A laptop with two graphics chips hands an application the integrated one
unless it says otherwise: this machine has a GeForce RTX 5060 and was
drawing on a Radeon 780M. libglvnd makes the choice when it first
resolves a gl entry point, which is after main starts, so setting the
environment here is enough - no wrapper script and no launcher. Names
already set by hand are left alone, and -integratedgpu undoes it for
anyone who would rather keep the battery.
The tiles moved out of the scenery's pak into an archive of their own: a
sorted index of fixed-size entries found by where a tile sits in the
quadtree rather than by a name, each entry compressed on its own so a
tile can be read without touching the rest. A name apiece would have been
seventy megabytes of index to read before the first tile could be drawn.
An entry carries what its tile's ground asks for - how far it departs
from the finest mesh, and how long its triangles are - because which
level to draw has to be settled for tiles that are not on the card yet,
and the index is the only thing about them that is read up front. Until
now the figure came from the tile's level, which is the worst tile in it,
so one steep corner spoke for every tile that shared its level.
The mesh carries a normal per vertex, octahedron-encoded into two bytes,
in place of the face normal the fragment stage worked out from how the
view position changed across a triangle. A face normal makes the ground
read as facets, and a coarser level's facets face differently from the
ones they stand for, so the light over a hillside changed as the level
did. A geometry stage hands the fragment stage the three materials of the
triangle it is in and the barycentric weights, so ground changes over a
triangle rather than along its edge; 3.3 has no gl_BaryCoordEXT.
And the level a tile is drawn at now counts the camera's height. The
distance was measured on the ground plane alone, which puts the tile
underfoot nought metres away from a camera a kilometre above it - and
nought metres asks for the finest level there is, so a view from the air
wanted the whole ground at full detail, more tiles than the buffers hold,
and what would not fit came out as holes.
Archives written under the old rules are refused at the door and baked
again.
The ground in a scenery is already an irregular mesh, so the cook keeps it: it cuts the
triangles along the tile grid, welds what coincides, and builds each coarser level by
collapsing edges. Nothing is resampled, nothing is interpolated, and a hole the scenery left
stays a hole. What was here before threw the triangles away, kept their vertices and built a
fresh Delaunay triangulation over them, which filled every hole, merged the two heights of
every retaining wall into one, and - because it never erased the super-triangle - emitted
indices three too large together with triangles reaching out to the artificial corners.
Tiles form a quadtree, as Cesium Quantized Mesh intends. A distant tile is a large one, so the
horizon costs tens of tiles rather than the thousands a fixed grid needed, and the level to
draw follows from the tile's own geometric error against what the screen can resolve - which is
what the format carries that field for. Every vertex on a tile's border survives into all of
its coarser levels, and so does every edge between two of them, so two neighbours drawn at
different levels share their border exactly: there is nothing to stitch and nothing to morph.
The file layout is the format's, with two deliberate departures, both written down in
quantizedmeshformat.h: the geodetic header fields carry scenery metres, there being no
geographic reference to convert from, and the material a vertex belongs to rides in an
extension of ours. Texture coordinates are not stored at all - ground tiles in world space, so
the coordinate follows from the position and from how many metres one repeat of the material
covers, measured off the scenery once and kept in the terrain's table.
Tiles live in one archive beside the scenery, an entry each, compressed on its own. Baltyk
cooks to 546 kB against 2.2 MB of heightfields, quark to 361 kB against 9.6 MB, tomaszewo to
375 kB against 9.9 MB.
Drawing is one call a frame: every resident tile lives in the same vertex and index buffer,
with its vertices measured from the middle of the terrain so that the whole map fits in floats,
and a frame is one glMultiDrawElementsBaseVertex over the tiles the walk chose and the frustum
kept. On baltyk that settles at 291 tiles resident, 16 MB of mesh, 85 tiles and 6k triangles
drawn.
What counts as ground is what the scenery says, not what its triangles look like. A node under
origin, rotate or scale is an object standing on the ground; a material with an alpha channel is
a tree, a bush, a fence or a catenary mast. Everything else is laid whichever way it faces, so
the side of an embankment or a cutting goes in with the rest instead of leaving a hole along
every bank. Whatever is left out is recorded per node, and the scenery goes on drawing it.
A cook is a cache: it is used again only when the archive, the note of which nodes went into it
and the manifest of what the scenery was made from all agree. A manifest listing a file the
scenery names but does not have used to fail that test for ever, since an absent file never
matches - quark has such an include, and nothing beside it could ever be reused.
tools/terraincook cooks the same way outside the engine, and meshtest beside it checks what the
cook promises: a tile reads back as written, neighbouring tiles agree on their border to the
millimetre, a hole is not filled in, a wall is laid exactly once, and a material's tiling comes
out of the source. seamcheck.py goes: it tested the morphing constants of a heightfield nothing
reads any more.
Replaces heightfield with Cesium Quantized Mesh format (TIN).
Full replacement, not hybrid - TIN is the only terrain format.
Implemented:
- Format: quantizedmeshformat.h, quantizedmeshreader.h
- Triangulator: terraincooker_tin.{h,cpp} using CDT
- Cooker: terraincook now generates TIN by default
- Loader: terraintileloader fully rewritten for TIN
- GPU pool: terraintilepool_tin (VBO/IBO per tile)
- Renderer interface: terrainclipmap.cppm updated
- Shaders: terrainclipmap_tin.{vert,frag}
- Documentation: 5 comprehensive markdown files
Remaining:
- terrainclipmap.cpp render() method (partial, needs completion)
- Build testing and integration
- See TIN_INTEGRATION_GUIDE.md for next steps
Stats: 2700+ lines new code, ~50% vertex reduction expected
A baked heightfield was trusted for as long as its file was readable, and the logic
container while the .scn was older than it: an edited include went unnoticed by both,
and node keys leaned on file dates, so merely touching a file changed them. And cooking
only ever happened as a side effect of starting a scenery.
Every cooked artifact now has a manifest beside it - <scenario>.ehf.deps,
<scenario>.logic.deps - listing each file it was made from with its size, modification
time and a hash of its content (simulation/cookdeps). The parser records what it
includes as it reads, so the list is exactly what the load read. An artifact is current
when every listed file still has its content: size and time are the quick check, and
when either differs the content is hashed, so a touched or copied file does not force a
rebuild and an edited one always does. A file added to the scenery cannot be in a
manifest; the files read are compared with it after the load, and a mismatch sends the
terrain to be baked on the next start - the new file's nodes, whose keys are unknown,
are drawn as geometry meanwhile - and the logic, cooked from the loaded state, is cooked
at once.
The terrain depends on what the scenario includes and not on the scenario file itself,
which the launcher rewrites on every start; the logic depends on both, since events can
be written in either, and costs milliseconds to cook again. With the manifest vouching
for file content, a triangle node's key is its file, line, include parameters and
placement, without the date.
-cook loads a scenery, brings whatever is stale up to date, and quits: the step that
prepares a scenery ahead of play. It closes the window directly, as there is no
simulation to wind down; a quit command waited a minute in the queue for one. Baking on
a first start stays as the fallback, and -nobaketerrain still turns it off.
Tried on a copy of baltyk with a terrain include of its own: a fresh cook bakes and
cooks everything, a second finds nothing to do, a touched file changes nothing, an
edited include is caught by its hash and both artifacts are rebuilt, and a new include
has the logic recooked at once and the terrain on the next run. A cooking run of baltyk
takes 5.5 s with nothing stale and 10.7 s with everything to do.
Three changes to what goes into a baked heightfield and how it meets the geometry drawn
beside it.
The heightfield held every triangle laid in it, including those it does not replace and
that are therefore drawn as geometry too - so they were drawn twice, and a roof or a
bridge deck left in it bent the ground. Now the first pass lays everything and decides
what is replaced; the second lays the replaced triangles together with the neighbours
that support them - the owners of the corner samples of the cells around them, without
which those cells are not drawn - and only that is written. A first attempt that
repeated passes, dropping whatever bordered a hole, eroded the whole terrain away one
ring per pass: the edge of the ground is always a hole. The replaced triangles stay
replaced in the second pass, since their supports are all there and they have fewer
rivals. The triangles bake no longer keep in memory; they wait in a spill file beside
the heightfield while the scenery parses.
Triangles were refused for an edge longer than 200 m, meant for a water plane under the
whole map. On baltyk and tomaszewo that refused thousands of ordinary fields of a
hand-modelled terrain, and nothing there comes near a whole-map plane. Only triangles
standing on their edge - walls, trees, fences - are refused now. The bake log says why
the triangles not replaced are not: standing on edge, too small, under other surfaces,
next to a hole.
Hand-modelled terrain has triangles hundreds of metres long, elevation-model terrain
small ones along the line, and one grid step either loses the detail or samples a field
every metre. A bake now writes three heightfields a factor of four apart -
<scenario>.ehf, <scenario>.1.ehf, <scenario>.2.ehf - and each triangle goes to the
coarsest one that still puts sixteen samples along its longest edge, and to every
coarser one too, so the coarsest holds all the ground and a triangle at the edge of a
finer field can be replaced by a coarser field beneath it. All of them are picked up
together, and a missing or stale one sends the whole set to be baked again.
Where fields overlap each other or the geometry, the depth test decides: the terrain is
drawn with a polygon offset, negative for the reversed depth buffer and larger for each
coarser field, instead of being lowered by 2 cm - so geometry lying on the ground wins
at any distance and on any slope, and a finer field wins over a coarser one.
On tomaszewo the heightfields went from 176 MB to 10 MB, the load that bakes them from
5:16 to 3:11 and its peak memory from 15.7 GB to 8.5 GB - the scenery alone takes 8.0 -
while the triangles replaced went from 37,130 to 40,402. On baltyk from 19 MB to 2 MB,
with 6,670 triangles replaced against 4,040.
Each resident tile had a height texture and a material texture of its own, uploaded with
glTexImage2D, and was drawn once per material it held with every other material's
fragments discarded, all of them whether in view or not. Thousands of draws a frame.
Now the tiles of a level live in one pair of texture arrays (terrain_tile_pool), a layer
per tile. A pool starts at sixteen layers and doubles when full, copying what it holds on
the gpu, so a level takes only the memory its tiles need; uploads go through a pixel
buffer orphaned for each tile. The ground textures of every material are gathered into
one sRGB array with mipmaps (terrain_ground) - drawn into it layer by layer by a small
shader once the engine has each texture ready, since the scenery's textures are
compressed and cannot be blitted - with a table of placeholder colours and measured
tiling beside it.
Each frame the tiles in range are tested against the frustum by bounding sphere and
gathered per level into a buffer texture of instance data, and each level is drawn in
one instanced call. The fragment shader reads the four samples around it, drops the cell
when any of them carries no ground - exactly the rule the bake already assumed - and
blends the ground textures of their materials by distance, so material boundaries
follow the samples smoothly instead of stepping along triangle edges.
State the copy pass changes is captured and put back, and the texture unit cache is
reset after raw binds, so the rest of the renderer finds things as it left them.
On baltyk: all 30 materials land in the array, 663 of the 1120 tiles in range are drawn
after culling, and a frame is at most six terrain draws. The residency line in the log
now says how many tiles in range were drawn.
Levels changed at fixed distances - 512 m, 1 km, 2 km and on - whatever the window, the
field of view or the zoom. Now a level is used out to the distance where one of its sample
spacings covers terraindetail pixels on screen, 3 by default, which follows from the
viewport height and the vertical field of view with the zoom applied; each level doubles
the spacing, so each doubles the distance. Zooming in or a taller window pushes detail
further, a coarser setting pulls it in. The finest range never drops below four tile
sides, which the morphing needs to close the seams.
Levels and morphing both follow the finest range the last scan used, so they always agree,
and a change of more than a percent - a zoom, a resize - triggers a rescan.
tools/terraincook/seamcheck.py now also tries finest ranges of 5.7 and 11.3 tiles, and
finds no gap in any of them.
terraindetail is in the ini and on a slider in the debug panel; the panel's level table
and the residency line in the log show the resulting ranges. On baltyk the current window
works out to 484 m, so the 512 m minimum applies and nothing changes by default; a
scenery setting terraindetail 1 got 1349 m, and the samples on the gpu went from 3 MB to
10 MB with it.
Tiles drawn at the same level met exactly, but a finer tile beside a coarser one has
samples on their shared edge that the coarser one skips, and the ground between them
could open up. This closes the gap the standard way (CDLOD): each vertex slides towards
the next coarser level as its distance nears the end of its level's range, so that by the
end of the band a tile is, vertex for vertex, the coarser tile beyond it. Levels also stop
popping when they change.
The mip chain is point sampled, so the next level's grid is this level's even samples. An
odd vertex slides onto its even neighbour and takes its height, and the grid, whose
diagonals all run the same way at every level, collapses exactly into the coarser
triangulation. No second texture and no neighbour information is needed; the morph is
computed per vertex from its distance to the camera, the same distance the level is
chosen by.
Two things the geometry demands, both settled by tools/terraincook/seamcheck.py, which
walks random camera positions over 128 m, 256 m and 12.8 km tiles and counts every edge
vertex where the finer side has not finished morphing or the coarser side has started:
- the band ends a rescan's worth of travel (a quarter tile) short of the level's range.
Levels are chosen where the camera stood at the last rescan, the morph follows it every
frame, and a band ending at the range itself left gaps at every level boundary;
- the band is a fifth of the range wide, and level ranges are counted in tiles (four for
the finest level) rather than fixed metres, which keeps the coarser tile's band clear of
any edge a finer tile reaches, whatever the tile size.
The script finds no gap with these values. In the game on baltyk, a low view differs from
the unmorphed shader only in a strip of distant ground at the horizon, and nearer than
the first band the picture is identical.
The cooker's comment claimed tile seams were identical at every level; that held only
for neighbours at the same level, and now says so.
Today's work on the binary scene format plan (binary-scene-format-plan.md), from the
first measurement to a terrain that bakes itself on first run and streams from a
thread. The step-by-step history is kept on backup/scene-streaming-format-presquash-0916.
Measuring first. -loadprofile times a scenery load by stage and by node kind and reports
node counts and resident memory. On l204 terrain is 95% of the bytes and 99.8% of the
nodes; on quarkmce2007 vehicles, not models, dominate what is left. The terrain turned
out to be a heightfield in all but name, a ribbon covering 12% of its bounding box.
The heightfield. scene/heightfieldformat.h is the .ehf layout: Morton-ordered tiles
of 128 samples with one shared edge sample, a point-sampled mip chain, zstd per tile,
a nodata sentinel, a material palette with a measured tiling per material, and a stamp
of the rules that decided what counted as terrain. scene/heightfieldreader.h reads it
for the tool, the renderer and the loader alike, and scene/terraincooker.h writes it,
shared by the terraincook tool and the engine so a first-run bake is the file the tool
would have made. The cooker rasterises every triangle where it is drawn (through
scene::place_in_world, the placement the region applies too), keeps an upper surface
laid within half a metre over another and the lower of two further apart, and then
asks each triangle whether the heightfield shows it: laid, reaching a sample, keeping
nearly all it claimed, and with the ground around it drawn at every level out to 2 km.
Baking on first run. simulation/terrainbake bakes <scenario>.ehf the first time a
scenery is opened, choosing the grid step from the first 200,000 vertices, and on later
loads leaves out the triangles its node table says the heightfield shows. Everything
else - trees, grass tufts, gantries, bridge decks - is drawn as written, so a bad bake
can at worst draw something twice, and the heightfield sits 2 cm under the triangles so
the original wins there. Nodes are keyed by file, line, include parameters and
placement; an edited file changes its keys and those nodes are drawn again. A bake under
older rules, or without its table, is thrown away and redone. A launcher copy named
$scenario shares the scenario's files. -nobaketerrain shows the triangles as written.
A scenery can still name hand-cooked fields, several at different steps.
Drawing and streaming. rendering/terrainclipmap draws resident tiles at the mip level
their distance earns, deriving the grid from gl_VertexID and texturing in world space
with the scenery's own materials. rendering/terraintileloader reads and decodes tiles on
a thread of its own; the renderer rescans what is in range only after moving a quarter
of a tile, asks for what is missing nearest first, uploads at most 32 tiles a frame,
drops arrivals the camera has moved past, and keeps drawing a tile at the level it has
until the right one arrives. The terrain has its own range, terrainrange in the ini and
a slider in the debug panel, 20 km by default and independent of basedrawrange; the far
plane of an ordinary view is pushed out to it. The debug panel shows the range, tiles in
view, on the gpu, wanted and queued, and how many sit at each level.
Events. The event queue moved out of the events into a binary heap with a deterministic
tie-break, and an event is handed its activator instead of storing it, so basic_event
is a definition and nothing else.
The logic container. scene/logicformat.h and scene/logiccontainer write a scenery's
events, their targets and sibling chains as flat, deterministic arrays with a schema
hash that refuses a mismatched build. simulation/logiccook cooks it from the loaded
state whenever it is missing, stale or unreadable, reads it straight back to check it,
and resolves each target to a position in the registry its event kind searches.
-cooklogic forces a cook, -verifylogic only checks. Nothing reads it at load yet.
Prefabs. scene/prefab describes a signal by the aspects it can show - lights, speeds,
animation channels, a sound - instead of pasting an .inc per variant, and expands an
instance to ordinary scenery text. Tried on a light signal and a shape signal.
Screenshots without a person present: -shot x,y,z[,yaw[,pitch]], -shotsettle, -shotstay.
What the plan tool was missing was not geometry but the ordinary business of
editing: taking a mistake back, dividing what is drawn, and seeing what the next
click will do before making it.
Cofanie. The document is plain data, so a step is a copy of it. What counts as one
step is decided by watching the document - a fingerprint over the tracks, turnouts
and catalogue - rather than by dressing every edit with a call: a frame that
changes anything opens a step, and the step closes when the drawing settles. That
is what makes a drag, or a number typed digit by digit, one step and not thirty,
and nothing can be forgotten on the way in. Ctrl+Z / Ctrl+Y; the scene editor
listens for the same keys, so it hands them over while the plan has the keyboard.
Złącze i wycięcie. `insert_joint` divides whichever element a station lands in and
leaves one chain: the ground does not move, the curvature at the new joint is the
one that was always there, and the fit reads the chain the same way afterwards. It
is refused on a straight, because two straights end to end are one straight and the
skeleton will not read that, and under a turnout, which is one piece of geometry on
one element. `cut_track` takes a piece out and does leave two tracks - a gap is the
ground between them, not a hole in a chain.
Przejście rozjazdowe. A turnout may now stand against another one: where it stands
is then worked out from that one's frog, not authored, so moving either carries the
other with it and a trapez stays shut. The frog has to land on the line leaving the
other frog - not on the track's axis, which is 1,469 m further on for a 1:9 - so
the station is settled by laying the turnout and walking it along until it does.
Drawing is a tool now, not what a click happens to do: until Rysuj is on, a click
picks what is under it. With it on, the cursor at the growing end shows the corner
and the straight the next click would lay, worked out by the same fit_corner the
click itself uses. A chain only grows at a loose end, and the first element of a
track is put down by two clicks instead of landing on the plan's zero.
Also here: the curvature strip under the panel, pikietaż in the rows and at the
cursor, how far the laid axis runs from the clicks it was drawn from, the file
format saying which frame and which units it is in (v9), plans chosen from the
directory instead of typed, and editor/samples/demo.m0s - two parallel tracks and a
trapez - with the generator that wrote it and a test that opens it and demands it
lays clean.
Two things about the map behind the plan, neither of them the plan's own doing.
stb keeps one flip flag for the whole program - there is a single stb_image.c in
the build - and the model texture loader turned it on for its own sake, because
models want the bottom row first. A WMS tile is the other way about: its first row
is its northern edge, which is what the drawing code takes v=0 to be. So every
tile came out upside down as soon as any model texture had been loaded through
that path, which is why this has come and gone with the order things load in. Both
decoders now say which way they want it per thread, where nobody else can reach it.
And libcurl is no longer required to configure. A tree without it builds an editor
that draws and exports as usual, only with nothing behind the plan, and says so
once rather than showing a silent grey screen; the CI installs it, so what comes
out of the build has its imagery.
The track layout library came in as a subproject of its own -- its own
CMakeLists, its own install and export rules, its own include tree -- which only
made the build harder to follow now that everything else is a module. It is part
of the editor, so it lives where the editor lives.
- editor/plan_*.cppm / plan_*.cpp: the fifteen interface units and twelve
implementation units, flat in editor/ beside editorTerrain and orthophoto, with
the plan_ prefix keeping them clear of what was already there (editor/turnouts
is the preset catalogue, so the geometry is plan_turnout_geometry).
- modules are named eu07.editor.plan_*, like every other module in the tree.
- namespaces move under the editor's own: editor::plan, editor::plan::geometry,
::io, ::render. The library's old name is gone from the sources and the docs.
- every moved file carries the MPL header the rest of the tree carries; it had
none. version.hpp went with the subproject: macros only, and unused.
- the umbrella header is dropped -- nothing imported it.
- the unit tests survive the fold: editor/tests, behind EU07_EDITOR_PLAN_TESTS
(off by default), compiles the interfaces a second time into a small library of
its own, because a test can only import modules from a target it links and
linking the game would drag the whole simulator in. The code under test needs
nothing but a compiler. 7/7 pass.
- a switch may now be written with two dashes as well: --editor beside -editor.
Builds with clang 22 (-DWITH_BETTER_RENDERER=OFF); -editor and --editor both come
up in the editor on empty ground.
The editor branch was written against headers and the PCH, both of which this
branch has done away with, so its own files were ported rather than merged: the
five headers it added are now interface units and their .cpp files module
implementation units.
- editor/{puwg1992,polandmap,turnouts,orthophoto} become eu07.editor.* modules,
application/planpanel becomes eu07.application.planpanel. The plan tool keeps
to its own module rather than joining the eu07.simcore partitions: it stands on
the editor's libraries and eu07.application.uilayer, not on the simulation core,
so :editoruilayer can export import it without a cycle.
- glm arrives by import eu07.glm, never by include, as the rest of the tree does.
maj0sted is header-only and glm-free, so its headers stay includes in the global
module fragment - every unit that names those types includes them itself, since
a global fragment is not visible across units.
- the one textual conflict was :editoruilayer's include block against this
branch's imports; resolved this branch's way plus the planpanel import.
Builds with clang 22 (-DWITH_BETTER_RENDERER=OFF) and starts into the editor on
empty ground with -editor alone.
The failing chain showed the same directory reached two ways - one
spelling with doubled separators, one without - so #pragma once did not
recognise the file and the SDK headers were parsed twice: every COM
interface then looked like a uuid mismatch and BroadcastSystemMessageEx
like a conflicting redeclaration.
vcvars64 was putting the SDK in INCLUDE while clang was also locating it
itself. Dropped the vcvars call; clang's own detection is enough, and one
spelling means the include guards work.
Six module interfaces carry windows.h in their fragment, and a fragment's
include guard does not cross the boundary, so a unit that imported first
and included the SDK afterwards parsed it twice: clang reports it as
conflicting redeclarations of BroadcastSystemMessageEx and mismatched
uuids all through the SDK. Same shape as the basic_filebuf failure, one
level further out - the header arrives through d3d12.h, not directly.
Thirty units get a guarded windows.h ahead of everything; on Linux it is
a no-op. Vendored single-header libraries are left alone, and the block
goes after a BOM rather than before it.
The BOM did not help: CMake preprocesses the .rc with clang before
handing it to llvm-rc, and the marker does not survive that, so llvm-rc
still met a byte it refuses to guess at. Dropped the diacritic from the
file description - it is metadata for the properties dialog.
The inserter aimed at the spot before the first import, which in a module
unit is past the module declaration - so the includes landed in the
purview, attaching those declarations to the module, and re-running piled
up duplicates. It now targets the global module fragment for module units
and keeps the before-first-import rule only for ordinary units.
LPT.cpp and PoKeys55.cpp never imported the modules holding WriteLog,
ToLower and contains. Neither file is compiled on Linux, so nothing had
caught it; Console.cpp, which is built the same way, already had its
imports.
eu07.rc.in was stored in CP1250 - the o-acute in "pojazdow" is a single
0xF3 byte. rc.exe guessed a codepage, llvm-rc refuses to. Re-encoded as
UTF-8 with a BOM, which both recognise.
Only Platform/Windows-MSVC.cmake passes /DWIN32; neither Windows-GNU nor
the GNU branch of Windows-Clang does. Switching drivers silently took the
macro away from every translation unit, and third-party code that tests
the unprefixed WIN32 took its non-Windows path: FSR2 defined wcscpy_s as
wcscpy, which then failed against the SDK's own three-argument call. asio
reads the same macro.
Defined before any add_subdirectory so every target sees it.
betterRenderer gated on WIN32, and WIN32 is true for clang's GNU driver
too, so /arch:AVX2 reached a driver that reads it as a file name.
Swept every compile and link option in the tree for MSVC-style flags
without an MSVC guard; this was the only one left.
clang's GNU driver read /DELAYLOAD:dxcompiler.dll as a file name. CMake's
LINKER: prefix emits -Xlinker for it and passes it bare under MSVC.
The other raw linker flags in the tree sit inside if(MSVC), which is
false for this driver, so they never reach the command line.
CMake writes the compiler paths verbatim into CMakeRCCompiler.cmake and
friends, where a backslash starts an escape sequence, so %ProgramFiles%
arrived as "C:\Program Files/LLVM/..." and tripped on \P. Same trap
waits for the Python root at C:\hostedtoolcache.
CMake wires up C++20 module scanning for Clang only when the frontend
variant is GNU; for clang-cl it sets nothing and refuses to generate:
"the compiler does not provide a way to discover the import graph
dependencies". Not a missing tool, a hard gap - Compiler/Clang-CXX.cmake
guards the whole scandep block on the GNU variant.
clang++ on Windows still targets x86_64-pc-windows-msvc, so the ABI, the
SDK, the vcpkg packages and D3D12 are unchanged. CMake then takes the
GNU path and leaves MSVC false, which is what the /MD /Zi /O2 block wants
anyway; the version resource had to move to a WIN32 guard to survive.
clang-cl keeps the MSVC ABI, so the Windows SDK, the vcpkg packages and
D3D12 all still link; only the front end changes. That front end is the
one whose module support has been giving way - global module fragment
merging and type completeness at a template's instantiation point.
Switches the job to Ninja, which is where CMake's module scanning is best
supported, with vcvars64 for the SDK paths. Ninja is single-config, so
the artifacts move from build/bin/<config>/ to build/bin/.
CMakeLists: the MSVC-only blocks tested CMAKE_CXX_COMPILER_ID, which
clang-cl reports as Clang, so the version resource, the icon and /utf-8
would have been dropped silently. They test MSVC now, which is true for
both. /MAP and /MP stay behind a real-MSVC check - the first is a
link.exe spelling, the second clang-cl rejects.
windows.cpp and piped_proc.cpp reached windows.h through an imported
module's global module fragment, which does not cross the boundary.
LPT.cpp and zmq_input.cpp never had it at all - the first was in an
MSBuild batch that aborted before reaching it, the second is behind
WITH_ZMQ and has not been compiled by either CI job.
The earlier sweep only looked inside #ifdef _WIN32 blocks, but these
files are gated by CMake instead, so they have no such block.
The _WIN32 branch of the physics-log setup calls a Win32 API the file
never included; Linux takes the mkdir branch, so it never showed up
locally. Swept the tree for the same shape - Win32 calls inside _WIN32
blocks without windows.h - and this was the only one.
The template-header propagation only walked files with a module
declaration, so EU07.cpp, windows.cpp, export_e3d_standalone.cpp and
every betterRenderer source were outside it - including the file with
main(). They reach cParser and friends through eu07.simcore and
instantiate their templates without the headers those bodies name.
Adds audit_modules.py, which checks the whole tree for the seven classes
this migration hit; all seven are clean now.
AnimModel.cpp reaches cParser through the simulation core, not a direct
import, so the previous pass missed it; the propagation now follows the
whole import closure. Train.cpp, drivermode.cpp and Globals.cpp call
Console:: inside _WIN32 guards without importing eu07.console - invisible
on Linux, where those blocks are skipped.
A template defined in a module interface is instantiated in the importing
unit, where its body's types must be complete. A declaration reaching the
importer only through the interface's global module fragment is not
enough for MSVC: cParser::operator>> uses std::stringstream, and every
unit that parsed a float got an incomplete type.
526 headers across 161 units, derived from what each interface's template
bodies actually name.
Ordering them before the imports in the same file was not enough: a
project header on line 1 can import a module several levels down whose
global module fragment already pulled in the same standard header, and
that include guard does not cross the module boundary. Putting them
first means the guard is set before anything can import.
Skips vendored single-header libraries and anything inside an #if.
The previous pass dropped imports into the middle of HR_RETURN's
line continuations, breaking nvrenderer_d3d12.cpp, and its scan gave up
on the first line of any file with a BOM - which is why nvtexture.cpp
still had five includes sitting after its imports. Rewritten to skip
macro continuations, tolerate a BOM, and insert only immediately before
the first import.
A header first pulled in through an imported module's global module
fragment leaves no include guard in the importing unit - macros do not
cross a module boundary - so a later textual include parses it again and
MSVC rejects the redefinition. Redoes the betterRenderer sweep with
placement that stays out of #if blocks, which is where the previous one
dropped <fstream> and broke nvrenderer_d3d12.cpp.
Its sources reached the standard library through the project headers they
included; those are imports now. 178 headers across 56 files, plus 18 C
library headers elsewhere for memcpy/printf and friends. MSVC's STL does
not supply these transitively the way libstdc++ does.
Unqualified uint8_t/size_t and friends were reaching 24 units through
some other header on libstdc++; MSVC's STL does not do that, which broke
gl/ubo.cppm (and made its UBS_PAD macro look like garbage). Also
silences C5244, MSVC's counterpart to the clang warning about glm being
included in the module purview, which is deliberate here.
Verified locally now that Vulkan and DXC are installed: all 31 targets
compile. glm has to come from eu07.glm only - having it both as a plain
include and as an import made every overload ambiguous. nvrenderer.h is
self-sufficient, since register.cpp includes nothing else. Fixes the
vcpkg triplet, which was pinned to x64-windows.
The module file set moves to its own eu07_modules library: a target can
only consume C++20 modules by linking the target that owns them, and
betterRenderer is a separate add_library. Its project includes become
imports.
CXX_EXTENSIONS must match across the two - a BMI built with -std=c++20
is refused by a unit compiled with -std=gnu++20.
Every module unit now includes the standard headers it uses. Declarations
reaching a unit only through an imported module's global module fragment
are reachable but not visible, so ADL misses the free operators older
libstdc++ declares beside a type - reproduced on ubuntu 24.04 (libstdc++
13), where gl/vao.cpp could not compare vector iterators.
Also drops a stale editor/editorSettings.hpp from the MSVC source list,
which broke the Windows configure step, silences the deliberate
include-in-purview warning in the glm wrapper, and removes GCC support:
clang or MSVC only.
Linux CI switches to clang 22; fixes what that exposed - export on
out-of-line member definitions, missing std includes, missing glm
imports. GCC 16 still builds.
make_event() lowercased the event name twice: cParser::getToken() already
folds ASCII via toLowerChar() (which intentionally leaves bytes >= 0x80
alone), then ToLower() ran std::tolower() over every byte on top of that.
With a non-"C" global locale, that second pass remaps cp1250 uppercase
accented letters (e.g. 0xAF -> 0xBF). Every lookup path - TTrack::AssignEvents(),
child-event resolution in multi_event, event-launcher name resolution -
normalizes only through the parser, so a registered name containing such a
letter can no longer be found: the track logs "can't find assigned event"
and the binding is silently dropped.
Removing the extra ToLower() makes the registration key match every lookup
site (ASCII-only folding, non-ASCII bytes compared verbatim), and matches
the direction toLowerChar()'s own comment already points at for UTF-8.
A whois with no triggering vehicle previously just wrote nothing to
the target cell with no trace in the log, which is a real pain to
diagnose from the scenery side (this is exactly the kind of thing that
cost real debugging time on a station script recently).
Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
whois_event::run_() dereferenced m_activator unconditionally in every
branch (this predates this PR). getvalues_event::run_() and
putvalues_event::run_() already guard the same way elsewhere in this
file; whois_event was missing it. Single early return covers every
branch, old and new.
