BuildingField.place() dropped a structure at the fleet centroid with no
overlap check of any kind, so they could be stacked.
It reads worse than it looks. BuildingField decides which structure the
fleet is standing on by testing the centroid against interact_radius and
returning the FIRST match, so two structures inside one another make that a
question of list order rather than of where the player flew - you could fly
to your workshop and open the crafter.
The separation is interact_radius itself (150) rather than a new number,
because that is exactly the circle the ambiguity comes from. Buildings draw
to a radius of roughly 60, so this leaves a visible gap rather than letting
them touch.
Gated in three places: the info popup (the button greys out and says which
structure is in the way), _confirm_build, and _try_place - the preview goes
stale as the fleet keeps moving.
Ordering matters in one case. A unique structure replaces itself by
demolishing the old one BEFORE placing the new one, so a placement refused
afterwards would leave the player with neither. The check runs before the
demolition and excludes the structure being replaced. Case D pins it.
Two existing tests encoded the old permissive behaviour and now fly clear
first; neither is about placement:
test_build_info_popup builds every menu option from one position, so
each structure blocked the next
test_payments_use_fleet_hold builds a mothership on the ship lab that
an earlier case injects at the origin
Also fixes a bug from 37e2e89: Sfx.warm() called add_child() from _ready(),
which Godot refuses outright - it printed an error and never pre-built the
pool, defeating the point of warming it. Deferred, like Music.ensure().
173/173.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019AJMdB7iuHuLFZM1aBu3GD
Reported from the game: sounds sometimes start a bit late. The samples were
the suspect, so they were measured first - and they are clean. True digital
silence at the head of each file:
seven of nine start on sample 1
sfx_gas_blast 0.2ms
sfx_black_hole 1.5ms, then a deliberate swell
The black hole's slow attack is the sound it was written to be - make_sfx.py
says it "swells across most of its length instead of decaying from the first
sample" - so there is nothing to trim without breaking it.
The delay was in the playback path, and "sometimes" was the clue:
play() called load() every time. A cache hit after the first, but the
FIRST play of each sound was a synchronous read and decode on the frame
the sound was asked for. Ten sounds, ten occasional late starts.
_pool() built its twelve AudioStreamPlayer nodes lazily, on the first
play() call - one bigger stall, every session.
The catalog is preloaded now, so play() never touches the disk, and
Sfx.warm() builds the voices next to Music.ensure() at startup.
Case F pins the thing that was originally suspected, since make_sfx.py
regenerates these and a lazy envelope would hide the delay somewhere nobody
would look. It reads the source .wav's RIFF chunks rather than the imported
stream: these import as QOA (compress/mode=2), so AudioStreamWAV.data is not
PCM. The first version of that check was vacuous for exactly that reason and
passed at a 0.5ms threshold; it now reports "9 samples checked, worst attack
sfx_black_hole.wav at 1.5ms" and fails when tightened.
Not touched: project.godot has no [audio] section, so output_latency is at
its default. On Android that buffer is usually the dominant latency, but
lowering it risks crackle on real hardware - worth doing with someone
listening, not blind.
172/172.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019AJMdB7iuHuLFZM1aBu3GD
The repo tracks a .import beside every other .wav; the four #208/#210/#213/
#214 sounds were committed without theirs, so a fresh checkout would import
them from scratch and a web export would carry different resource ids than
the tree it was built from.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019AJMdB7iuHuLFZM1aBu3GD
#204 the Beam Collimator widens the beam instead of adding range to one that
already crosses the screen; beam width becomes a per-shot vessel stat.
#205 the Energy Propulser gains three upgrades - dash recharge, dash reach,
and boosted-cruise drain - which turns three constants into vessel stats.
#206 three further doctrines carry the fleet from ten to twenty.
#207 the prologue's World Ship gets a marker, through the scanner's own
guidance arrow rather than a second one.
#208#210#213#214 four new synthesised sounds: the breach opening, cargo
reaching the hold, a rock breaking, and a gas bomb blooming.
#215 building health bars are larger, plated and framed.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_019AJMdB7iuHuLFZM1aBu3GD
"Shot sounds sound like a toy" and "death sounds from ship are really
cheap", and there was no arguing with either. The shot was a bare
descending sine, which is the sound every toy ray gun has ever made. The
death was one band of filtered noise with a sine under it - thin, and over
before the eye had finished with the explosion.
The synthesis is rewritten rather than tuned, because the shape was the
problem:
LAYERS, not a waveform. A bang is a crack, then a body, then a tail, and
they have different lengths. One envelope over one source cannot be any
of those things, so the explosions now run three clocks - the crack is
done in a fifth of the length, the roar runs most of it, and sparse
debris keeps falling afterwards so there is an aftermath rather than a
stop.
A SUB. Below about 80Hz is where "big" lives; without it an explosion is
a hiss however loud you play it. Measured: 74% of a ship death's energy
and 86% of a structure's now sits under 200Hz, against almost none
before.
SATURATION. Soft-clipping folds in harmonics, so a sound reads as loud
rather than merely being loud - and stops peak normalisation flattening
everything into the same quiet mush.
DETUNED PARTIALS on the shot. Three of them at deliberately wrong ratios
beat against each other and read as a mechanism; the single sine was the
whole of the toy problem. A noise transient and a sub thump under it give
it something to hit with.
Lengths went up with the weight: the ship death from 0.85s to 1.45s and the
structure from 1.6s to 2.3s, both audible to the end rather than gone in the
first third.
SPACE_4.wav is the source the shipped theme is made from, so it belongs
under version control - but not in the game.
It lives in audio/masters/ with an empty .gdignore beside it, which makes
Godot skip the folder outright: nothing in it is imported into res://, and
nothing in it is carried into an export. Left where it was, in the project
root, its 59MB would be imported and then packed into every Android and web
build for a file nothing opens - the game loads audio/space_4.mp3, the same
5m36s at 6.7MB.
Its .import file goes with it, since there is no longer anything to import,
and the test runner stops copying the folder for the same reason it already
skips android/ and builds/: 59MB copied on every single run, for a folder no
test can even see.
README.md next to it records which shipped file each master became and the
command that made it, so the pair cannot drift apart silently.
Three events get a sound: a shot leaving a gun, a vessel dying, a structure
coming down.
The sounds are SYNTHESISED rather than downloaded - audio/make_sfx.py
generates all three from scratch, and is committed alongside them. The
ticket asked for free sources and the good ones are in its comments
(Kenney's CC0 packs, OpenGameArt's CC0 sets, Freesound filtered to CC0),
but a generated sound has no licence to track, no attribution to keep, and
no binary in the repository whose provenance nobody can check a year from
now. It also makes the sounds DATA: too sharp a shot is a number in that
file and a re-run, not a hunt through a library. Dropping in a pack file
later means replacing a .wav and nothing else.
A voice POOL, not a player, because these overlap constantly - a fleet
firing is several shots inside one shot's length. Twelve voices, and when
they are all busy the one furthest through its own sound is recycled, which
is the one with the least left to lose.
Parented to the tree root, for the reason that matters: every one of these
sounds is started by something that is being freed on the same frame. A
dying ship cannot own the sound of its own death.
Levels are per sound rather than shared - a shot fires constantly and a
death happens once, so equal volumes are not proportionate ones - and shots
get a little pitch spread so a stream of them does not flatten into one
machine-gun note. The deaths get none: a wobbling explosion sounds broken
rather than varied.
The test goes at the failures a play-through would hide for weeks: firing
three times the pool size must leave exactly the pool's worth sounding and
no leaked voices, and the ship-death case is checked from a node that really
has been freed on that frame.
"All the time" is the whole of the requirement and it decides the shape.
The theme has to survive the main menu becoming the game, the game becoming
a planet surface, and the surface becoming the game again - so the player
cannot live in any of those scenes, because change_scene_to_file() frees
whichever one is current.
It is parented to the tree root instead, beside the GameState autoload,
which is the same trick for the same reason. Music.ensure() is idempotent by
node name, so each scene asks for the music on the way up and gets the one
already playing rather than a second copy over the top of it. Two call
sites cover everything: MainMenu, and GameController - which is the root
script of BOTH the space scene and the planet surface.
Done this way rather than as an autoload deliberately: project.godot carries
the developer's own uncommitted window settings, and adding music should not
drag those into a commit.
It is also PROCESS_MODE_ALWAYS. Since #128 a dialog stops the tree, and a
soundtrack that cuts out every time someone reads a line is worse than no
soundtrack.
The theme ships as MP3 rather than the 59MB WAV that was sitting in the
project root. Same 5m36s, 6.7MB, encoded with LAME at 160k - 89% smaller,
which is the difference between a repository that carries its music and one
that carries a WAV forever in its history. The original file is untouched
and still there, untracked, so nothing is lost.
The test checks CONTINUITY rather than "a sound started": the same player,
never a second one, through two scene changes and a dialog.
