Replies: 19 comments 15 replies
|
Started on the Linux/libspnav phase: #635 Opened as a draft on purpose. It builds, is off by default with zero cost to the normal build (verified from clean configures both ways), links correctly against libspnav when the option is enabled, and I confirmed for real — this sandbox genuinely has no spacenavd running — that a build with the feature compiled in stays completely silent and doesn't crash when no daemon or device is present, which is the state almost every user of an opt-in build will actually be in. What I cannot verify without a device: which physical axis is which, sign, and sensitivity. I picked X/Y → pan, Z → zoom per the scope above, but that's untested against real hardware. Flagged clearly on the PR, with the constants isolated so fixing any of it is a one-line change. If anyone here owns a SpaceMouse/SpacePilot and is willing to build this and try it, that's exactly the missing piece before it's ready to come out of draft. |
|
Quick update: #635 now has a That's in preparation for phase 2 (Windows/macOS via 3DxWare), not phase 2 itself — I don't have the SDK or a Windows/macOS toolchain in this environment, so I couldn't compile, link, or run a single line of that code, and didn't want to ship platform code that's never been built as if it were verified. The seam means whoever does have that SDK can add a backend without touching the Linux path or the (already-tested) view-navigation logic. Details on the PR. |
|
Button-mappable actions are now in #635 too — generic across backends (any Unlike the pan/zoom motion path, the parts downstream of "a button was pressed" don't need real hardware to verify, and I tested them properly this time: the settings-backed button→action map, The one part that's still unverified is the same one as before: detecting an actual button press needs a real device or daemon I don't have. |
|
Topic related https://qelectrotech.org/forum/viewtopic.php?pid=15911#p15911 |
so.... I don't have a mouse to test either. Will it create any harm leaving it untested in the source or somebody to test some day? |
|
Extend 3D mouse (SpaceMouse/SpacePilot) support to Windows and macOS Goal Technical findings Platform Mechanism License / Driver FreeCAD: integrates NavLib for a full 3D viewport (camera matrix, frustum) — less transferable, designed for pure 3D. |
|
First step is up: #1022 makes the 3D mouse work in the element editor, the gap you found in #635. Could you try it with your SpacePilot Pro? On your plan: the backend interface and per-platform CMake choice already exist from #635. For Windows/macOS I suggest trying |
|
Second step: #1023 adds settings under Configuration > Souris 3D: pan and zoom speed, a dead zone, inverting each direction, and zooming by twisting the cap instead of pushing it. Speed no longer depends on how often the driver sends updates, which matters once Windows/macOS drivers are involved. It builds on #1022. If you try it with your SpacePilot Pro, the most useful thing to know is whether the default speeds feel right. |
|
@scorpio810 step 1 for Windows/macOS is up as a draft: #1025. QET can now read the 3D mouse directly over USB, with no 3Dconnexion driver. To check it against a real device, could you record your SpacePilot Pro on your Debian machine? The script walks you through about a minute of movements and button presses and saves one file. It sends nothing anywhere. Please attach that file here. The same script is in #1025 as spacemouse-capture.py#!/usr/bin/env python3
# Copyright 2006-2026 The QElectroTech Team
# This file is part of QElectroTech.
#
# QElectroTech is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 2 of the License, or
# (at your option) any later version.
#
# QElectroTech is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with QElectroTech. If not, see <http://www.gnu.org/licenses/>.
"""Record raw USB reports from a 3Dconnexion 3D mouse, for QElectroTech.
QElectroTech's Windows/macOS 3D mouse support reads the device directly
over USB, so it has to decode each device model's raw reports itself.
This records what your device sends while you make a few guided
movements, and saves it to one file you can attach to the discussion.
Nothing is sent anywhere.
sudo python3 spacemouse-capture.py # finds the device itself
sudo python3 spacemouse-capture.py --list # just show what it finds
sudo is needed because /dev/hidraw* is usually readable by root only.
spacenavd can keep running. If the recording comes out empty, stop it
(`sudo systemctl stop spacenavd`) and try again.
Only the standard library is used, so it runs on any Linux with Python 3.
"""
import argparse
import datetime
import glob
import json
import os
import platform
import select
import sys
import time
VENDORS = {0x046D: 'Logitech (older 3Dconnexion)', 0x256F: '3Dconnexion'}
STEPS = [
('rest', 'Do not touch the device.', 3),
('right', 'Push the cap to the RIGHT and hold it, then let go.', 4),
('left', 'Push the cap to the LEFT and hold it, then let go.', 4),
('away', 'Push the cap AWAY from you and hold it, then let go.', 4),
('toward', 'Pull the cap TOWARDS you and hold it, then let go.', 4),
('down', 'Press the cap DOWN and hold it, then let go.', 4),
('up', 'Lift the cap UP and hold it, then let go.', 4),
('twist_cw', 'TWIST the cap CLOCKWISE (seen from above) and hold, then let go.', 4),
('twist_ccw', 'TWIST the cap ANTICLOCKWISE and hold, then let go.', 4),
('tilt_away', 'TILT the cap AWAY from you and hold, then let go.', 4),
('tilt_right', 'TILT the cap to the RIGHT and hold, then let go.', 4),
('buttons', 'Press each button once, slowly, one at a time, in any order.', 15),
]
def find_devices():
"""Return [{hidraw, name, vendor, product, sysfs}] for 3Dconnexion devices."""
found = []
for sysdir in sorted(glob.glob('/sys/class/hidraw/hidraw*')):
try:
with open(os.path.join(sysdir, 'device', 'uevent')) as f:
uevent = dict(line.strip().split('=', 1) for line in f if '=' in line)
except OSError:
continue
# HID_ID=0003:0000256F:0000C635 (bus:vendor:product)
try:
_bus, vendor, product = (int(x, 16) for x in uevent.get('HID_ID', '').split(':'))
except ValueError:
continue
if vendor not in VENDORS:
continue
found.append({
'hidraw': '/dev/' + os.path.basename(sysdir),
'name': uevent.get('HID_NAME', '?'),
'vendor': '%04x' % vendor,
'product': '%04x' % product,
'sysfs': sysdir,
})
return found
def read_descriptor(sysdir):
try:
with open(os.path.join(sysdir, 'device', 'report_descriptor'), 'rb') as f:
return f.read().hex()
except OSError as e:
return 'unreadable: %s' % e
def record(fd, seconds):
"""Read every report arriving within `seconds`; return [[ms, hex], ...]."""
reports = []
start = time.monotonic()
while True:
left = seconds - (time.monotonic() - start)
if left <= 0:
return reports
ready, _, _ = select.select([fd], [], [], left)
if not ready:
continue
try:
data = os.read(fd, 64)
except BlockingIOError:
continue
if not data: # only a test FIFO with no writer does this
time.sleep(0.01)
continue
reports.append([round((time.monotonic() - start) * 1000, 1), data.hex()])
def main():
ap = argparse.ArgumentParser(description=__doc__.splitlines()[0])
ap.add_argument('--list', action='store_true', help='only list matching devices')
ap.add_argument('--device', help='hidraw path, if more than one device is found')
ap.add_argument('--descriptor', help=argparse.SUPPRESS) # testing without a device
ap.add_argument('--yes', action='store_true', help=argparse.SUPPRESS) # no Enter prompts
ap.add_argument('-o', '--output', help='output file (default: spacemouse-capture-<product>.json)')
args = ap.parse_args()
devices = find_devices()
if args.list:
for d in devices:
print('%(hidraw)s %(vendor)s:%(product)s %(name)s' % d)
if not devices:
print('No 3Dconnexion device found under /sys/class/hidraw.')
return 0 if devices else 1
if args.device:
dev = next((d for d in devices if d['hidraw'] == args.device),
{'hidraw': args.device, 'name': '?', 'vendor': '?', 'product': '?', 'sysfs': None})
elif len(devices) == 1:
dev = devices[0]
elif not devices:
sys.exit('No 3Dconnexion device found. Is it plugged in? (try --list)')
else:
sys.exit('Several devices found, pick one with --device:\n' +
'\n'.join(' %(hidraw)s %(name)s' % d for d in devices))
if args.descriptor:
with open(args.descriptor, 'rb') as f:
descriptor = f.read().hex()
elif dev['sysfs']:
descriptor = read_descriptor(dev['sysfs'])
else:
descriptor = 'unknown'
try:
fd = os.open(dev['hidraw'], os.O_RDONLY | os.O_NONBLOCK)
except PermissionError:
sys.exit('Permission denied on %s -- run with sudo.' % dev['hidraw'])
print('Recording from %s (%s, %s:%s).' % (dev['hidraw'], dev['name'], dev['vendor'], dev['product']))
print('For each step, press Enter, do the movement, and wait for the next prompt.\n')
result = {
'tool': 'spacemouse-capture.py 1',
'date': datetime.datetime.now(datetime.timezone.utc).isoformat(timespec='seconds'),
'system': platform.platform(),
'device': {k: dev[k] for k in ('name', 'vendor', 'product')},
'report_descriptor': descriptor,
'steps': [],
}
try:
for i, (key, text, seconds) in enumerate(STEPS, 1):
print('[%d/%d] %s' % (i, len(STEPS), text))
if not args.yes:
input(' Press Enter to start (%d s)... ' % seconds)
reports = record(fd, seconds)
print(' %d reports recorded.\n' % len(reports))
result['steps'].append({'step': key, 'instruction': text, 'reports': reports})
except KeyboardInterrupt:
print('\nStopped early -- saving what was recorded so far.')
finally:
os.close(fd)
out = args.output or 'spacemouse-capture-%s.json' % dev['product']
with open(out, 'w') as f:
json.dump(result, f, indent=1)
total = sum(len(s['reports']) for s in result['steps'])
print('Saved %s (%d reports in total).' % (out, total))
if total == 0:
print('Nothing was recorded. Try stopping spacenavd first: sudo systemctl stop spacenavd')
else:
print('Please attach this file to discussion #599. Thank you!')
return 0
if __name__ == '__main__':
sys.exit(main()) |
|
Step 2 is up as a draft: #1026 builds the Windows package with 3D mouse support, with no 3Dconnexion driver needed. Looking for a Windows tester with a SpaceMouse. A build is here until the next nightly replaces it: Installer_QElectroTech-0.200.1-qt6-r10004-f2f105191_x86_64-win64-1.exe. Please say whether panning and zooming work, and whether the view moves twice when 3Dconnexion's own driver (3DxWare) is installed. |
|
I don't have a mac or a 3D mouse. I do have windows but not in my build enviroment I'm building you a solution now, |
|
If you own a SpaceMouse/SpacePilot and can spare a few minutes: |
|
Thanks — checked it against the decoder. Every step reads correctly except Could you redo just the |
|
Also pushed That triples every step to ~12s (buttons to ~45s). Just the full run again is fine — no need to isolate |
|
Also still after that re-recording, on Linux with your SpacePilot PRO, whenever you get a chance: Same as before, just a firmer push on |
|
Full instructions, in case they're useful: Needs |
|
|
Found it: Follow the prompts, then attach the JSON it saves here. |
|
Hum, not working on macOS with 3DxWare software installed.. |
|
That's expected — Your Linux recording already covers what the decoder needs. There's no macOS build to test the actual backend against yet (Windows has one in #1026); macOS is next once the |
|
macOS build with hidapi backend enabled on script 9883570 |
|
Nice — that's the first macOS test possible for this. Could you try it with your SpacePilot Pro: does panning/zooming work, and does the view move twice if 3DxWare's own driver picks it up as well as the hidapi backend? |
|
The 3D mouse should not be detected on macOS; in QET it has no effect, but it does work in the 3DxWare software. |
|
Tested on macOS arm64 (with the hidapi build + brew install hidapi): the SpacePilot Pro is not detected by QET, but it is detected fine by 3DxWare. Additional diagnostics:
So the device is clearly visible at both the OS and hidapi levels — the problem isn't a macOS permission issue or a conflict with 3DxWare, but very likely a vendor ID filter in QET's backend. The SpacePilot Pro still uses the old Logitech VID (0x046D), a holdover from when 3Dconnexion belonged to Logitech — not the current 3Dconnexion VID (0x256F). In fact, the VENDORS = {0x046D: 'Logitech (older 3Dconnexion)', 0x256F: '3Dconnexion'}If QET's hidapi backend only filters on 0x256F, that would exactly explain this behavior: newer devices (VID 0x256F) would get detected, while older models like the SpacePilot Pro (VID 0x046D) would be silently skipped. |
|
The vendor filter isn't it: Likely cause: on macOS, hidapi opens devices exclusively by default, which fails while 3DxWare has the SpacePilot open. Could you quit 3DxWare completely, then start QET and move the puck? If it works, the fix is small. |
|
The 3D mouse works on Windows 11 Edit: (3DxWare not installed ) |
|
We need you run the test again, see comments in this thread above. The logging was too short, i've increased the logging time |
|
Please Shane merge you new version of spacemouse-capture.py. |
|
Done: #1035. The script keeps On your Mac: On Linux, same file: Please push firmly on |
Record your 3D mouse for QElectroTech (Linux, macOS, Windows)If you own a 3Dconnexion 3D mouse (SpaceMouse, SpacePilot, SpaceNavigator…), a two-minute recording helps QElectroTech read that model correctly. A script guides you through a few movements and button presses, then saves one Each step says what to do: press Enter, make the movement, and hold it until the next prompt. Push firmly. LinuxOpen a terminal: If macOSOpen Terminal (Applications → Utilities): No sudo. If
Rather not use Terminal? The double-click SpaceMouse Check app does the same in a window (instructions on that page). Windows
No administrator rights needed, and 3DxWare can keep running. Every systemThe file is saved in the folder you ran the script from, named |
|
No terminal? SpaceMouse Check now exists for all three systems. Download, double-click, press Start, then attach the file it saves here:
The page has the steps for each system. |
Uh oh!
There was an error while loading. Please reload this page.
Forum thread: https://qelectrotech.org/forum/viewtopic.php?id=2480
Problem
Users with 3Dconnexion 6-DOF input devices (SpaceMouse, SpacePilot / SpacePilot Pro, Navigator) have no way to use them for panning/zooming QET's diagram view — the natural mapping for a primarily-2D schematic editor. This came up in the forum's zooming thread: a maintainer owns a SpacePilot Pro and confirmed device support "need[s] a lot of work and coding" and isn't something there's currently bandwidth for, so it's stalled as an idea rather than a concretely scoped task.
What exists today
DiagramView::zoom(qreal zoom_factor)(diagramview.cpp:335-350) and the view's ownhorizontalScrollBar()/verticalScrollBar()(driven directly inDiagramView::wheelEvent,diagramview.cpp:661-687). Whatever feeds a 3D mouse's motion into QET only needs to call these two things — not reimplement navigation.DiagramView::wheelEventalready branches on input-source characteristics: it has agestures()check to treat trackpad wheel deltas differently from a physical mouse wheel (diagramview.cpp:665-680) — precedent for handling a different navigation-input source distinctly. A background 6-DOF device is continuous/ambient rather than tick-driven like a wheel event though, so it's better implemented as its own listener object callingzoom()/the scrollbars directly, rather than folded intowheelEventitself.option(...)(e.g.option(PACKAGE_TESTS "Build the tests" ON),CMakeLists.txt:52) — the same pattern would let 3D-mouse support be compiled in only when wanted, with zero effect on default builds or users without the device.Proposed scope (staged, Linux-first)
libspnav: integrate withspacenavd/libspnav(BSD-licensed, already the de facto standard other open-source CAD/3D tools — Blender, FreeCAD — use for this class of device, and vendor-neutral rather than requiring 3Dconnexion's proprietary SDK). A small listener polls translation/rotation deltas and maps them onto the existing primitives: X/Y translation →horizontalScrollBar()/verticalScrollBar()(pan), Z-axis push/pull →DiagramView::zoom()(zoom in/out) — the same two calls the wheel handler already uses, so no new navigation logic is needed, only a new input source feeding it.3DxWare) rather than a portable open library, a materially bigger lift (vendor driver dependency, separate integration code per platform). Flagging this explicitly as future work rather than bundling it into an initial PR, in line with the maintainer's own assessment in the thread that full device support is a large undertaking.Related, not proposed here
Button-mappable actions on the device (dedicated buttons for rotate/mirror/undo, etc.) are a natural follow-up once basic pan/zoom motion works, but out of scope for a first cut.
Happy to build the Linux/libspnav phase if the scope above sounds right.
All reactions