Repository navigation
Pipe v0.49.2
The /update result reaches every channel
- After
/updatethe result ("Pipe updated to ...") reached only the channel that reconnected to the restarted
backend first. It was usually Telegram, andpipe webgot nothing (the message is not in the dashboard inbox
either). It could also happen the other way round - Every event now carries the time it was published, and a client that reconnects says what it saw last: the backend
first sends what it missed.pipe webdoes it through the browser's standard SSE reconnect (Last-Event-ID, and per
device in the polling mode of/web), so it also gets alerts that came during a dropped connection. The Telegram bot
remembers its last event, and after its own restart (an update replaces its container too) it picks up the update
result without repeating older alerts - Protocol:
subscribetakes an optionalsince, and events have atsfield. Older clients work as before
Fixes
- The
pipe webbridge no longer waits in a connection's cleanup after it was interrupted. Such a connection could
linger and later close a socket that already belonged to another request, which gave an empty answer now and then
Images (linux/amd64, linux/arm64):
docker pull ghcr.io/kvmyk/pipe:0.49.2
docker pull ghcr.io/kvmyk/pipe-telegram:0.49.2
Signature (cosign, keyless):
cosign verify ghcr.io/kvmyk/pipe:0.49.2 \
--certificate-identity-regexp '^https://github.com/Kvmyk/Pipe/\.github/workflows/release\.yml@' \
--certificate-oidc-issuer https://token.actions.githubusercontent.com