fix: OK/Cancel buttons clipped in Flushing Volume dialog (#13511)
The WipingDialog renders its UI inside a wxWebView. When the dialog
was clamped to screen size (many filaments, small displays, high DPI),
the HTML content exceeded the WebView bounds and the OK/Cancel buttons
fell below the visible area.
HTML fix:
- Convert .container to flexbox column with overflow-y: auto
- Pin .button-container with flex-shrink: 0 so it stays visible
- Allow .scroll-container to flex-grow for the table area
C++ fix:
- Replace heuristic extra_size with accurate fixed_overhead estimate
- Use correct cell height (25 DIP) matching HTML table row height
- Add screen margin for window decorations
- Enforce minimum dialog size when clamped
Co-authored-by: SoftFever <softfeverever@gmail.com>
* Add Optimized Gyroid infill (auto-tuned wavelength + amplitude)
New infill geometry derived from FillGyroid. Two parameters are
auto-computed per-region from density, line spacing, and layer height
(no user inputs):
omega = sqrt(density_adj) / sqrt(1 + layer_height/spacing)
clamped to [0.5, 2.0]
-- Euler-Bernoulli buckling: critical load ~ 1/L^2,
so shorter wavelength under higher load (denser infill)
raises buckling resistance.
amplitude = 0.55 / omega^2, clamped to [0.20, 0.65]
-- Curved-beam bending stress: peak stress ~ A * omega^2,
so amplitude is reduced as omega rises to keep peak
fiber stress bounded while preserving stiffness.
Files:
- src/libslic3r/Fill/FillOptimizedGyroid.{hpp,cpp} (new)
- src/libslic3r/Fill/FillBase.cpp (factory case)
- src/libslic3r/Fill/Fill.cpp (switch case)
- src/libslic3r/Layer.cpp (switch case)
- src/libslic3r/PrintConfig.{hpp,cpp} (enum + label)
- src/libslic3r/CMakeLists.txt (build sources)
User-facing: appears as "Optimized Gyroid" in the Fill Pattern dropdown.
Density still chosen by user; omega/amplitude are internal.
* Fix build: layer_height is in FillParams, not Fill base
* Add ipOptimizedGyroid to multiline infill list in ConfigManipulation
* Refactor: replace ipOptimizedGyroid enum with gyroid_optimized boolean
Per @RF47's review feedback, fold the optimized wave math into FillGyroid
itself behind a per-region boolean instead of a separate infill enum.
What changes:
- New ConfigOptionBool "gyroid_optimized" on PrintRegionConfig (default
false). When unchecked, gyroid behavior is byte-identical to before.
- Optimized wave math (compute_omega_factor, compute_amplitude_factor,
f_opt, make_*_opt, make_optimized_gyroid_waves) lives inside
FillGyroid.cpp. _fill_surface_single branches on params.gyroid_optimized.
- FillParams gains a bool gyroid_optimized field, populated in Fill.cpp
from region_config alongside fill_multiline.
- UI checkbox added under Strength > Infill in Tab.cpp, label
"Optimize gyroid wave (experimental)". Toggle is hidden by
ConfigManipulation when sparse_infill_pattern != ipGyroid.
- "gyroid_optimized" added to s_Preset_print_options for preset I/O.
What goes away:
- ipOptimizedGyroid enum value, factory case, switch cases, dropdown
label, string key.
- FillOptimizedGyroid.cpp / FillOptimizedGyroid.hpp (math moved into
FillGyroid.cpp).
- Net diff drops by ~250 lines.
Existing profiles using gyroid are unaffected.
* Wire gyroid_optimized through SurfaceFillParams to FillParams
Linux build failed because line 921 in Fill.cpp populates a
SurfaceFillParams (the dedup struct), not FillParams directly.
Add the field there, in operator< / operator==, and copy it to
FillParams at both conversion sites.
* Use toggle_line for gyroid_optimized: hide row when pattern != gyroid
* Account for multiline wall thickness in omega correction (per @RF47)
When fill_multiline = N, each gyroid wall is N lines thick, so the
geometric scale fed into the buckling correction term should be
spacing * N rather than spacing. Increases omega (tighter wavelength)
when multiline is enabled, consistent with the thicker wall being
more buckling-resistant.
* Optimized gyroid via marching squares on the implicit scalar field
Per @RF47 review: replace the analytical f_opt / make_one_period_opt
wave generator (which had visible kinks at vertical-horizontal
transitions) with a marching-squares iso-extraction on the gyroid
scalar field, modeled on FillTpmsFK.cpp.
- New marchsq::GyroidField in FillGyroid.cpp evaluates
F(x,y,z) = sin(fx*x)cos(fy*y) + sin(fy*y)cos(fz*z) + sin(fz*z)cos(fx*x)
where fx = omega * baseline (anisotropic in x), fy = fz = baseline.
- get_gyroid_polylines() runs marching squares at iso=0 and converts
rings to polylines.
- _fill_surface_single() optimized branch now builds GyroidField,
runs marching squares, and skips the bb.min translate (field
output is already in absolute coords).
- Dropped: f_opt, make_one_period_opt, make_wave_opt,
make_optimized_gyroid_waves, compute_amplitude_factor. Amplitude
has no clean analog in iso-zero extraction.
- Standard (non-optimized) gyroid path unchanged.
* Mass calibration: compensate period by cbrt(omega) for x-anisotropic field
Per @RF47: optimized vs standard gyroid had different masses at the
same sparse_infill_density setting. Cause: scaling fx by omega while
leaving fy=fz at the baseline raised the surface-area-to-volume ratio
by approximately omega^(1/3) (the geometric mean of the three
frequencies).
Fix: multiply the base period by cbrt(omega) so the geometric mean of
(fx, fy, fz) returns to the standard baseline. Net effect:
fx = omega^(2/3) * baseline_orig
fy = fz = omega^(-1/3) * baseline_orig
which preserves total mass at the same density setting while
preserving the load-direction anisotropy this PR introduces.
* Switch optimized gyroid anisotropy from X to Z (per @RF47)
Z is the typical compression-load axis for FFF parts and is not at
delamination risk under compression — so the dominant failure mode
is column buckling of the vertical strands themselves. Tightening
fz directly shortens the effective vertical strand length, which
improves Z-axis buckling resistance.
Mass calibration via cbrt(omega) period compensation still applies
(scaling exactly one of three frequencies by omega; the geometric-
mean preservation argument is symmetric across axes).
* Update src/slic3r/GUI/Tab.cpp
Co-authored-by: Rodrigo Faselli <162915171+RF47@users.noreply.github.com>
* Address review feedback (Copilot + @RF47)
- Fill.cpp: gate params.gyroid_optimized on (params.pattern == ipGyroid)
so non-gyroid surfaces don't differ in SurfaceFillParams by an
irrelevant flag (would unnecessarily split fill batching).
[Copilot suggestion, RF47 confirmed correct]
- PrintConfig.cpp: drop "amplitude" from the tooltip; only wavelength
is parameterized (the marching-squares iso=0 extraction is invariant
to a uniform field scale, so amplitude has no effect).
- FillBase.hpp: shorten gyroid_optimized comment to match the actual
carried state (no amplitude term).
- FillGyroid.cpp: shorten the marchsq namespace comment block; the
ODR concern was overstated (FillTpmsFK uses the same pattern fine).
* Drop redundant marchsq bb expansion (Copilot)
bb is already offset by 10 * scale_(spacing) above for edge-artifact
margin; the second offset on bb_field doubled the raster area for no
geometric benefit and hurt CPU time on large parts.
* Update src/slic3r/GUI/Tab.cpp
Co-authored-by: Ian Bassi <ian.bassi@outlook.com>
* Fix density mismatch + rename to Z-buckling bias optimization
Issue (per @ianalexis): at the same sparse_infill_density setting,
the optimized branch produced denser fill than standard. Verified via
Python sim (sim_gyroid_compare.py) using marching squares on the
implicit field across multiple z slices.
Root cause: the omega formula was inverted from the buckling-physics
intent. The naive sqrt(density_adj) factor produced omega < 1 at
typical print densities (10-30%), which LENGTHENED the Z wavelength
instead of shortening it -- net loss in both mass and strength.
Fix:
- compute_omega_factor: invert to sqrt(1 / density_adj), clamp to
[1.0, 2.0]. Now omega = 2.0 at low density (long strands need
most help) and clamps to 1.0 above ~30% density (no-op, since
standard gyroid is already short enough).
- Remove the cbrt(omega) period compensation. Empirically (sim
table embedded in FillGyroid.cpp comment) the inverted formula
keeps line length per area at ~1.000 of standard across all
densities with no period scaling needed.
Predicted gains (sim, Z-axis Euler buckling proxy):
density line/std strength/std
10% 1.000 2.84x
15% 1.000 1.89x
20% 1.000 1.42x
30%+ 1.000 1.00x (no-op)
Rename per @ianalexis: "Optimize gyroid wave" oversells (now no-op
above 30% density and Z-only). Renamed user-facing label to
"Z-buckling bias optimization (experimental)" with updated tooltip
that scopes to vertical compression and discloses the density cutoff.
Internal config key (gyroid_optimized) unchanged for diff size.
Real-world Instron compression tests at Brown's Prince Lab to follow.
---------
Co-authored-by: Rodrigo Faselli <162915171+RF47@users.noreply.github.com>
Co-authored-by: Ian Bassi <ian.bassi@outlook.com>
Co-authored-by: SoftFever <softfeverever@gmail.com>
* Added UI force-sync button and fixed bug that didn't sync in one case and caused orange highlight
* Fix sync preset race: join old thread before starting new one
---------
Co-authored-by: Mykola Nahirnyi <mnahirnyi@amcbridge.com>
Co-authored-by: SoftFever <softfeverever@gmail.com>
* feat: add UI feedback on http error and some logs
* spelling fix
* show error dialog only once per session
* show errors with plater notification when on developer mode
* remove return
* remove irrelevant logs
GridCellSupportEditor::DoActivate dereferenced
grid->GetSelectedBlocks().begin() without checking against end(). In
wxWidgets 3.1.5 that range always contained at least one block; in 3.3.2
it is empty after the user deselects, and the dereference crashes inside
wxGridBlockCoords::GetLeftCol().
Fall back to the (row, col) that triggered activation when the selection
is empty, so the existing single-cell branch handles it. While here,
drop a dead local_table cast that was never read.
Fix macOS orcaslicer:// deep links after wxWidgets 3.3.2 upgrade
Install an OrcaSlicer-owned kAEGetURL Apple Event handler from
on_init_inner(). The wxWidgets 3.3.2 handler registered in
applicationWillFinishLaunching: stopped delivering URL events to
GUI_App::MacOpenURL on macOS (#13119), so links from Printables /
Thingiverse opened a blank project instead of importing the model.
Registering our own handler late in startup is last-writer-wins on
NSAppleEventManager and routes back to the existing MacOpenURL ->
start_download path, restoring the pre-upgrade behavior without
touching wxWidgets.
Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
Use glBlendFuncSeparate in GLTexture::render_sub_texture so destination
alpha stays at 1.0. The Wayland compositor honors framebuffer alpha for window compositing; the previous straight-alpha blend reduced dst alpha
at anti-aliased icon edges, making them semi-transparent against the
desktop. RGB blending is unchanged, so X11/Windows/macOS are unaffected.
Delay opening the object-list filament dropdown on wxGTK until the editor
window is mapped, avoiding popup creation without a valid native toplevel.
Also set the GTK transient parent for dropdown popups created from data-view
cell editors.
* unify footer
adds "Done" Buttons,
standardize "Reset" Buttons,
adds disabled logic to "Reset" Buttons
* adds ImGuiWrapper::COL_WARNING = ColorRGB:Warning,
replaces ColorRGB::Warning
* consistently display warnings,
moves assermbly warning after footer,
make cut warnings warning color
* adds separator above cut warnings
* adds tooltip entry in move, rotate & scale for auto-drop
* use wrapper tooltip in wrapper->button for consistent text color
* restructures brim ears button,
renames "Reset" button in painting tools to "Erase All",
small styling fixes
* small adjustments,
moves "Edge detection" checkbox in painting tool underneath,
moves "Section View" in Brim ears tool after buttons,
moves warnings in Assemble tool under footer
* small brim ear tool structure changes,
adds separator between "Head Diameter" and "Max angle",
adds tooltip to "Auto-generate" button,
rewords shortcut descriptions to match other tools
---------
Co-authored-by: Hanno Witzleb <hannowitzleb@gmail.com>
* adds UI and preview plane for multiple dovetail cuts
* adds multiple dovetail cuts to perform_with_groove()
* adds ui text info for spacing, adjusts max_val for gap to respect plate dimensions
* adds spacing before multpile UI
* adjusts wording, adjust gap max by count
---------
Co-authored-by: Hanno Witzleb <hannowitzleb@gmail.com>
Co-authored-by: SoftFever <softfeverever@gmail.com>
* Add logic to handle the left, middle and right buttons being assigned to do nothing, pan or rotate
* Add entries for setting the drag actions to preferences
* Allow the label text in preferences to wrap
* Show mouse mappings in Help -> Keyboard Shortcuts
* Re-add preferences in updated layout
* Add camera mouse options under camera
* Change mouse action strings to use L() for localization
* Display "None" when it is selected instead of blank in keyboard shortcuts
---------
Co-authored-by: Rob O <robertolabode@gmail.com>
Co-authored-by: Noisyfox <timemanager.rick@gmail.com>
[How to Download Pull Requests Artifacts for
Testing](https://www.orcaslicer.com/wiki/how_to_download_pr_artifacts)
This can be enabled in Preference->Developer->Keep painted feature after
mesh change.
<img width="731" height="633" alt="init"
src="https://github.com/user-attachments/assets/8b195486-538e-4eda-9e77-bfdf1a794306"
/>
TODO:
- [ ] Bug fixes
- [ ] Make it faster
- Keep painting after other mesh operations such as reload from
disk/simplify/boolean operation etc:
- [x] Planar cut
- [x] Dovetail cut
- [x] Cut with part assigned to other side
- [x] Split to parts
- [x] Split to objects
- [x] Mesh boolean gizmo
- [x] Mesh boolean in right click
- [x] Reload from disk/replace stl (won't work well if mesh changed too
much)
- [x] Fix model
- [x] Simplify/smooth (this two won't work well due to too much mesh
changes)
- [x] Add options in settings since I think this will be experimental
for a long time until being tested by a lot of ppl
* Expose Antialiasing velues
Expose Antialiasing multipliers.
Default to 4 as current implementation but enables the user to disable it to improve performante or increase sampling to improve quality.
* FXAA
* Improve descriptions
* Require restart when MSAA setting changes
Detect changes to the OpenGL MSAA (multisample anti-aliasing) preference when opening Preferences and prompt the user to restart the application to apply the change. Adds a constant key for the MSAA setting, stores the previous value, checks for changes (similar to the existing FXAA handling), and shows a warning dialog explaining the restart will close the current project without saving. If the user accepts, recreate_GUI is invoked to apply the MSAA change immediately.
* Revert "Require restart when MSAA setting changes"
This reverts commit dde134d346c3849598c91d025d2faed1b51c8a22.
* Menu and FPS options
* Fix FPS limiter and remove VSYNC
* Grouped FPS settings and mode up rigth fps counter
---------
Co-authored-by: Noisyfox <timemanager.rick@gmail.com>
FlowRateCalibrationDialog was missing the SetSizeHints call that every
other calibration dialog in calib_dlg.cpp uses (Cornering, Pa, Pa Pattern,
Max Volumetric Speed, Temperature, Retraction, VFA, Input Shaping). Without
it, the dialog has no enforced minimum size: on wxGTK, iconizing the main
window while the dialog is open triggers a layout/refresh that re-runs
Fit() with transient zero-sized children, collapsing the dialog to ~222x249
px with the OK button clipped and unreachable. The window manager then
honors the (incorrect) small geometry hint, so the user cannot resize it
back manually.
Two changes:
* Add v_sizer->SetSizeHints(this) after Fit() in the constructor, matching
the pattern used by all other dialogs in this file. This locks in the
correct minimum the first time Fit() runs, before any iconize event.
* Remove the redundant Fit() from on_dpi_changed. With SetSizeHints in
place the WM enforces the minimum, but the unconditional Fit() on every
DPI/refresh signal was the trigger for the shrink path; Refresh() alone
is sufficient here.
Co-authored-by: Packerlschupfer <packerl@schupfer.at>
* Add tooltips to Transfer buttons
Set contextual tooltips for the Discard, Transfer and Save buttons when a dependent preset is present. The code computes the previous and new profile names and provides localized explanatory tooltips that clarify that switching will discard changes, transfer "New Value" settings, or save changes into the current profile. This improves UX by making the consequences of each action explicit.
* added some endlines
Co-Authored-By: yw4z <yw4z@outlook.com>
---------
Co-authored-by: yw4z <yw4z@outlook.com>
* Remove unused `get_printer_preset` calls.
* `Plater::check_printer_initialized` should skip nozzle flow type checks if nozzle is unknown.
Also remove unnecessary `is_multi_extruders` check.
* Remove unused `CalibUtils::is_same_nozzle_diameters()` fn.
* Simplify `CalibUtils::check_printable_status_before_cali`.
The single-arity can delegate to the vectorized arity.
* Add `DevExtruderSystem::NozzleDiameterMatchesOrUnknown` to simply checks.
* Update `CalibrationPresetPage::update_sync_button_status()` to use `NozzleDiameterMatchesOrUnknown`.
Simplify logic by iterating over each extruder and checking for diameter and volume type match.
The previous code had several mistakes (from what I could tell):
- `curr_obj->is_multi_extruders()` doesn't imply exactly 2 extruders
- the single/multi branch served no purpose
- the single branch failed to check the volume type
* Specify `std::fabs` and add explicit import.
Ref: https://github.com/OrcaSlicer/OrcaSlicer/pull/13330#discussion_r3133613736
Not sure how idiomatic this is in C++ / OrcaSlicer codebase, but CoPilot suggested it and it seems reasonable.
---------
Co-authored-by: SoftFever <softfeverever@gmail.com>
* Support file uploads and the device details page for CC2 printers.
* Resolved build issues for Linux and macOS.
* 1. Added `ElegooPrinterWebViewHandler` to handle WebUI messages for Elegoo printers. Other printers will keep the current behavior.
2. Added a static `get_print_host_webui` method in `PrintHost` to retrieve the printer WebUI URL.
* Improved timeout handling for CC2 file upload and SN info APIs.
---------
Co-authored-by: SoftFever <softfeverever@gmail.com>
* perf: speed up startup and show progress in splash screen
Cold-start was ~15 s. Profiled in GUI_App::on_init_inner: 5 s in
preset_bundle->load_presets (sequential vendor loading of ~3000 JSON
profile files), 3.7 s in new MainFrame(), and a 1.5 s splash screen
that closed long before init finished — so the user stared at a
frozen blank screen for several seconds with no feedback.
Changes:
- PresetBundle::load_system_presets_from_json: parallelize vendor
loading with TBB. ORCA_FILAMENT_LIBRARY is loaded first
synchronously (it is the inheritance base for filaments); the
remaining vendors are loaded in parallel into separate PresetBundle
instances, then sequentially merged. On a typical setup this drops
load_presets from ~5 s to ~3.5 s; the saving scales with vendor
count and CPU cores.
- SplashScreen: remove the wxSPLASH_TIMEOUT flag so the splash stays
visible until the main window is shown explicitly, and add status
updates at each slow init phase ("Loading printer and filament
profiles", "Creating main window", "Loading current preset",
"Showing main window"). The splash is destroyed right after
mainframe->Show(true).
- MainFrame: FileHistory::LoadThumbnails moved to a detached
background thread. The previous synchronous call opened every recent
3MF file at startup — on macOS that triggered the TCC permission
dialog for ~/Downloads mid-launch, freezing the whole app until the
user clicked. Letting it run in the background means the UI is
responsive from the start and thumbnails populate when ready.
* revert: keep LoadThumbnails on the main thread (review feedback)
SoftFever pointed out the detached thread introduces race conditions
(e.g. MainFrame::open_recent_project() reading while the thread writes
m_thumbnails) and that thumbnails may not appear if the homepage shows
before it finishes. LoadThumbnails already uses tbb::parallel_for
internally, so the call-site offload gave no real speedup. Reverted to
the original synchronous call. The macOS TCC-dialog concern that
motivated this will be handled properly in the lazy-init refactor.
---------
Co-authored-by: SoftFever <softfeverever@gmail.com>
* bbl device selector search box
* align file and more menus with its button
* fix multiline inputbox border color on windows
* fix multiline inputbox border color on windows
* preferences: use content width on combobox dropdowns
* about: match version text size on macOS
* msg dialog improvements
* fix canvas menu overlapping with sliced plates toolbar
* bbl bind dialog button placement
* bbl color picker
* Update StepMeshDialog.cpp
* drop file dialog
* drop dialog revert fonts
* revert windows multiline border
* Update StepMeshDialog.cpp
* update
* Flushing Volumes: match style of combobox
* fix hyperlink color on canvas notifications
* fix possible issues with shared profiles notification
fix: eliminate UI freeze during cloud preset sync on startup
Move synchronous HTTP calls off the main thread in
GUI_App::start_sync_user_preset():
- Call scan_orphaned_info_files() and process_delete_presets() in
the background sync thread instead of the UI thread, so their
HTTP DELETE calls don't block startup.
- Call reload_settings() directly from the background thread
instead of via CallAfter, so get_user_presets() (HTTP GET) and
load/save_user_presets (file I/O) run off the main thread.
- Guard update_side_preset_ui() and app_config->save() with
is_main_thread_active() / CallAfter so they're safe when called
from a worker thread.
- Simplifiy finishFn lambdas: the progress-dialog case only
destroys the dialog; the no-dialog case is a no-op.
Two related fixes for LAN printing with newer Bambu networking plugins:
1. FileTransferUtils: guard against missing ft_tunnel_* symbols
The eMMC tunnel constructor calls ft_tunnel_create / ft_tunnel_set_status_cb
without checking if the plugin exported them. Older plugins (e.g.
01.10.01.01) don't have these symbols, so the calls dereference null and
crash when sending a print. Now throw a clear exception so callers can
fall back gracefully.
2. PrintJob: disable eMMC print path by default
Plugin 02.03.00.62's eMMC tunnel code hangs indefinitely at the upload
phase (30%) on Bambu H2D, blocked inside the plugin waiting on a future
that never resolves. Cancel doesn't work because the plugin doesn't
check cancel_fn during the wait. The plain FTP path works reliably.
Default to try_emmc_print=false; opt-in via AppConfig setting
"disable_emmc_print" = "0". Also wrap the eMMC tunnel creation in
try/catch so a missing-plugin exception cleanly falls back to FTP
instead of killing the worker thread.
Co-authored-by: hyiger <hyiger@gmail.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Co-authored-by: SoftFever <softfeverever@gmail.com>
Fix Windows single-instance file handoff
On Windows, opening a model file while OrcaSlicer is already running could
fail when single-instance mode was enabled.
The second launch correctly detected that another instance already existed,
but it could no longer find the running main window to forward the file-open
request. As a result, the new process exited and the file was dropped.
This happened because the existing-instance lookup depended on the native
window title containing "OrcaSlicer". After the titlebar change from #12706,
the title may contain only the project name, so the running instance was no
longer discoverable that way.
Fix this by identifying the existing main window through stable window
properties instead of title text:
- require the expected wxWidgets window class
- require the instance hash properties to exist
- match against the current instance hash
This restores forwarding of externally opened files to the already-running
instance without depending on the current titlebar text.
Hide tree support parameters not in use
- hide independent support layer height for organic tree support
- hide threshold overlap for tree supports
Co-authored-by: SoftFever <softfeverever@gmail.com>
* Fix speed popup refresh and sizing in Preview
Fix two UX issues in the Preview actual speed profile popup.
Keyboard-driven horizontal slider changes could leave the popup visually stale
until a later input event, which made keyboard navigation feel delayed.
Also, the speed-profile popup could briefly render at an incorrect height
before resizing on the next frame.
This change makes keyboard slider updates explicitly request a follow-up frame,
so refresh no longer depends on incidental mouse activity.
It also removes auto-resize from ToolPositionTableWnd and sizes the popup
before begin() each frame. The popup height is computed from plot height,
table rows, and window paddings, then clamped to at least the height of the
adjacent ToolPosition window.
The popup width is now computed from translated header text and padding terms,
with a 16:9-derived minimum for a more stable appearance. The width is cached
for the session and recalculated when UI scale changes.
Result:
- keyboard navigation updates reliably
- popup no longer shows the transient wrong-height first draw
- sizing is more stable across scale and localization
* Fix ToolPosition window also + other fixes
* Fix compile error when ENABLE_ACTUAL_SPEED_DEBUG disabled
IMGUI_DEFINE_MATH_OPERATORS is needed even if ENABLE_ACTUAL_SPEED_DEBUG is not enabled
* Guard seam fallback vertex index in GCodeViewer
Fix legend hover hit-testing
Fix GCode viewer legend where the eye icon and right edge don’t respond to hover/click unless you click slightly left (works after switching view types unless the vertical slide is used).
Fix hybrid tree first-layer support base flow and UI consistency
Fix incorrect first-layer generation for hybrid tree supports.
The support base on the first layer used brim flow instead of support material flow,
which could lead to incorrect extrusion behavior and wrong material usage in
multi-material prints.
Additionally, spacing and density for the regular support part were not consistently
derived from the actual first-layer support flow when first-layer and regular
line widths differed.
Hybrid-specific behavior is clarified:
- first-layer expansion applies only to the regular support part
- tree-only regions keep their existing behavior
The first-layer support pattern is aligned with normal and organic tree supports
to ensure consistent and meaningful density behavior.
Also make first-layer support expansion and density settings visible whenever
supports are enabled, as density was already affecting hybrid supports but was
hidden in the UI.
Avoid calling Show() on the main frame when it is already visible from BaseTransparentDPIFrame::on_hide(). On GTK this can re-enter size/layout handling after filament sync and leave the UI unresponsive.
