser_vessel run. The fill was drawn anyway, so nothing looked wrong,
but it buried real errors in test logs.
_filled_hull cut each hull edge against the fill line itself, which is
Sutherland-Hodgman, and Sutherland-Hodgman is only sound for a CONVEX
subject. Hulls are not: the gas bomber has a tail notch and the distorder a
waist. On those it stitched the pieces back together with a zero-width
bridge, and a polygon with a zero-width bridge is not one the triangulator
accepts.
Geometry2D.intersect_polygons does the clipping properly and returns a list
of polygons, which is also the truer answer - a notched hull genuinely fills
as two puddles until the level rises past the notch. A weld-and-area guard
stays in front of the renderer for repeated vertices and slivers.
Zero triangulation errors across the vessel tests now.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The shot was one flat line that appeared at full length and faded. Nothing
happened at the ship, nothing moved, and a hit looked exactly like a miss.
The blade now runs out of the hull over the first 30% of the flash on a
cubic ease-out, holds white-hot, then falls away, with a hum in its width,
a bright cap at the leading end, a flare at the emitter, and sparks off the
hull that fired it. MuzzleSparkVFX takes an optional colour so the laser
throws cyan where the ordinary gun (#87) keeps its gold.
The extra light is a white-hot rod down the middle at 42% of the width,
deliberately not a wider halo: the note at BEAM_WIDTH_SCALE records that the
drawn beam spans exactly what the beam hits, and a halo past that would put
the "it visibly missed" complaint back the other way round.
The damage is untouched - dealt in full, along the whole beam, on the frame
of the shot. A beam that had to arrive would change how the weapon plays,
which the ticket did not ask for, so the test pins it down: a target at the
far end of the range must be hit before the drawn blade has visibly reached
it.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The card is authored at a fixed 704x1200 around a portrait phone. On a
window wider than it is tall the stretched canvas is about 1080 units high,
so the card sat at y=-60 and hung over both edges - and the confirm and
cancel buttons are anchored to its own bottom edge, so they went off screen
with it. The popup was not merely ugly on a desktop, it was a dead end:
nothing on it could be pressed.
NeonPanel now fits itself to the viewport and re-fits when the window
changes. It only ever shrinks; growing the card would stretch proportions
that are tuned rather than fluid. The fitting adjusts offsets around a
fixed centre, so it applies only to a card pinned to a point - a panel
stretched between different anchors already follows its parent and is left
alone.
Verified across portrait, 16:9, ultrawide and small-laptop shapes, on the
card and on both buttons.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Two separate faults, both reported on the same ticket.
The menu was inert under a mouse. Input reaches the tree as raw events
through BuildController._unhandled_input, because a Control that consumes
them through the GUI system swallows the drag, and the tree only ever read
touch events. That was invisible while the project set
emulate_touch_from_mouse and every click arrived as a synthetic touch; with
that setting gone, nothing selects and nothing scrolls, and there is no
other way to research anything on a desktop. The press and the pan are now
the behaviour, with touch and mouse as two ways in to them. Motion pans only
while the button is held - a hovering pointer must not drag the tree out
from under the reader - and the wheel scrolls, which is the first thing a
desktop player reaches for.
The panel also still had the buy button's hole in it. That button went away
in #83 and nothing moved down into the space, leaving ~180px of dead panel
below the readout while the tree above it stayed cramped. The readout moves
down 154px and the canvas takes the height: 1190px to 1344px.
The test drives clicks through BuildController rather than into the view,
since the missing piece was the routing - calling the view directly would
have passed against the bug as reported.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
On a gas planet the vents were drawn on top of the player. SurfaceFeatures
is added after Fleet in PlanetSurface.tscn, so at the default z_index of 0
the vent field won the tie purely by being the later sibling, and the fleet
flew under its own gas source.
Same one-line fix, for the same reason, as #50's capture rings and the
rocks on a rocky surface. It goes on the field rather than on children
because this one draws in its own _draw().
The test checks the rule instead of the node that was wrong: every drawing
feature, on every surface profile that has features, must end up behind the
fleet - 22 of them today, and any added later without editing the test.
Enemies are excluded and the test says why; they are combatants and share
the fleet's plane on purpose.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
run_tests.sh only knew one way to reach a Godot: `proot-distro login ubuntu`
and the arm64 build at /root. On WSL there is no container and no arm64
binary, so the suite could not be run from a desktop at all.
The platform is detected from /proc/version, and decides two things. Which
Godot: on WSL a Linux `godot` on PATH if there is one, otherwise the newest
Windows build already sitting on /mnt/c - the *_console.exe specifically,
since the plain .exe is a GUI-subsystem binary that detaches from the
terminal and would leave every test log empty. And whether to copy the
project first: that copy exists to get off Android's FUSE bridge, and a
Windows Godot already reads C: natively, so WSL runs in place.
_run_inside.sh now takes the binary as its first argument instead of
hardcoding it, and run_tests.sh calls it directly where there is no
container to enter. The scheduling, timeouts and verdicts are unchanged and
shared by both paths.
120/120 on WSL in 48s with JOBS=6.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
test_desktop_hides_touch_controls failed outright on any machine whose
headless DisplayServer reports no touchscreen. Godot on Android answers
yes, so case C ran on the phone; Godot on Windows answers no, so the test
bailed before reaching cases A and B and reported a failure that said
nothing about the game.
Case C now runs where a touchscreen is reported and is skipped where none
is, which leaves A and B - and the static guard on _note_keyboard's list
and touchscreen check - exercised everywhere.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Every other test script's .uid is tracked; this one was written after its
commit landed. Left untracked it would keep showing up in git status and
would be regenerated with a different id on another machine.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Aiming with a thumb on a stick is never exact, and at 950 units a couple of
degrees is a long miss. A shot fired roughly toward an enemy now goes at it.
28 degrees of cone, chosen so that pointing at something hits it while the
shot still visibly goes where the player pointed. Past about 45 the gun starts
picking its own targets, which is the behaviour #80 deliberately removed.
Enemies only, never asteroids. Snapping onto a rock while the player is trying
to hit the thing shooting at them would be the assist working against them,
and rocks are large and stationary enough to hit on purpose.
The aim guide is drawn through the SAME helper the gun fires through, so the
dotted line bends with the shot. A guide showing the raw stick heading while
the bolt took an assisted one would be the same class of mismatch #86 removed
between the guide's length and a bolt's actual reach - and this one would have
been harder to notice, because the line would look right most of the time.
Range 620 -> 950, the other half of the ticket.
Also swept the project root again: Godot keeps resurrecting pre-#84 copies of
moved files there from its own cache. None this time.
Suite 120/120.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
#85 - on a computer the sticks and the pivot pair are dead weight: WASD
steers, the mouse aims, Q/E pivot. All four go together the first time a
keyboard or a mouse is used, and only where there is no touchscreen. A touch
device with a keyboard attached keeps everything, because a finger is still a
way to play it.
#86 - the default gun shoots faster (500 -> 850) and further (320 -> 620).
The interesting half is that a bolt now STOPS where the gun's range ends.
Projectile lifetime is derived from fire_range at launch rather than set
independently, so the distance a shot covers and the distance the aim guide
draws are one number. They were two, and they disagreed: the guide promised
320 while a bolt actually flew speed * lifetime = 750, so shots landed on
things well past the line saying where they would stop.
#87 - the ship that fires throws sparks, so a shot is visibly attached to the
vessel it came from; with a whole fleet shooting there was nothing tying the
two together, which started mattering once the player had to aim. Enemy deaths
gained a shockwave ring and roughly double the debris - a kill used to read as
a small red puff that was easy to miss in a fight.
#88 - everything moves 1.5x faster: fleet cruise, every vessel's max_speed and
acceleration, and every enemy's. Scaling the enemies too keeps it a change of
TEMPO rather than a difficulty cut, and scaling acceleration alongside speed
keeps the handling from #77's tuning intact.
Three tests carried assertions that were really about one balance pass:
test_enemy_mothership pinned a literal 35.0 and now compares against the
chaser, which is what "never reaches ramming speed" actually means;
test_planet_atmosphere_transition flew far enough at the new speed to reach
the landed garrison, and now clears it, since that test is about the handoff
and not about a fight it never meant to pick.
Also repaired real data loss, caught only because test_gas_burner failed:
Godot had silently re-serialised GasBurnerBuildingDefinition.tres and dropped
collected_outputs, capabilities, map_radius and map_phase, and had recreated a
stale copy of it at the pre-#84 root path. Without collected_outputs no
transporter ever collects a burner's energy. Both files restored from git.
It had also resurrected pre-#84 SNAPSHOTS of Main.tscn, Planet.tscn and
lunar_surface.gdshader at their old root paths - the root Main.tscn still
carried 27 references to the flat layout. They are dead weight the engine
never loads, since project.godot and every preload point into the folders, but
they are exactly the kind of thing that wastes an hour later when someone
edits the wrong copy. Deleted, and the root is now down to ui_theme.tres,
icon.svg, web_shell.html and the two config files.
Suite 119/119.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Everything lived in one flat directory, so a diff gave no clue what area it
touched and finding the file you wanted meant knowing its exact name. Twelve
folders now, grouped by what a thing IS rather than what type it is - a
vessel's definition, its component scenes and its upgrade resources sit
together, because that is the set you read when you change one of them.
buildings core controllers enemies logistics planets
resources shaders surface ui vessels vfx world
332 files moved and every res:// reference rewritten - 306 files touched,
including the autoload and main-scene paths in project.godot. Nothing renamed:
this is purely where files sit, so the diff is a pile of pure moves and no
behaviour is buried in it. ui_theme.tres, icon.svg and web_shell.html stay at
the root, where Godot and the export preset expect them.
Verified three ways, because a suite alone would not have caught all of it:
the full 117 pass, the web export builds clean, and the game renders.
The export is why that matters. The FIRST export after the move failed with
"Attempt to open script 'res://GameState.gd'" - the old path - while
project.godot on disk plainly said res://core/GameState.gd. That is this
project's documented Android FUSE mtime trap: Godot served a cached copy of a
file whose content had changed. A `touch` fixed it. Nothing in the test suite
would have found that, because the suite runs from a copy on a native
filesystem - the one place the bug cannot happen.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
#80 - desktop combat. The mouse aims, the left button fires, and the number
keys pick the weapon: 1 the fleet's guns, 2 and up the abilities. The twin
sticks are built for thumbs; on a desktop they mean dragging a virtual stick
with a pointer that could simply BE the crosshair.
The pointer path hands its heading to FireStickController rather than driving
weapons itself, so exactly one thing in the scene aims a gun, and a held stick
wins - a finger on it is a deliberate act, a mouse resting somewhere is not.
Abilities gained fire_at(), which aims and fires in one call instead of
pretending to be a press-drag-release touch.
It is gated on the absence of a touchscreen, because Godot emulates a mouse
button from every touch and without that a tap on a phone would fire the guns.
The gate has an explicit override, and not for tidiness: headless reports a
touchscreen, so with the automatic answer alone the aiming could never be
exercised by a test and the only covered behaviour would be the guard.
#81 - the bottom-left control hint is gone, from both scenes and from
GameController.
#82 - a chaser is destroyed by the impact when it reaches a vessel. It used to
pass through the wreck untouched and line up another pass, which made a
ramming enemy strictly better than any other. Data-driven per enemy
(dies_on_contact), because a sniper that happens to drift into a ship has not
spent itself doing it - and routed through take_damage, so it dies exactly as
it dies to gunfire. The enemy is told about the impact BEFORE the ship
resolves its own damage, since resolving it can free the ship.
#83 - the research tree's buy button is gone; tapping a node twice buys it.
Deliberately "the same node twice" rather than a timed double-tap: no window
to miss on a slow phone, no accidental purchase while browsing, and the first
tap always does something visible, so the second is never a blind commitment.
The status line now says "Tap again to research" - with the button gone it is
the only thing distinguishing an affordable node from an unaffordable one.
test_enemy_collision asserted the exact behaviour #82 reversed, and now checks
the ram costs both. Suite 117/117.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
#77 - the Distorder becomes the Energy Propulser. Renamed everywhere the
player reads it, reskinned to match: a symmetric hull flaring into an engine
skirt instead of the lopsided warped lump, an engine bell with a thrust plume
instead of the wobbling ring glyph, and energy's own yellow instead of purple,
since energy is what it burns. The internal id stays "distorder" - it is the
capability string, the unlock key and the component scene name, and renaming
the wiring would be a large mechanical change with no player-visible effect.
Its speed ceiling now grows with the LOGARITHM of how many fly. Each one used
to add a flat 1.0, so a fleet that could afford six simply drove at six times
cruise and the vessel stopped being a choice. Base-2 makes the rule one
sentence - every doubling adds another 1.0 - and leaves two propulsers worth
exactly what they were worth before, so nothing already built is devalued.
#78 - a gas burner can be paused. It converts gas the player may want for
bombs into energy they may not need yet, and the only way to stop it was to
demolish it. The action is data-driven like every other: the visual's
ActionProvider lists "pause", and any converter can adopt it. Pause is the
burner's PRIMARY action, so it is one tap rather than a menu; the button shows
what it will DO, bars while running and a play triangle while stopped; and the
building itself goes visibly cold, because a paused building that looked
identical to a running one would be a setting to remember rather than see.
Conversion progress is left where it was rather than reset, so switching off
and on again does not discard a part-finished unit.
#79 - picking something up says so. A "+N" with the resource's own mark rises
and fades where it happened. Built from the SAME glyph the resource tables and
transfer rows use, so a material added to the catalog later cannot end up with
one mark in the HUD and a different one in the world. A pickup a full hold
REFUSES raises nothing.
Two Godot traps cost real time here and are now written down where the next
person will hit them:
- a `#` comment inside a .tres [resource] block is not a comment. It parses
as part of the preceding value and silently blanks the property after it.
One in DistorderVesselDefinition.tres left definition_id empty, which
surfaced as "<blank> did not resolve its definition" in a test far from
the edit. The note lives in VesselDefinition.gd now.
- a global class_name is not reliably resolvable, INCLUDING a script's own
inside its own static function. ResourcePickupVFX.new() failed to compile
when reached through a preview scene, taking Rock and the gas vents with
it. It has no class_name now and constructs itself by path.
Suite 114/114.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Both sticks float - the base jumps to wherever the finger lands, so a player
does not have to hit an exact spot on a screen they cannot feel. Releasing put
only the KNOB back onto the base and left the base parked wherever that touch
had been, so an idle stick sat off-centre in a different place every time and
read as a control still in use. Release now returns base and knob to the
middle of the rect.
Extractors produce at 0.5 ore per second, halved again: at 1.0 they still
outran the rest of the economy. An extractor now pays back its 25-ore cost in
fifty seconds rather than twelve.
The ship lab and the crafter were both overlapping themselves. Enlarging the
shared resource table to 136px (so it could breathe and show bigger type)
pushed its bottom edge 34px into the vessel row beneath it in one and the
status text in the other - my regression, visible immediately on a real
screen and not at all in any test. The lab's second row of upgrade cards was
also already running into the detail text below it, by 13px, before any of
this. Both menus are re-spaced end to end rather than nudged, since moving one
collision into another is how the second one got there.
Two tests carried waits tuned to an extraction rate that has now changed
twice. Both derive them from EXTRACTOR_RATE instead, so the next balance pass
does not report a broken extractor that is working perfectly - the same fix
test_extractor_income already needed one round ago, in the file next to it.
Suite 112/112.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Asteroids were pure hazard: the only thing one could ever do was cost you a
ship. They now take damage like anything else and break open, dropping ore the
fleet can fly over and collect - the one resource reachable in open space
without conquering anything, and something for the new fire stick to do out
there.
Two ways to destroy one, deliberately not equivalent. Shot apart yields the
ore; flown into does not - the rock shatters against a hull rather than being
mined, and a ship or a shield charge has just been spent on it. Without that
split a shield vessel could farm asteroids by ramming them.
Asteroid.take_damage() matches Enemy's contract exactly, so every weapon that
already knew how to hurt an enemy hurts an asteroid through the same branch
rather than a second special case: bolts, the laser sweep, and the gas bomb's
blast all pick both groups up together.
Dropped ore carries the asteroid's own drift, so it reads as debris rather
than as a spawned pickup, and it expires after a minute. Rock grew an optional
lifetime for that, defaulting to 0 (never) so hand-placed surface rocks are
unchanged - a planet surface is bounded and authored, open space is not, and a
field shot up and abandoned would otherwise leak nodes for the session. Rock's
header note that asteroids and rocks are never on screen together is no longer
true and now says so; the hexagon-versus-lump shape language matters more now
that they sit side by side, not less.
Five weapon tests silenced the asteroid field but never cleared what it had
already placed. That was harmless while asteroids were untouchable and is not
any more - two of them assert exact hit counts, and a drifting rock in the
beam would have made those fail intermittently.
Suite 111/111.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The fleet's guns picked their own targets and fired on their own, so combat
happened whether or not the player did anything - there was nothing to aim and
nothing to decide, and the right half of the screen held nothing at all.
Firing is now an action. VesselWeaponComponent no longer seeks a target; it
shoots along a heading, and the only thing that supplies one is the new
right-hand stick via FireStickController. A fleet arriving from another scene
starts with its guns silent, which is the correct resting state.
A subtle dashed guide runs from the fleet along the aimed heading while the
stick is held, out to the guns' OWN range - read from a weapon that is
actually aboard, so the line can never promise reach the fleet does not have.
Layout: the two sticks take the bottom corners, the pivot pair is centred
between them at the bottom edge, and the laser and gas bomb sit on an arc up
and left of the fire stick where a thumb already is. Those four are now round
- IconButton grew a round_frame mode, and the in-world action controls use it
while panel and menu buttons stay square, so the two kinds of button read as
different kinds of thing.
Worth recording, because it cost a screenshot to find: a script-declared
property set BEFORE the `script = ExtResource(...)` line in a .tscn is
silently dropped. Godot's parser has no idea the property exists yet. The
round frames were all correctly set in both scenes and none of them applied.
Also fixed researtch_lab_garrison.tres, which was authored under the void
profile - that is the black hole surface, so a lab garrison was spawning on
black holes.
Suite 110/110. test_enemy_combat's whole premise was auto-fire; it now holds
the real stick toward the enemy rather than poking weapon components, so it
exercises the path the player uses. test_fire_stick is new and covers the
change itself: an idle stick leaves a target parked well inside range
untouched, which is precisely what the old gun would never have done.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The fleet had a second of inertia in it: every vessel's acceleration sat at
roughly its own top speed, so flipping the stick while cruising took 0.58s to
take effect and let go coasted 222 units, while the formation anchor the ships
chase turns instantly. Acceleration is now ~3x top speed, chosen by measuring
rather than taste - past 3x the reversal time stops improving because
max_speed becomes the limit. Measured after: 0.18s and 23 units.
test_fleet_agility guards all three numbers and proves the bounds discriminate
by re-running the same measurement at the old value.
The laser reaches 1450 instead of 950 and strikes 38 units either side instead
of 22. Its drawn width is now derived from the hit width: the two were
independent numbers that had drifted, so the beam killed things it visibly
missed. LaserController's aim range also stopped hardcoding a second copy of
950. Tapping the laser button without dragging used to do nothing at all - the
press was claimed, the aim begun, then thrown away on release. It now fires at
the nearest enemy in range; dragging still aims precisely. That lives in the
shared aimed-ability controller, so the gas bomb gains it too.
Planets sat one per ~29 chunks - some 35,000 units of nothing between them.
Now one per ~12, measured over 14,641 chunks. The two thresholds moved by
different amounts deliberately: every SPACE-biome planet is a gas giant, so
lowering both equally would have tripled the gas-giant share as well. This
changes how often you find a planet, not what you find. Extractors produce
half as fast to match, so expanding is what grows the economy.
The research screen gave its tree 386px of 1280 and the text below it 520.
The text is now at the bottom, the tree has 550, and the resource table has
room and legible type - and its LOCAL column is no longer hidden behind the
close button.
Two map-editor bugs turned out to be one cause: a setup's id and its FILENAME
are not the same thing, and both save and delete rebuilt the path from the id.
ResearchLabSetup.tres holds setup_id "research_lab_1", so deleting it looked
for a file that does not exist and silently refused - which read as "the first
item in the list cannot be deleted" - while saving it wrote a second file
under the derived name and left the original behind, so the pool served the
stale twin. A third bug hid behind those: rescan() re-read through Godot's
resource cache and handed back pre-save values even when the file on disk was
correct. All three present to a player as "my change was not saved".
Research labs no longer grant a point of their own; what conquering one is
worth comes entirely from its authored map, as asked. The shipped lab map
carries 1 so the balance is unchanged - only where it is configured moved.
Suite 109/109. Four tests changed because the behaviour they pinned changed:
the joystick coast is now derived from cruise_speed^2/(2*acceleration) rather
than a constant tuned to one acceleration; the extractor wait derives from
EXTRACTOR_RATE; the research-point test asserts the delegation; and the black
hole test no longer assumes nobody will ever author a void-profile map.
test_planet_setups and test_planet_map_assignment stopped depending on which
maps happen to be in the content folder, which was making them fail when maps
were added or deleted elsewhere.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Construction already pooled the two; research, ship-lab upgrades and vessel
unlocks still charged the bank alone. That split was arbitrary from the
player's side - the resources are a couple of taps away from being banked
either way, so refusing a purchase over where they happen to be sitting was
ceremony. Worse, it was inconsistent: the same ore could buy a mothership and
not a skill node.
So the pooling moved down into can_afford/spend themselves, and the
construction-specific pair is gone. can_afford_resources now pools per
resource type, and a new spend_resources() charges bank first and hold for
the shortfall, settling affordability in full before taking anything - a
multi-resource price must not charge for the half it can cover and then fail.
unlock_vessel goes through it instead of withdrawing from the bank directly.
The skill menu's "you have %d" quoted the bank, which would have told the
player they could not afford something they could.
Ammo stays out of it. AimedFleetAbilityController reads the fleet hold
directly, and should: a gas bomb is fired from what the ships are physically
carrying, which is a fact about where the matter is rather than a convenience
about where the money is.
test_build_uses_fleet_hold became test_payments_use_fleet_hold and now walks
every route - skill tree, ship lab, vessel research, multi-resource
atomicity, and ammo staying unpooled. Two of its own setup steps were being
silently clamped by the fleet hold's capacity, which is what made an
affordable price look unaffordable; capacity is not what those checks are
about, so it is taken out of the way explicitly.
Full suite 106/106.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The setup editor and the landed surface drew different backgrounds for the
same planet. Each scene carried its own hand-written ShaderMaterial, and only
the game side ever swapped in a profile's shader, so a gas giant authored in
the editor showed recoloured craters while the game showed cloud bands. Both
now instance one SurfaceBackground.tscn and configure it through one method,
and every profile names its own shader - the old "null means keep whatever
the scene already had" convention was the drift, since each scene "already
had" something different. The shared material is resource_local_to_scene, or
two live surfaces would repaint each other.
The gas giant's bands broke into disjoint segments once per cell: the second
turbulence octave doubled only the INTEGER half of a lattice coordinate,
which is a different point in every cell rather than the same point at twice
the frequency. Measured, the old form steps 0.153 at a cell boundary against
0.0002 now. A second discontinuity went with it - the camera cell index
arrives pre-wrapped, so hashing it raw rearranged the whole sky every 32
cells; hashes are now wrapped per octave and the field is exactly periodic.
Bands also gained low-frequency octaves and per-row differential drift.
WorldMap swallowed every touch in _input(), which runs before the GUI system,
so CloseButton never saw one: the map could be opened and not closed. Android
and desktop hid it behind Godot's emulated mouse event, which this never
swallowed; a phone browser gives no such second chance. Its test had been
closing the map by calling the method the button was failing to reach.
Construction now draws on the fleet hold as well as the bank. With an empty
starting bank and ore bankable only at a mothership, charging construction to
the bank alone made the first mothership cost ore that could not yet exist.
Also: a conquered planet no longer re-spawns its garrison (guarded on having
capture points, since a setup with none registers as vacuously conquered);
planets keep the map they drew, with the candidate list sorted rather than
left in filesystem order and assignments pruned rather than cleared on
rescan; and the map editor lists setups at a tappable size and can delete
one, as an explicit armed mode rather than a tap away from destroying work.
Full suite 106/106.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The web export reached the main menu and went no further. Two separate bugs,
one hiding the other.
BUTTONS NEVER FIRED. A browser on a phone delivers the touch DOWN and never
the release. IconButton emits `pressed` only on release, deliberately, so
that dragging off cancels - so on the web it never emitted at all, on every
button in the game. fire_on_press (default OS.has_feature("web")) completes
the press there instead, and does not latch, since nothing is coming to
unlatch it. The cost is drag-off-to-cancel, which every other platform keeps.
Measured on the device rather than guessed: an on-screen probe reported
`touch=1 mouse=1 held=y` after one tap - one down, one emulated mouse down,
the button latched through ordinary GUI routing, then nothing.
THE ENGINE THEN TRAPPED. With the button working, pressing Play loaded and
built Main.tscn and the whole engine stopped mid-frame, page still alive.
An error overlay in a custom HTML shell finally showed why:
Uncaught RuntimeError: remainder by zero
An integer % 0. Wasm TRAPS on that; native builds silently tolerate it. So
the bug was equally present on Android and in all 100 tests, doing nothing
in both - which is why nothing runnable here could ever have caught it, and
why reasoning from frozen screenshots got nowhere for several rounds. Every
variable modulo is now guarded: the hangar grid's `i % columns`, planet
naming's `% names.size()`, and two in the map editor.
WHAT IS KEPT, AND WHY
- web_shell.html mirrors console errors onto the page, hidden until one
occurs. Godot's errors go to the browser console, which is unreachable on
a phone; this found in one shot what inference had failed to find all
evening. It costs nothing when there is nothing to report.
- MainMenu._start does load -> instantiate -> add_child by hand instead of
change_scene_to_file. That call does all three opaquely, so a failure is
indistinguishable from an unpressed button; split, each step reports.
- PlanetAtmosphereController defers the forced world populate out of the
frame that builds Main.tscn. It did not fix the hang - the trap did - but
25 chunks of asteroids and 49 of planets do not belong in the one frame
that already constructs 197 nodes, and nothing can paint until it ends.
Removed the direct-input-routing experiment and its test: it was written on
the theory that touches were not reaching Controls, which `held=y` later
disproved, and its set_input_as_handled() swallowed the very events the
probe was reading.
Full suite green at 100/100.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Every button in the web build was dead - reported as "Play does nothing".
Diagnosed on the deployed build with a temporary on-screen input probe,
because none of it reproduces here: the suite passes, and pressing Play under
the real gl_compatibility renderer (the same path WebGL2 takes) reaches
Main.tscn fine. The probe reported, from a phone browser:
#6 touch down @ 362,781
play rect 296,725 128x128
hit=YES
Three taps, six events, every one a DOWN and not a single UP - and the button
left drawn in its bright held state, pressed_down stuck true because nothing
ever came to clear it. So touches arrive, their coordinates are correct, and
they land inside the button; the browser simply never delivers the release,
in _gui_input or in raw _input.
IconButton emits `pressed` only on release, deliberately, so that dragging
off a button cancels it. With no release that means it never emits at all -
on every button in the game, not just Play. Which also explains why the
earlier suspects went nowhere: the scene change was never reached, so it was
never the scene change, and hit=YES ruled out coordinate mapping.
fire_on_press defaults to OS.has_feature("web"). There, a press completes the
tap itself and does NOT latch, since nothing is coming to unlatch it. Every
other platform is untouched and keeps drag-off-to-cancel.
The cost, stated plainly: a web tap can no longer be cancelled by sliding off
before lifting, and that applies to destructive buttons too. The alternative -
keeping release-firing and treating a new press as completing the previous
one - preserves cancellation but leaves the very first tap dead, which is
worse.
tests/test_icon_button_press_fires.gd drives the flag directly, since a
headless Linux run cannot BE a web build: one tap emits exactly once, a late
release does not emit again, the button does not latch, and with the flag off
the release still drives the press and dragging off still cancels.
MainMenu keeps the load/start bisection introduced while chasing this - a
failed change_scene_to_packed otherwise leaves the menu up, which looks
exactly like an unpressed button.
Full suite green at 100/100.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
DISTORDER STACKING. Each space-time distorder vessel past the first raises
the fleet's speed ceiling by 1.0, so a second one is worth crewing instead of
being dead weight. GameState gains max_fleet_speed_multiplier, pushed by
Fleet whenever composition changes - the same direction as every other
fleet-derived figure, since GameState does not reach into the scene tree to
count ships. The energy drain is normalised against the CURRENT ceiling, so
full boost costs the same per second however wide it is: extra vessels buy
more speed, not cheaper speed. Losing one re-clamps an over-ceiling boost
rather than leaving the fleet travelling faster than its crew can sustain.
TRANSFER MENU. Rows overlapped, and the last one ran into the hint text.
row_height is divided from the available space, but each row's contents used
a fixed pixel ladder - glyph at +10, buttons at +88, selector at +142 - which
needs about 244px. Three resources fit; the fourth cut rows to 180 and every
one spilled 64px into the next. Every offset is now a fraction of the row, so
it cannot overflow. Also dropped the per-row AMOUNT caption (four copies of a
word the footer hint already explains) and moved each glyph and name to the
TOP of its row, where they read as its heading rather than the next row's.
HANGAR. Six vessel types in one row gave each about 100px for a 155px label,
so the names ran together. Now a 3x2 grid per section, sized from the space
between the first heading and the status line, using the empty lower half.
LAB RESEARCH REWARD. A harder enemy lab is worth more research points, and
how hard a lab is depends on the layout authored for it - how many capture
points, how many defenders - which the planet TYPE cannot know. PlanetSetup
carries research_point_reward, set in the map editor, overriding the planet
type's flat rate. 0 means "no opinion", so every map authored before this
keeps its old value.
RESEARCH LAB BACKGROUND. Its own shader instead of the crater field with a
red tint, which read as "rocky planet, tinted" rather than someone's
facility: plated floor, seams, service conduits, lit junctions. Loop-free by
construction (#68's lesson applied up front) - conduits stay continuous
across cells by keying a horizontal run to cell.y and a vertical to cell.x,
so neighbouring cells agree without any cell scanning its neighbours.
ENEMY DETAIL. Every enemy was a single flat outline, which left the artillery
and the turret spawner as the same octagon at two sizes - the silhouette
carried no information about what the thing did. EnemyDefinition gains
detail_lines and detail_rings as DATA, so adding an enemy still means adding
a definition rather than editing Enemy.gd, and the four non-default hulls are
redrawn: a long-barrelled sniper, a squat mortar, a hive with launch bays, a
capital ship with spine and ribs.
Three things drew an enemy from its definition - the enemy, the map editor's
markers, and the palette buttons - each reimplementing it from hull_points
alone, so the new detail appeared in the game and nowhere else and an editor
marker stopped matching what it placed. They now share
EnemyDefinitions.draw_definition().
ARTILLERY. The 1.8s charge-up before the tower fired arrived long before the
threat was legible and mostly read as the tower being stuck. It fires
immediately now; the warning is the shell you can see in the air plus a small
cross on the ground at the impact point. Deliberately not a blast-radius ring
- a large translucent area was easy to read as the explosion having already
happened, where the cross says "move off this spot".
MAIN MENU. No Quit in any build (a web build has nothing to quit to, and
Android's back gesture already does it), a plain black background, and the
game is "Spaceship" - in the title and in config/name, which is what drives
the browser tab for the web export. change_scene_to_file() returns an error
and otherwise fails silently, leaving the menu up, which is indistinguishable
from an unpressed button; it is now checked and reported on screen.
Full suite green at 99/99.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Adds a Dev Mode entry to the main menu and takes the play-testing
conveniences out of a normal game, which turn out to be the same change
from two directions.
DEV MODE. Banks 2000 of every resource, grants 500 research points, and
plants one of each planet type in the chunks around the spawn. Built ON TOP
of reset_new_game() rather than beside it, so a dev game is a normal game
plus a stock rather than a second definition of "fresh game" to keep in
sync. Offered only when running from Godot - OS.has_feature("editor"), the
tag an export clears - as is the map editor. Hidden AND guarded: pressing
either anyway is refused, because visibility is presentation and the rule is
the rule.
The menu now lays its stack out in code. Hiding two of four entries with
scene-authored offsets would leave holes in a shipped build - the same
failure the HUD columns had - so it measures the entries that actually exist
and centres on them.
A NORMAL GAME NOW HANDS OVER NOTHING. No 500 banked ore (removed from both
reset_new_game and the autoload's own default, so the two cannot disagree),
and no gas giant guaranteed in the origin chunk. Both were play-testing
conveniences that had become part of the real world's shape: every player
got the same landmark in the same place. Play-testing wanted all four planet
types anyway, and a normal game wanted none of them - different needs, now
different mechanisms.
WorldGen cannot read GameState back (an autoload is not resolvable as a
global identifier that early, the hazard PlanetSetups.gd already works
around), so GameState.dev_mode's setter PUSHES the flag down. World
generation stays static and dependency-free, GameState stays the single
source of truth, and dev planting is applied after every rng draw so a dev
world and a normal one differ only in those four chunks.
FLEET SIZE. BASE_FLEET_SIZE was doing two jobs - how many slots exist AND
how many ships you start with - which forced a new game to open with a full
fleet, leaving the first ship crafter nothing to do. Split into
BASE_FLEET_SIZE (3 slots) and STARTING_SHIP_COUNT (1 vessel), so the opening
has two empty slots to fill. A third copy of the old value turned up in
PlanetAtmosphereController, hardcoded as spawn_ships(5) and silently
degraded into "however many slots exist" once max_fleet_size() clamped it.
The doctrine branch now runs 3 -> 10 slots. Wing (30 -> 15) and Squadron
(70 -> 35, 2 -> 1 points) are cut hardest: they are no longer a mid-game
luxury but the ordinary way out of the opening, and pricing them as a luxury
made the early game feel stuck. Armada follows at 90/2. Two new tiers,
Battlegroup and Grand Fleet, grant +2 each - seven single-vessel nodes in
one lane would be a long monotonous chain for a branch that only ever does
one thing. DEFAULT_FORMATION grows to 10 slots, because max_fleet_size()
clamps to its length and a doctrine past that would charge real resources
for nothing; test_fleet_growth.gd asserts that invariant directly.
TEST FALLOUT, all of it from the deliberate change rather than a defect:
- Tests that silently inherited the old 5-ship default now ask for the fleet
they need. Several had to crew it from their first frame rather than
_initialize(), where the engine's deferred tree-entry finalization undoes
it - the same hazard those tests already documented for set_process(false).
- test_planet_landing_proximity and test_return_position_matches_planet take
the FIRST chunk holding a planet and land there. The guaranteed gas giant
had been masking that this can now be a black hole, which destroys ships on
contact. They exclude it, and clear the asteroids the field already placed.
- test_pivot_controls measured rotation with angle_difference, which wraps at
+/-pi, so a hold that turned a full circle read as no rotation. Latent
flakiness that only ever showed under load; it now reads the controller's
unwrapped pivot_hold_amount.
- test_research_points was failing on ore, not points: can_unlock() checks
the ore price too, and a fresh game banks nothing now.
- test_gas_system's "a reachable gas source near the spawn" premise moved to
dev mode, which is where that guarantee now lives.
Full suite green at 97/97.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Seating the HUD columns below the resource readout moved the build button
from the scene-authored 160 to 288. _layout_context_buttons() then wrote the
action row straight back to a hardcoded 324 on the next context change - and
it runs on the first frame, so the action button overlapped the build button
immediately.
Same shape of bug as the overlap that seating fixed: a literal coordinate
that meant "directly below the thing above me" and silently stopped being
true when that thing moved. It now measures instead:
row_top = build_button.position.y + build_button.size.y + HUD_ROW_GAP
Both the action button and the expanded context row derive from that, so
wherever the build button ends up, the row below follows.
tests/test_hud_layout.gd asserts the RELATIONSHIP rather than any
coordinate, so a freshly hardcoded y fails whatever value someone picks:
every HUD button clears the readout's real bottom, and no two
simultaneously-usable buttons overlap - checked at load, collapsed and
expanded, which is the path that broke. BuildingActionButton and
HangarButton are excluded as a deliberately exclusive pair.
Verified to fail against the previous code with the reported symptom:
build_button y288..436 overlapping action_button y324.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The build button overlapped the resource list. ResourceReadout builds one
row per resource type AT RUNTIME (58px each), so how far down it reaches is
not knowable when the scene is authored: PlanetSurface/Main declared
offset_bottom 154, four resources actually reach 270, and BuildButton sat
at a hardcoded 160. A 110px collision that grew with every resource type
added. Because the readout spans the full width - the fleet hold column is
at x 356 - it hit both button columns, TransferButton and
EnterAtmosphereButton as well as the left one.
Rather than nudge the offsets, the layout now derives from the readout:
ResourceReadout.rows_height()/content_bottom() report the real extent and
_ready() stops the node understating its own size, and
BuildController._seat_hud_columns() seats both columns below that at load.
Adding a fifth resource type pushes the buttons down on its own, so the
overlap cannot come back.
Two details worth keeping: the height is derived from the live resource
list rather than cached in _ready(), because BuildController readies before
UI's children in scene-tree order and a cached value would answer zero. And
it translates the block rather than re-flowing it, since
BuildingActionButton and HangarButton deliberately share a row as mutually
exclusive options.
The planet name moved from the centre - where it sat across resource rows
3 and 4 - to the top-right corner, right-aligned. Right alignment is what
keeps it clear: short names sit against the right edge, well clear of the
fleet hold column, and long ones wrap downward into empty space rather than
growing left into it. That scene edit rode along in the #68 commit, both
being changes to PlanetSurface.tscn.
Also adds three .uid files orphaned by earlier commits this session; the
repo tracks them for every other script.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Rotating the fleet and spamming a direction let vessels fall out of sync
and fly at wildly high speeds. The pivot cooldown added earlier throttled
how often the exploit could be triggered; it never touched the cause.
Fleet.stop() did `ship.velocity = ship.velocity + ship.fleet_velocity`,
compensating for the fact that Ship.velocity read zero during cruise. The
next steer refilled fleet_velocity, so every press-release cycle added a
whole cruise velocity to the ship's own, decaying only at `acceleration` in
between. Tapping faster than that bled it off stacked them without limit,
and ships desynced because each bled off at a rate set by its own distance
to its slot. Measured before this change: spamming a direction while
pivoting flew the fleet 32737 units in 3s against 784 for simply holding
the stick, with a vessel reaching 48x its own top speed.
The compensation existed because `velocity` was not the ship's real motion:
Ship snapped to its slot every tick and zeroed it, so the truth lived in a
second field and anything wanting it had to add the two together. Fixed at
that root rather than by clamping the symptom:
- fleet_velocity is now fed FORWARD into the steering instead of carried
alongside it. The ship aims to travel at its slot's velocity plus a
correction closing the gap, so holding formation costs no error and
`velocity` is always true world motion.
- The snap only applies when the slot is standing still, which is the one
case where converging asymptotically would leave a residue. While the
slot moves the ship steers continuously - a snap there was the teleport
that made `velocity` lie in the first place.
- A hard bound on the INTEGRATED result every tick, not merely on the
request, so no sequence of commands can leave a vessel over its own top
speed.
- Fleet.stop() injects nothing; it reports that the slot has halted.
tests/test_fleet_speed_bounded.gd covers it, and fails against the previous
code with the numbers above: no vessel exceeds its own top speed however
hard the direction is spammed, spamming cannot cover more ground than
holding the stick, both hold while the formation is pivoting, and the fleet
stays together.
test_space_time_distorder.gd now clears the asteroid and enemy groups at
setup, the same clean-slate idiom test_ship_lab.gd uses. Silencing the
spawners was never enough - what Main.tscn had already placed kept moving,
and losing the distorder vessel to a rock resets the multiplier correctly,
which the test then reported three stages later as a failed scene swap. It
also checks it still crews a distorder vessel before relying on that.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
The crater shader ran on every pixel of a fullscreen rect every frame and
was by far the most expensive thing in the surface scene. Nothing else came
close: RockField keeps 14 rocks alive, Rock/Ship/Enemy already avoid
per-frame queue_redraw() for transform-only changes, and LunarBackground.gd
is a handful of scalar ops.
Three reductions to the field itself:
- 5x5 -> 3x3 neighbour scan. That width was never the geometric answer;
it was insurance against craters seen clipping at cell boundaries
on-device, i.e. a crater reaching past its nominal radius because a hash
returned a value outside [0,1] and extrapolated mix(). Clamping the
crater CENTRE into its own cell, alongside the radius clamp already
there, bounds the reach whatever the hash returns - so a crater two cells
out is >= 1.0 cells away, a normalized distance of >= 2.63, past every
threshold the shading reacts to (the last saturates at 1.15). The outer
ring provably cannot change a pixel. This removes the clipping bug's
cause instead of hiding it behind 16 extra samples.
- One hash42 per cell instead of hash21 twice plus hash22, decorrelating
on output lanes rather than input offsets.
- Squared distances in the loop, square-rooted twice at the end rather
than once per sample.
Measured with the new tests/profile_shader.sh, above softpipe's own ~657ms
fullscreen floor: 9482ms -> 405ms of shader work, a 23x reduction.
Then the larger change: the field is a fixed function of world position, so
every frame recomputed an identical answer. It is now made periodic (the
cell index is wrapped before hashing, which makes a Worley field tile
exactly rather than merely blend), baked once at load into one tile, and
scrolled with repeat wrapping by lunar_tiled.gdshader - one texture fetch
per pixel. Two honest costs: the field repeats every 7040 world units, and
a fixed-resolution tile softens when zoomed past 1:1. Lunar regolith is
forgiving of both, having no landmarks and being smooth gradients already.
The bake waits a frame and asks the material whether it actually samples a
field_texture, so a gas giant - which swaps in its own background shader
(#51) - does not pay for a 2048-square tile it will never read. Headless
runs skip it entirely, having nothing to rasterize.
tests/test_lunar_field.gd guards the reasoning, since a fragment shader
cannot be asserted on directly: the premises are still in the source, the
worst-case ring-2 geometry stays clear of every threshold independent of
the hash, a 3x3 and a 5x5 scan render identically over 4000 sampled points,
the field repeats exactly in both axes, and the miss sentinel still cannot
overflow mediump into +Inf and then NaN.
test_surface_profiles.gd asserted "a rocky landing leaves the background
alone" by naming lunar_surface.gdshader; it now reads the shader from the
scene, which is the property it actually cares about.
PlanetSurface.tscn's edit also carries the planet-name relocation described
in the HUD commit - same scene file, so the two cannot be split.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Strayed 2.6x over its old allowance (77.7 vs 30.0) under one heavily
loaded JOBS=1 full-suite run, then passed cleanly in isolation
immediately afterward - device jitter, not a regression. The wider
margin still leaves orders-of-magnitude headroom below what a real
formation-desync bug (#72) would actually produce.
#64: Load used to blindly cycle to "the next" saved setup with no way
to see what existed or jump straight to one by name. It now opens a
real selectable ItemList of every saved setup (name + surface profile),
tap one to load it.
#65: a single-finger drag over empty canvas could never pan - the
touch handler placed an item on PRESS, before a drag had any chance to
happen, so the "pan" branch below it was unreachable dead code (it only
fired when selected_palette == "", which nothing in the UI can ever
produce). Placement now waits for release: released within
PAN_DEADZONE of where the finger went down is a tap (place), crossing
it partway through is a pan - the same tap-vs-drag deadzone idiom
GasBombController already uses for its own aim gesture.
#66: added a persistent "Selected: <kind> #<n>" label, refreshed
wherever selected_item changes (touch-select, new/load a setup,
delete), so a tap actually confirms what got picked instead of the
player inferring it from which dot on screen turned white.
Also widened test_space_time_distorder.gd's formation-lag margin after
it strayed 2.6x over its old allowance under one heavily loaded
JOBS=1 full-suite run and passed cleanly in isolation immediately
after - device jitter, not a regression; the wider margin still leaves
orders-of-magnitude headroom below a real desync bug's stray distance.
#74: RESEARCH_LAB_CHANCE raised from 0.08 to 0.16 - research points
gate mid/late skill-tree nodes and vessel/lab research now (#63), so
0.08 was too scarce for how central they'd become to progression.
#75: an extractor placed on a research_lab planet would sit on an
empty resource id forever, since it yields nothing by design. Refused
in BuildController's placement confirm, with the same "explain the
refusal" idiom every other build gate uses. register_capture_points()
gained an optional definition_id parameter so the refusal/reward logic
can tell a research_lab planet apart without threading an extra
argument through every existing caller.
#76: conquering a research_lab awarded a research point with no
on-screen acknowledgement beyond a HUD number the player has no reason
to be watching at that exact moment. CaptureLabel's existing
"CONQUERED" flash now names the reward directly on a research_lab.
test_extractor_building.gd and test_planet_conquest_gate.gd both use a
"first planet chunk found" scan with no regard for definition_id; in
this world's fixed seed that chunk happens to be a research_lab, so
#75's new gate started refusing their otherwise-unrelated extractor
placement checks. Both now skip a research_lab chunk when picking
their target, same as test_research_lab_extractor_gate.gd does on
purpose for its own dedicated coverage of that refusal.
_finish_pivot_tap() granted a full PIVOT_MIN_STEP rotation on every
button/key release with no rate limit, so tapping fast enough rotated
the formation far beyond PIVOT_SPEED's smooth held-rate cap - well
above what holding the button down can ever produce. Each of those
artificially-frequent re-seats (Fleet.apply_formation()) forced every
ship into a fresh max-speed catch-up dash toward its new slot with no
time to settle between dashes, and since outer-wedge ships swing
further per pivot than inner ones, spamming made them visibly diverge
in speed while each ship's own bounded max_speed made the fleet read
as constantly pinned at its speed ceiling.
Added a per-direction cooldown (the time a genuine hold needs to reach
PIVOT_MIN_STEP on its own) so tap-spam can't out-pace holding the
button. Deliberately NOT shared between CW/CCW: a quick correction tap
the other way right after an overshoot is normal play, and
test_keyboard_input.gd already covers Q and E each independently
producing their own minimum step back-to-back - only repeatedly
re-tapping the SAME direction needed capping.
Chasers (EnemyPursuitComponent) and snipers (EnemyStandoffComponent)
used to beeline/close in from anywhere on the map the instant they
spawned - no matter how far the fleet was, which read as omniscient
rather than alert. Both now sit idle until the fleet is within their
own detection_range. Artillery's slow drift is gated the same way, but
reuses its own existing fire_range rather than a new tunable - outside
its firing envelope it has nothing to be drifting toward yet.
The enemy mothership deliberately keeps the opposite behaviour: it is
meant to read as a persistent, unavoidable stalking threat, not a
local sentry, so it gets its own EnemyPursuitComponentMothership.tscn
variant with an effectively unlimited detection_range - the same
scene-variant pattern EnemySpawnerComponentChaser/Artillery already
use for per-definition property overrides on a shared component script.
int_speed/shd_speed granted +max_speed/+acceleration per level to a
single vessel, which is redundant now that #59/#62 gave the whole
fleet a shared speed fader instead - a per-ship speed stat has nothing
left to meaningfully affect since #59's multiplier is applied at the
point of use in Ship.gd regardless of the ship's own base stats.
Deleted both upgrade resources and their vessel-definition references
outright rather than leaving them unused.
test_ship_lab.gd's per-vessel-type isolation check used int_speed as
an arbitrary interceptor-only upgrade to prove it doesn't leak into
transporter stats - swapped to int_range, which still exists. Its
upgrade-card-count assertion updated for 3 remaining interceptor
upgrades instead of 4.
SpeedFader was missing anchor_top/anchor_bottom = 1.0 in both Main.tscn
and PlanetSurface.tscn, so it rendered TOP-anchored at a fixed y while
its siblings (LaserStatus, GasBombStatus) are bottom-anchored - on a
720x1280 viewport both landed in the same screen region, directly on
top of each other. Properly bottom-anchored SpeedFader now, positioned
above both status labels instead.
Separately, the ship lab's 7 vessel-name labels (#57-#59 grew the
roster from 4 to 7 without widening their row) were laid out 120px wide
on only a 90px pitch - a fixed 30px overlap between adjacent labels
regardless of text length, worse for genuinely long names like
"Interceptor"/"Transporter". Narrowed each label to fit its own lane
with real gaps, dropped the font size, and enabled word-wrap; shifted
the upgrade-card rows below down to give the now-taller label row room.
Squadron Doctrine and Power Systems both sat at depth 3, x=0.50,
landing on the exact same screen position - Squadron's own
prerequisite chain (coordination -> wing -> squadron -> armada) zigzags
x from 0.50 to 0.82 back to 0.50, colliding with Freight's branch at
that depth. Kept the whole fleet-size doctrine branch at x=0.82 (wing's
own lane) instead, so it runs straight.
Added a structural regression test: no two SKILL_TREE nodes may share
the same (skill_depth, x), computed the same way SkillTreeView itself
places nodes.
GameState._advance_distorder_drain() runs in GameState._process() -
not the distorder component's tick() - alongside extractor mining and
production, so it (and the fleet's travel) keeps working across the
space<->surface scene swap. Drain scales with how far above 1.0 the
multiplier is: (fleet_speed_multiplier - 1.0) / (MAX - 1.0) *
ENERGY_DRAIN_PER_SECOND (2.0 at full boost), accumulated fractionally
the same way extractor mining_progress is so small multipliers still
drain a nonzero amount instead of rounding to zero every frame.
Running the fleet hold's energy to exactly zero force-resets
fleet_speed_multiplier to 1.0 and raises a one-shot
boost_depleted_notice flag, since GameState can't reach into whichever
scene's SpeedFader may or may not currently exist. SpeedFader.gd
consumes the flag on its next _process(), snaps its knob down (it was
already reading the multiplier live, so this falls out for free) and
shows "Out of energy!" - also clearing active_touch_index so a
still-held drag can't immediately re-boost from the position the
finger never moved away from.
test_space_time_distorder.gd (#59) updated to stock fleet energy
first, for the same reason #60/#61 needed it: boosting is no longer
free, and that test's own multi-second boosted phases would otherwise
get cut short mid-measurement by running out.
Mirrors #60's gas bomber ammunition exactly: ENERGY_PER_SHOT (8) drawn
from GameState.fleet_inventory before consuming the firing ship's
cooldown, the button's enabled rule requires affordability too, and an
unaffordable-but-otherwise-ready press still explains itself through a
new LaserStatus label rather than silently doing nothing.
Added two small generic GameState helpers this needed:
deposit_fleet_resource()/deposit_bank_resource(), counterparts to
#60's withdraw_fleet_resource() for resources (energy) with no
dedicated fleet_cargo/fleet_gas_cargo-style property. Along the way,
confirmed deposit_resource_to_building() on a "bank" building deposits
into that building's own per-record inventory, NOT bank_inventory
itself - only building_resource()/building_inventory() special-case
the "bank" capability to redirect there.
test_laser_vessel.gd (#58) updated to stock fleet energy first, since
shots are no longer free - same fix#60 required for the gas bomber's
own pre-existing tests. Also fixed test_laser_vessel.gd's own instance
of the bare-ClassName.CONST script-corruption trap (documented in
#60's commit): var skill: LaserController and a bare
LaserController.BEAM_HALF_WIDTH reference, mirrored into a local const
instead.
Shots cost GAS_PER_SHOT (5) from GameState.fleet_inventory, not the
bank - the fleet has to be carrying its own ammunition. The button's
enabled rule now also requires affordability, and a press on a ready-
but-unaffordable bomber still consumes the touch to explain itself via
a new status label (GasBombStatus, IconButton's own "greyed but still
pressable" convention) rather than either silently doing nothing or
consuming the bomber's cooldown for a shot that never fires.
Also documents a real, reproducible GDScript trap hit while writing
tests/test_gas_bomber_ammo.gd: a bare `ClassName.SOME_CONST` reference
anywhere in a -s entry test script can corrupt method resolution for
the WHOLE script, making an unrelated, earlier `some_node._process()`
call fail with "Nonexistent function '_process' in base 'Nil'" despite
the node being valid and the method existing (confirmed via
has_method()). Root cause unconfirmed; fixed by mirroring the constant
's value into a local const instead of referencing the class directly,
in all three gas-bomber tests.
tests/test_gas_bomber.gd and test_gas_bomber_cooldowns.gd updated to
stock fleet gas first, since shots are no longer free.
