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
A release refuses a tag that would pin a release JitPack never built.PinsResolveGate asks JitPack for
every pin a cut module's tag carries on an upstream outside the release, and names the flag to add. On
2026-10-12 SDK v1.4.3, Studio v1.4.3 and dashboard v0.3.1 were cut pinning botmaker-shared v0.2.4, which never
built, and all three were broken for good.
A release refuses a module that does not compile against the releases its tag pins.BuildsAtPinsGate
compiles a clean worktree of the module with every upstream outside the release pinned to its newest tag, and
quotes the compiler. SDK v1.4.3 called a shared method no shared release had; CI, building against shared's main, was green.
Releasing shared forces the dashboard, which pins it; every pin is a forcing edge now, which a test derives
from the modules' pins.
A module whose tag would pin a newer upstream is released when named, not skipped as "only docs": its
published pom differs (ChangeKind.pinsMove).
A gate whose Maven run could not find an upstream's -SNAPSHOT in ~/.m2 skips with the install command
instead of refusing the module: the redirect-pointer, plugin, template and builds-at-pins gates all read the
failure through MissingSnapshot. It also names the missing artifact, not the project Maven was building, and
reads a miss Maven cached ("was not found in … during a previous attempt").
--no-wait-jitpack stops before tagging on top of an upstream JitPack already failed to build, asked once
per upstream this run tagged.