Area alone would pass on a result whose pieces were merged across tiles
or which kept the cut edges of the clip. The rings are compared after
rotating each to its lowest point and sorting, so only the ordering is
free. The fixture now also asserts it really is split into more than one
tile, which is the path being tested.
append(const ExtrusionEntity &) clones; the collection each island
produced was deep-copied into the layer's loops and then thrown away.
The no-overlap areas are moved as well.
The Windows build stopped on test_kdtree.cpp: it calls std::iota without
including <numeric>, which libstdc++ happens to pull in anyway. Added
there, and the same for <limits> and <algorithm>/<cmath> where the
recent changes rely on them being included by something else.
bridge_over_infill's helper for splitting polygons by proximity named
its locals near and far. The Windows headers define both as macros that
expand to nothing, so "Polygons near;" declared nothing and the uses of
it did not compile. Renamed; no behaviour change.
Collecting every candidate within the radius into a vector cost more
than the search itself. Same order, so the seams are unchanged;
align_seam_points ~19.6 s at 0.1 mm / 2000k, was ~21.
detect_surfaces_type, process_external_surfaces and the vertical shells
each waited on their own heaviest layer in turn. The LOTR map plate
slices in ~10.5 min at 0.1 mm / 2000k, was ~11.5; ~87 s at normal
settings, was ~97.
MultiPoint had no rvalue constructor, so the derived move constructors
bound to the const reference and copied; append reserved exactly, so
collecting pieces one by one was quadratic. Colour segmentation ~3 s at
0.1 mm / 2000k, was ~40, and ordinary prints gain too.
Only the slices within the deepest shell offset decide the result, so
the work is done per tile of the face. Top and bottom segmentation
~130 s at 0.1 mm / 2000k, was ~180.
ClipperLib slows down with the number of edges on a scan line, and a
layer cut through a fine relief has tens of thousands of pieces.
detect_surfaces_type ~50 s at 0.1 mm / 2000k, was ~145.
The merge took anything from 3 to 38 minutes at 0.1 mm / 2000k, now
~2.5. Every region is grouped with the islands it overlaps, so the
result is the same.
A colour texture baked at 0.1 mm / 2000k made the top layers thousands
of islands and slicing never finished. Colour segmentation runs per
island, the merge subtracts piece by piece, the support check tests only
nearby islands, and the travel ordering finds crossings through a grid.
A layer split into ~1000 colour fragments (a colour texture baked over a
large top face) made several per-fragment loops redo whole-layer ClipperLib
work, so slicing took ~33 min; it now takes ~3 min with the same output.
- make_fills: clip the layer's no-overlap area to each expolygon's box
before intersecting
- discover_vertical_shells: small-piece filter compares only against the
nearby part of the layer
- bridge_over_infill: whole-layer union/diff/intersections restricted to the
candidate's neighbourhood; fill boundary expanded once per spacing; anchor
tree built only from lines crossing the scan range; bbox pre-check in the
collision test; limiting outline taken directly instead of through
expand(..., 0.3 * flow.spacing()), which offsets by 0.135 scaled units
(flow.spacing() is in mm) and only cost a whole-layer pass per candidate
notarytool submit --wait has no timeout. On 2026-10-02 it hung for
over 5 hours in a main build. Since #16044 a new push no longer
cancels a running main build, so nothing stopped it and six waiting
main runs were replaced without starting.
Over the last 30 days the step succeeded 206 times, with a median of
4.3 minutes and a maximum of 14.8.
* Fix debug build after qhull upgrade
We upgraded qhull from 8.0.1 to 8.0.2 in 504a5d3b70, which contains a commit qhull/qhull@16159c648c `use same CMake target name for Debug and non-Debug`, so this target name check is no longer required
* Fix issue like `IMPORTED_LOCATION not set for imported target "opencv_world" configuration "RelWithDebInfo".` when build Debug config
* Keep User Preset Values on Extruder Variants They Don't List
A user preset stores the variant list its parent had when it was saved.
When the parent later gains variants, update_diff_values_to_child_config
matched variants by name only and left the new ones at the parent's
value, so the user's settings were silently replaced there, and a
re-save wrote the system values into the user's file.
An unmatched parent variant now takes the child's first variant of the
same extruder, the rule slicing already uses in get_config_index_base.
A child without a variant list covers the parent's first extruder. The
name match also no longer indexes the child's extruder ids when it has
none.
* Share One Variant Column Rule Between Slicing, User Presets and Projects
Three places chose which variant column a value comes from, each with
its own copy of "the same variant and owner, else the owner's first
column": get_config_index_base when slicing, the user preset merge in
update_diff_values_to_child_config, and normalize_filament_values_to_variants
for projects and the CLI.
find_variant_column now holds that rule and map_variant_columns applies
it to a variant list, so a change to how missing variants are filled
reaches all three. Each caller keeps its own copy step. There is no
behaviour change: G-code is identical before and after. The one
relaxation is that get_config_index_base no longer reads past a short
id list when its two lists differ in length, which its assert already
rules out.
* Rename variant column helpers to variant index
---------
Co-authored-by: SoftFever <softfeverever@gmail.com>
* fix: include bundled UV binary for arm64
* fix: update unit test CI
* Install unit-test numpy only with the bundled uv
* Simplify the unit-test script's uv lookup
---------
Co-authored-by: SoftFever <103989404+SoftFever@users.noreply.github.com>
Co-authored-by: SoftFever <softfeverever@gmail.com>
The ARM64 jobs pinned CMake 3.31 because CMake 4 dropped pre-3.5 policy
compatibility and its ARMASM support broke Boost.Context. Both are
handled now. deps/CMakeLists.txt sets CMAKE_POLICY_VERSION_MINIMUM on
CMake 4, and Boost.Context uses the winfib implementation on ARM64, so
nothing assembles with armasm.
CMake 3.31 also predates VS 2026. Its InstallRequiredSystemLibraries
treats the v145 toolset as v143, searches only the VS 2017-2022 install
directories and finds no runtime, so the ARM64 installer ships without
msvcp140.dll and vcruntime140.dll. CMake 4.2 and newer find the VC145
redistributable.
get-cmake also installs Ninja, so the ARM64 jobs now use its latest
release instead of the one already on the runner, as x64 does.
The install now fails when InstallRequiredSystemLibraries returns no
msvcp140.dll or vcruntime140.dll, after a configure warning naming the
CMake and MSVC versions. A CMake that predates the Visual Studio in use,
on a developer machine or after the next runner image update, then
stops the installer build instead of shipping one that cannot start.
The build_win.bat prerequisite installer drops its matching 3.31.8 pin.
This is the bulk of the profile changes for U1 that led to the cooling
catiant connection.
The ugly end gcode is used by the printer's UI to complain if normal
nozzle is installed but the file was sliced for highflow.
[How to Download Pull Requests Artifacts for
Testing](https://www.orcaslicer.com/wiki/how_to_download_pr_artifacts)
* Make Painted Multi-Material Slicing Deterministic
Painted (multi-material) models sliced to slightly different G-code on
every run: ±1 µm wall coordinates and reordered islands. Hashing each stage
of the segmentation across runs showed the projected painted lines and the
per-layer Voronoi segmentation were stable; the raw top/bottom projections
from slice_mesh_slabs() were not. Three causes, all thread-order dependent:
- slice_slabs_make_lines() appends each slab's intersection lines from a
parallel facet loop and never restored a canonical order, so the loop
start vertices and polygon order from make_slab_loops() depended on
scheduling. Sort every slab's lines with the same key slice_make_lines()
already uses.
- segmentation_top_and_bottom_layers() wrote a layer's shell projections
into neighbouring layers' vectors from the parallel loop, relying on a
parity double-buffer that assumes TBB ranges are exactly one group wide
and aligned, which blocked_range does not guarantee; two threads could
append to the same vector. Each source layer now records its projections
in its own slot and they are gathered per target layer in source order.
- The painted-line sort in post_process_painted_lines() was not a total
order: projections of one span from facets of different colours tied on
every key and the first one won the span. Colour and end points now break
the tie.
Three multi-threaded runs of each painted fixture now give one G-code;
unpainted output is unchanged.
* Test That Slab Slicing Does Not Depend on the Thread Schedule
Projects a dense, tilted sphere with slice_mesh_slabs() on one thread and
then three times multi-threaded, and requires the polygons to match exactly,
vertex order included. Fails without the canonical line sort, passes with it.
* Add Missing Includes Across src/slic3r/GUI
Every GUI source and header now directly includes the headers declaring what it uses, rather than relying on the precompiled header or transitive includes. Generated with clang-tidy misc-include-cleaner, plus one hand edit making CalibrationPanel.hpp self-contained.
* Drop the OS-Specific Includes Added Outside Their Platform Guards
GLib, GTK, D-Bus and POSIX headers are only used inside platform #if blocks, which already include them. Added unconditionally at the top of the file they broke the Windows build.
* Add the clang-tidy Configuration That Generated These Includes
Only misc-include-cleaner's missing-include check, with the headers it must never suggest: per-platform, internal and OS-specific ones that would break other platforms or are not meant to be included directly.
* Match Windows Paths in the clang-tidy Ignore List
Header paths use backslashes on Windows, so every / in a pattern is now [/\\]. The Windows SDK headers are ignored alongside the other OS-specific ones, and the list is one pattern per line. Suggested by @raistlin7447 from a Windows clang-cl run.
* Lock Config and Preset Files Across Instances and Write Them Atomically
Every running instance shares one OrcaSlicer.conf and one user preset
tree, and nothing kept their writers apart. Two instances saving at the
same moment, or the cloud preset sync thread writing while the GUI thread
saved, could interleave, and a reader in another instance could open a
preset JSON or .info file between truncate and close and get a partial
file, dropping that preset for the session with a parse error.
Add InstanceLock, a scoped guard that serialises the threads of one
process through a recursive mutex and other processes through an advisory
OS file lock: flock on POSIX, held on the guard's own descriptor so no
other close in the process can drop it, and LockFileEx on Windows. The
outermost guard opens the lock file and closes it on release, so nothing
stays open between saves and a data dir can be removed once nothing is
saving into it; the file itself is kept, since deleting it would let a
third instance lock a fresh file while the second still holds the old
one. It is best effort: when the lock file cannot be opened or locked, or
another instance still holds it after a second, the guard logs once and
lets the write proceed, then leaves the file alone for ten seconds, so a
hung instance never blocks every other one and a holder stuck in a
debugger does not cost a stall per save. The guard sits at the leaf
readers and writers: set_sync_info_and_save() calls save_info() under the
preset collection mutex, so a batch lock around save_user_presets() would
invert the order against the sync thread. The user preset scan reads its
files on worker threads without the guard, since the mutex would
serialise them, and takes it per file in the serial commit step, so a
save never waits for the whole scan. Each read keeps the bytes of the
preset and its .info as they were before parsing; commit compares them
with the disk under the guard and reads a file that changed again, so it
never deletes or writes back over another instance's newer save, nor
installs a .json and .info from two different saves; a preset another
instance removed in the meantime is not installed. Without the guard, in
a cool-down, the scan still loads the presets but leaves their files
alone: an unreadable file stays for the next scan, and a derived
compatible printer is not written back. Read-only scans, which is what
the CLI does, take no lock and create no lock file.
AppConfig holds OrcaSlicer.conf.lock in load() and save(); load is
included because the Windows path restores from the .bak copy. Every
user preset writer and reader holds user.lock: Preset::save(), which
writes no .info when the preset itself could not be written, since an
.info without its preset reads as a cloud deletion request, save_info(),
reload() and remove_files(), each preset the scan commits, the
bundle metadata reads and write, the .info removal after a cloud-confirmed
delete, the orphaned-.info scan on the sync thread, the bundle folder
removal on unsubscribe and the physical printer writers and delete
paths. A bundle import extracts under cache/ into a folder per process
and per import, where no scan reads.
Preset JSON, .info, bundle metadata, physical printer and config files,
and the caches and state files that already used a temporary by hand,
now go through write_file_atomically(), which writes <file>.<pid>.<n>.tmp
beside the target and renames it over, so a reader that never waits sees
a complete old or new file. A symlink is followed; a target that is not
a regular file is written in place; and when no temporary can be created
beside an existing target, or the rename itself is refused, by a Windows
reader holding the file open or a mount that cannot replace in one step,
the helper writes in place as before, since losing the save is worse
than a torn read. On POSIX the rename replaces the
target atomically where the old code removed it first and left a window
with no file at all; only a mount that refuses a one-step replace gets
the old remove-then-rename. A crash between temporary and rename leaves
the temporary behind, which no scan reads. Preset::save() returns
whether it wrote the preset, so the scan counts a compatible printer it
could not write back as an error. A failed config write keeps
the config dirty, and the idle handler waits ten seconds before retrying
while an explicit save always tries.
* Run the Cross-Process Lock Test on Every Platform
The test that checks the guard yields to a lock held elsewhere forked a
child to hold it, so it was left out on Windows. The OS lock belongs to
the handle on Windows and to the open file description elsewhere, so a
second handle in the same process is refused like another instance
would be. The test now holds the lock that way and runs everywhere.
extrude_infill() and extrude_support() reversed the layer's extrusion
entities in place while chaining them, so each export started from the
previous one's reversed toolpaths, and each copy of an object from the
copy before it. The export-time region lists now hold const pointers,
and chaining reverses a clone instead.
The display-name tests call set_user_session(), which persists the session.
The agent they built was in keychain mode, so every run saved a fake
OrcaSlicer/Auth session into the system keychain of whoever ran the tests,
replacing their real Orca Cloud login on any desktop with a working keychain.
Move them next to the other agent tests and build the agent the same way:
encrypted-file mode with a throwaway config directory.
An ironing line spacing of 0 reached the fillers from a 3MF, the CLI or
the per-filament override, which has no GUI guard. Concentric ironing
then never finished slicing, because a zero inset never shrinks the
region, and rectilinear ironing was silently dropped. Tiny positive
values produced an unprintable number of lines.
Top surface and support ironing now clamp the spacing to the 0.05 mm
floor the process GUI guard already enforces, so these configurations
iron at that spacing. Spacings at or above the floor, including every
shipped profile, are unchanged. The concentric filler also returns early
on a non-positive step so no other caller can hang it, and the filament
settings page now resets a too-small override the same way the process
page does.
# Description
With Xcode 27, building deps on macOS 26 fails at the Python install
step with a segfault, and a libpython built with Xcode 27 crashes on
macOS 12–26 the first time anything calls `os.pipe()`, which every
plugin `subprocess` call does. The macOS 27 SDK declares `pipe2()` and
`dup3()` as macOS 27-only, and CPython 3.12 calls them without a runtime
check once configure finds them, so on older systems they resolve to
NULL. This keeps CPython on the `pipe()`/`dup2()` fallbacks it already
uses with older SDKs; upstream fixed it in 3.13+
([python/cpython#153711](https://github.com/python/cpython/issues/153711))
but not in 3.12.
No change for builds with Xcode 26 or older, or on Linux and Windows.
# Screenshots/Recordings/Graphs
<!--
> Please attach relevant screenshots to showcase the UI changes.
> Please attach images that can help explain the changes.
-->
## Tests
Rebuilt deps with Xcode 27 on macOS 26.6: the Python install step now
completes, the installed libpython no longer imports `pipe2`/`dup3`, and
CPython's `test_os`, `test_subprocess` and `test_posix` pass, apart from
one test that needs `_testcapi`, which `--disable-test-modules` leaves
out. Running CPython's configure against the macOS 26.5 and 27.0 SDKs
gives a byte-identical `pyconfig.h` for 26.5 with and without this
change, and for 27.0 with it. The x86_64 cross-build path was configured
on arm64 to confirm both of its configure runs pick up the change.
<!--
> A guide for users on how to download the artifacts from this PR.
-->
[How to Download Pull Requests Artifacts for
Testing](https://www.orcaslicer.com/wiki/how_to_download_pr_artifacts)
The plugin test fixtures start the interpreter before the test points
data_dir at its temporary directory, so PythonInterpreter creates
{data_dir}/python/packages and {data_dir}/log with an empty data_dir: a
python/ and log/ folder in whatever directory the tests run from. When that
is the test binary's folder, the next run's embedded-interpreter tests took
the stray python/ as their home and failed to start Python.
Give each fixture that initializes the plugin manager its own temporary
data directory, set up before initialize(), and only use the python/ folder
next to the test binary as the interpreter's home when it holds a standard
library.
With stealth mode off the home page asks for the login status every 2 s,
and while nobody is logged in that ends in clear_user_secret(), which
opened the system keychain and deleted the OrcaSlicer/Auth entry on the
UI thread every tick. A working keychain cost a D-Bus round trip per tick;
a keychain that never answers blocked the UI for 25 s per tick. It also
deleted a login another running instance had just saved, and ignored
use_encrypted_token_file, so opting out of the keychain did not help.
Remember whether this process read a secret from the store or wrote one,
and only then touch the store when a logged-out poll asks for a logout.
A logged-out instance never touches the keychain, and in encrypted-file
mode the poll no longer opens the keychain at all. A secret this process
cannot read, such as a token file encrypted for another OS user sharing
the data directory, is left alone by the poll. An explicit logout still
wipes both backends, so a token stranded by switching the token storage
option cannot sign the account back in later.
# Description
Include variant-aware validation, temperature and pressure controls,
named nozzle selection, and filament sidebar refresh.
## Support nozzle-specific filament cooling and tuning
Filament presets can require different cooling and tuning for Standard
and High Flow nozzle variants. This change makes part and auxiliary
cooling, pressure advance, temperature limits, and multitool ramming
settings follow the selected variant, with matching UI controls and
profile validation.
The Snapmaker U1 profiles in #15755 / Snorca illustrate why this
matters:
| Profile | Standard cooling | High Flow cooling |
|---|---|---|
| Snapmaker ABS | Part fan: 15–15% | Part fan: 10–60% |
| Snapmaker PETG HF, 0.4 mm | Part fan: 20–40% | Part fan: 30–60% |
| Snapmaker PETG-CF | Part fan: 0–20%; auxiliary: 0% | Part fan: 5–40%;
auxiliary: 20% |
| Snapmaker PLA Matte | Auxiliary: 80% | Auxiliary: 100% |
These differences need to survive variant selection, editing, and
slicing. Shared single values continue to apply across variants, and
short filament arrays fall back to element zero, matching the loader.
### Additional changes
- Preserve named nozzle selections when refreshing the diameter
selector.
- Enable or disable the pressure-advance input for the selected variant.
- Initialize profile-validation slicing with the printer’s declared
nozzle volume type.
- Refresh filament controls after preset loading.
Profile changes remain separate in `u1-hf`.
### Validation
- All 13 focused variant-tool tests passed.
- Broader checks encountered an existing Windows path-separator
assertion failure, reproduced with upstream code.
- C++ build and GUI validation have not been run.
[How to Download Pull Requests Artifacts for
Testing](https://www.orcaslicer.com/wiki/how_to_download_pr_artifacts)
## Summary
Adds a new paint-style **Texture Displacement** gizmo that stamps
grayscale
height-map textures onto a model's surface and turns them into real
relief -
engraved or embossed detail - either as a live preview or baked into
actual mesh
geometry.
You paint where a texture applies, stack up to **8 blended texture
layers**
(image-editor semantics: Add / Subtract / Multiply / Divide), choose how
each is
projected onto the surface (Triplanar / Cylindrical / Spherical / LSCM
unwrap /
From-view), and - for the LSCM projection - lay the charts out by hand
in a new
dockable 2D **UV Editor** pane. Coarse models can be **Subdivided** or
**Remeshed** first so there are enough vertices to carry fine detail,
and the
result is committed with **Bake**, restricted to the painted area only.
The tool only ever affects the **painted** region; everything left
unpainted
keeps its original surface, and bake blends the relief seamlessly into
it with no
remeshing or hole-filling at the seam.
---
## User-facing features
- **Paint the affected area** with Brush (circle/sphere), single-Face,
or
Connected-area flood fill; or **Select whole model** - reusing the
existing
`TriangleSelector` / `FacetsAnnotation` painting machinery, one full
mask per
layer slot.
- **Up to 8 texture layers**, each with its own paint mask, texture, and
parameters, combined in slot order like image layers.
- **Per-layer controls:** Depth, Tile size, Rotation, Midlevel
(bidirectional
emboss/engrave), Smoothing, Edge smoothing, Invert, Blend mode, Tile
(Repeat / Mirrored-repeat / decal), and Projection.
- **Projection methods:** Triplanar (blended, seam-free across sharp
edges),
Cylindrical, Spherical, **Unwrap (LSCM)** - a real CGAL conformal
parameterization - and **From view** (slide-projector decals).
- **UV Editor pane** for LSCM layers: move / rotate / scale / snap / cut
/ join
islands, average texel density, Checker and Distortion overlays, manual
mark-
seam workflow, live model update while dragging.
- **View modes:** Normal (true displaced geometry = what Bake produces),
Fast
(GPU bump-shaded approximation of the active layer), Checker, Distortion
heatmap, and an independent Wireframe toggle.
- **Mesh prep:** Subdivide (1–5* uniform 1->4 split) with cyan-wireframe
preview,
and CGAL isotropic **Remesh** to a target edge length.
- **Texture library:** 10 shipped seamless 512×512 grayscale height
maps, plus
import of any PNG/JPG/BMP (converted to an 8-bit grayscale map and
copied into
the user data dir so app updates can't clobber it).
- **Bake** runs in the background (off the UI thread); the preview is
free to
explore and only Bake changes the real mesh.
---
## How it works (implementation)
- **Bake is accumulate-then-displace and topology-preserving.** The
output mesh
has exactly the input's vertices and triangles in the same order; only
displaced vertex positions differ. Each layer is evaluated against the
**base
mesh** and folded per-vertex into a shared accumulator via its blend
mode, then
every touched vertex moves once along its precomputed undisplaced
normal. This
replaced an earlier sequential re-mesh-per-layer design that was the
root cause
of the "second layer never applies" bug and made blend modes impossible.
- **Boundary vertices are pinned.** Any vertex shared with an unpainted
triangle
is never displaced, which is what keeps bakes seamless with zero
remeshing.
- **LSCM unwrap** uses a new `MeshBoolean::cgal::parameterize_lscm()`
built on
CGAL's already-vendored `Surface_mesh_parameterization` package - **no
new
external dependency**.
- **Fast GPU preview** perturbs the shading normal from the height
gradient
(analytic mm-per-mm slope for triplanar; Mikkelsen's
screen-space-derivative
method for the conformal LSCM path), so apparent depth matches the bake.
- **Background jobs:** preview compute and bake both run off the UI
thread on the
shared job worker (queued, not `replace_job`, so a preview never cancels
an
in-flight bake); a generation counter discards stale results.
- **UV Editor** shares the app's single real `wxGLContext` and reuses
the
registered `flat`/`flat_texture` shaders; island drags update one affine
matrix
per island rather than re-uploading geometry.
---
## Backward compatibility & constraints
- **Feature is fully gated behind the new gizmo** - it adds no new
default
behavior and does not touch existing slicing, profiles, or defaults.
Models
that never open the tool are unaffected.
- **Cross-platform** - pure `libslic3r` / `libslic3r_gui` /
`libslic3r_cgal`
code; no new dependency and no `deps/` rebuild. (Built and tested on
Windows;
no platform-specific APIs introduced.)
- **`.3mf` compatibility:** **baked** relief round-trips fine, since it
becomes
ordinary mesh geometry via the existing serialization path. **Unbaked**
paint
masks and layer definitions are **not yet serialized** - see
Limitations. No
existing project data is affected.
---
## Testing
Unit tests in `tests/libslic3r/test_texture_displacement.cpp` - **run
and
passing** (7 cases, 116 assertions). Coverage:
- `decode_height_texture` round-trip
- Empty-layer no-op
- Full-cube uniform displacement
- Boundary-vertex pinning on a hand-built fan mesh
- Regression: a **second layer over the same area actually contributes**
(the
bug the bake rewrite fixed)
- Table-driven check of all four blend modes
- The lowest painted layer ignoring its blend mode
`BUILD_TESTS` is `OFF` in the checked-in cache; enable to run:
```bash
cmake -S . -B build -DBUILD_TESTS=ON
cmake --build build --config Release --target libslic3r_tests -- -m
./build/tests/libslic3r/Release/libslic3r_tests.exe "[TextureDisplacement]" --order rand
```
---
Building deps with Xcode 27 on macOS 26 failed at the Python install step,
and a libpython built with Xcode 27 segfaulted on macOS 12-26 whenever a
plugin started a subprocess.
Fan speeds, the recommended nozzle temperature range, minimal purge and
multi-tool ramming flow now take one value per extruder variant. Each
single value is repeated for every variant the filament lists, so every
preset loads exactly what it did before.
ExPikaPaka