[tatsuy]

Summary

Fix loss of precision in the latitude/longitude sent to Xacti cameras in the GnssStatus DroneCAN message.

Classification & Testing (check all that apply and add your own)

• Checked by a human programmer
• Non-functional change
• No-binary change
• Infrastructure change (e.g. unit tests, helper scripts)
• Automated test(s) verify changes (e.g. unit test, autotest)
• Tested manually, description below (e.g. SITL)
• Tested on hardware
• Logs attached
• Logs available on request

Took several photos at the same spot while the GPS position was jittering slightly:

• Before: the EXIF longitude was identical in every photo
• After: the EXIF longitude differs between photos

Description

com.xacti.GnssStatus defines latitude and longitude as float64, but they were calculated as loc.lat * 1e-7.
This rounds the longitude to steps of about 1.5e-5° (~1.4 m).
This PR casts to double before dividing, so the full precision of the int32 location is kept.

[rmackay9]

thanks very much for this. I guess the fix is the cast to double?

It's fine but I guess we could stick with the, "* 1e-7" part instead of changing to a divide.

[tatsuy]

1e-7 is treated as a float (due to -fsingle-precision-constant) and becomes 1.0000000117e-7.
I built and tested each value using Claude and compared the results.

Expression	Longitude example (expected 139.7671250)	Issue
loc.lng * 1e-7	139.767135620	Rounded to steps of about 1.4 m
(double)loc.lng * 1e-7	139.767126633	The steps are gone, but there is always an offset of about 0.15 m to the east
(double)loc.lng / 1.0e7	139.767125000	No error

[github-actions]

Global Size Summary (Total Flash delta in bytes)

Built from 415e61c rebased onto 7246115

Board	Periph	tracker	blimp	bootloader	copter	heli	iofw	plane	rover	sub
DrotekP3Pro	_	_	_	*	_	_	_	_	_	_
Durandal	_	-4	-4	_	-4	-4	_	-4	-4	-4
Hitec-Airspeed	*	_	_	_	_	_	_	_	_	_
KakuteF7	_	*	*	_	*	*	_	*	*	*
MambaH743v4	_	*	*	_	*	*	_	*	*	*
MatekF405-CAN	_	_	_	*	_	_	_	_	_	_
MatekF405	_	*	*	_	*	*	_	*	*	*
MatekH743-bdshot	_	-4	-4	_	-4	-4	_	-4	-4	-4
Pixhawk1-1M	_	*	*	_	*	*	_	*	*	*
disco	_	*	*	_	*	*	_	*	*	*
f103-GPS	*	_	_	_	_	_	_	_	_	_

[AP-Review]

Automated review note — AI-generated (Claude+Codex), validated against the live diff. Please sanity-check before acting.
Verdict: ACCEPT

Reviewed at head 415e61c584.
Full report: https://firmware.ardupilot.org/Tools/APReview/DevCallReviews/PRReviews/ardupilot/ardupilot/34925/1.html#pr34925

This two-line fix looks good to merge: dividing the int32 coordinates by 1.0e7 in double precision gives the Xacti float64 lat/lon fields full 1e-7 deg precision, where the old float multiply lost up to about 1.8 m. The author's reply to the cast-only suggestion checks out too, since keeping * 1e-7 leaves a small fixed bias because the constant is still a float.

Note

• libraries/AP_DDS/AP_DDS_Client.cpp:705 (outside this PR): the DDS goal topic has the same lat * 1e-7 float conversion into GeoPoint doubles. A follow-up PR could use the same (double)x / 1.0e7 form. This does not block this PR.

Not fully reviewed: behaviour on a real Xacti camera or DroneCAN bus, and builds for single-precision-FPU boards, were not exercised.

[peterbarker]

LGTM.

This is only one of many, many places that would require the same fix.

-fsingle-precision-constant is a mistake IMO. Not when we introduced it, but now with new tooling and whatnot I think we can do better.