You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
KiteUtils 0.14.0 (OpenSourceAWE/KiteUtils.jl#165, registration JuliaRegistries/General#172937) reverses the sign of azimuth_east: it now turns anti-clockwise seen from above, atan(y, x). That is its only breaking change. KiteGeometry pins KiteUtils = "0.13.1", so no package that depends on KiteGeometry can move to KiteUtils 0.14: SymbolicAWEModels needs to (OpenSourceAWE/SymbolicAWEModels.jl#430), and its resolve fails on this bound.
KiteGeometry never calls azimuth_east (checked on main: no match in src/), so the change is the bound alone: KiteUtils = "0.13.1, 0.14", the minimal resolve of the tracked manifest, a changelog line, and a patch release. Keep 0.13 in the bound so dependents can move one at a time.
Out of scope here: Transform.azimuth is documented as turning from east towards south (clockwise), the old azimuth_east sense. Whether that input convention should follow KiteUtils is being asked on OpenSourceAWE/SymbolicAWEModels.jl#430; do not change it in this task.
Widening the KiteUtils bound to "0.13.1, 0.14" (package and test env), with the minimal resolve of both tracked manifests and a changelog line. Independent of the open PRs (#71 touches calc_heading, not the bound). Leaving Transform.azimuth's convention alone, as asked; the patch release follows once this merges.
Activity
1-Bort-1 commentedon Oct 22, 2026
Widening the KiteUtils bound to "0.13.1, 0.14" (package and test env), with the minimal resolve of both tracked manifests and a changelog line. Independent of the open PRs (#71 touches calc_heading, not the bound). Leaving Transform.azimuth's convention alone, as asked; the patch release follows once this merges.