GitHub's contributors list is built from commit authorship, so it only ever shows the handful of people who have pushed code. Almost everything this driver knows about the hardware came from somewhere else: from people who plugged in a wheel none of the maintainers own, ran a diagnostic that told them nothing useful, and reported the result anyway.
This file is where that work is recorded. It is ordered by what people did, not by how much they did.
If you are listed here and would rather not be, or the description is wrong, open an issue and it will be changed the same day.
Support for a wheel we cannot touch is written blind and confirmed by proxy. Every one of these people ran tests on hardware in their own home, often several rounds of them, and often getting nothing back for their trouble until the last round.
- fenixadam - the G923 Xbox edition on Logitech's own SDK, from the first capture to the release. His USB captures decoded the SDK's stream on that wheel (#81), his hardware runs confirmed the rotation-range and strength fixes (#82), his kernel traces named two separate crashes and the load-order race behind one of them (#90), and a full day of precise runs on the SDK route found four faults in a row that no machine here could have shown: a marker written under the wrong name, an SDK left believing a tiny rotation range, a proxy that re-entered the SDK, and an installer that kept old apps (#91). He then compared the result against Windows on the same car. AC Rally (#83) is his too. And after weeks of a freeze that left nothing in any log, he disabled his iGPU to see the panic on screen and photographed it, which named the stale reply buffer fixed in 0.42.5 (#128). The Windows captures in #134 and #135 then showed that the wire traffic is the same on both systems, which put the pedal inversion on the Windows side and closed a long thread with evidence.
- DaveDos95 - RaceRoom on an RS50 under SteamOS (#92): read the game's own profile file to rule out the first diagnosis, then produced the proxy logs that proved the real fault, a DirectInput force aimed straight north, and confirmed the fix with a binary built for a system that cannot build its own. The same runs were the first live confirmation of the RaceRoom telemetry decoder.
- simonr2k4 - the entire G923 Xbox edition. Force feedback and TrueForce both work on that wheel because of his testing, and neither existed before it (#27). He also found that the installer shipped a udev rule without the helper it calls, which left the wheel looking dead, and produced the Wine log that root-caused the 90 degree steering lock: the TrueForce SDK looking for a G HUB that does not exist under Proton. The stutter report and USB capture in #121 started the hardware session that found the SDK proxy silently blocking ACC and AC EVO.
- pokesl0w - the first confirmed Steam Deck install, and the G923 Xbox edition's button and pedal layouts, captured button by button from his own wheel (#68). Both of those layouts differ from the PlayStation edition's, which is something no amount of reading could have established here.
- Maanikko81 - found that the G923 reports its pedals inverted, with the evtest reading that made it provable rather than anecdotal (#67).
- adnanmur - reported the G923 Xbox mode-switch rule hard enough to stop a machine booting, with a diagnosis better than the bug deserved (#52).
- sixtysecondstosmash and gmlinux - G PRO Racing Wheel captures (#8), which is how that wheel's protocol was mapped.
- SandSeppel - the earliest reports on the project, including the RS50 TrueForce capture (#5) that the haptic work is built on, plus the first build failures (#1, #2), the reconnection bug (#6), and the original request for TrueForce streaming (#4).
- Mhytee - author of TF4ALL, whose independent analysis confirmed the G923 shares the RS50 and G PRO TrueForce stream protocol, and corrected our understanding of how Windows drives these wheels (#20).
- PeposCJ - contributed to the same protocol discussion, including the Dynamic OLED's transport and the decode of its layout descriptor from the wheel's own firmware.
- WnDTech - had their tester read the OLED layout descriptors from an RS50, which showed the field widths are self-describing rather than a table to hardcode (#20).
- fsfarmscaper - decoded the dual
clutch page (
0x8127) on a G923 Xbox edition by listening to the wheel instead of writing to it, with predictions logged before the runs that tested them, and shared the raw logs (#97); then came back with the power-cycle and Reset runs that settled the enable flag and persistence. The first layout for that page from any source.
- kangaro0 - an RS50 that never showed up on first power-up (#141). The kernel log in that report named the real cause, a front-panel USB port failing the link before any driver ran, which is now a troubleshooting entry for everyone after him.
- Ismael-Torresan - the only quantitative measurements the force engine has: a fixed push over a fixed travel, timed, for damper, friction and inertia on both the firmware's path and the engine's, twice (#72, #87). Those numbers are what the engine's condition forces will be calibrated against, and his wedge report on the Xbox edition found a detector that reset itself on every error reply.
- purpleferret - a G PRO through
the whole DirectInput proxy route on DiRT Rally 2.0
(#105): the report that put Assetto Corsa Rally on
the right route, then a run of precise logs and USB captures that found
four faults nothing here could have shown (the proxy never aimed at the
wheel, its errors going nowhere a Steam user can read,
/dev/uhidroot-only where uhid is not loaded, and a shifter's gears dropped for want of buttons past 32), and finally the debug trace proving the game sent the wheel nothing until it was given an identity the game knew. - Lorenzoo78 - the Assetto Corsa EVO stutter report under the launcher (#74).
- matthiasvegh - twelve issues and the most thorough reporting the project has had: rotation range having no effect (#10), inverted force feedback (#12), cumulative forces (#16), intermittent FFB spikes (#31), rev-light brightness reverting (#29), slow module init (#30), and packaging assumptions that only break outside Arch (#17, #18).
- gondezee - the SDK revision question (#21), a hat-switch direction bug (#22), and Arch-specific test notes (#23).
- andrewexton373 - kernel 7.0.9 build breakage (#24) and the Le Mans Ultimate pedal binding problem through the logi-ffb virtual wheel (#50).
- sharkforce13 - the G923 PS
edition's oops on unplug and on module unload
(#127), with the
modprobe -rreproduction that made it a bench bug rather than a hardware one.
Regressions that turned out not to be ours, established by people who did the work to prove it.
- Mesche900 - held everything else constant across three kernels to show that a force feedback regression was Debian's 6.12 stable branch rather than this driver, and reported the result that made an earlier observation of their own wrong (#53).
- AX3Lino and LuanVSO - reported that the driver interfered with Logitech mice (#7, #9). That is why the module is scoped to wheels and no longer shadows the in-tree driver for every other Logitech device.
- marvicdigital and sugituber - additional reports and confirmations.
Commit authors appear in
the contributors graph, including
aderumier, whose pull request is in the
tree, gonzaloadrio, who wrote the
RS H-Shifter support (#140) and its desktop-mode gate
routing, then the Racing Adapter detection, the accessory list in
wheel_accessory and the standalone handbrake and adapter on the Test page
(#139), and tested every revision on their own RS50,
including the hot-plug paths nobody here could try, and the measurement
that put the H-Shifter's Reverse on its real button; and
schiffy91, who cleared the nixpkgs
deprecation warnings from the flake (#146).
Upstream projects this driver is built on are credited in the README.