Tomáš Glozar

I see. I'll have a look at it :)

Thank you for the hints! Memory leaks should be fixed now.

> I suppose this is for operations which don't take _that_ long, but where it would still be too expensive to transfer everything through the browser wire back and forth?...

> **1 CockpitRedirectChannel does not respect channel freezes.** > > Freezing is implemented in CockpitTransport on received messages, for a given channel. I believe it is necessary to send all...

I believe this works. Let's say I want to redirect from channel "a" on the left privileged bridge (let's call it bridge 1) to channel "b" on the second privileged...

Changes: - `cockpit_peer_handle` and `CockpitChannel` now ref the `CockpitRedirect` object instead of taking ownership of it. - `CockpitChannelRedirect` now forwards data to the underlying transport via `cockpit_transport_emit_recv` instead of using...

Hi @stefwalter, Unfortunately this breaks the redirect mechanism, because the bridge where the redirecting channel is created doesn't have to know about the redirect channel, e.g. when redirecting a superuser...

So it turned out that the only reason the issue didn't appear in tests was the fact that they use `mock-bridge` as the peer, which doesn't use `CockpitRouter`. Anyway I...

> So far I had the impression that this is just a "fast path" optimization, but does not allow anything new that the JavaScript side could not already do. I.e....

> That's correct, although it encourages behavior that we may later prohibit. I would say that it not only encourages it, but it doesn't make much sense without it. For...