We were retrying failed kelvin upgrades as many times as we had apps
that needed to be suspended, because suspending an app triggers an
attempt to run the next kelvin upgrade. This suspends all those apps in
one batch move, and then tries the next kelvin upgrade only once at the
end.
Fixes#6407
Partially addresses #6285
Adds a "mode" to channels, which can be set to either %json (current
behavior) or %jam. For %jam channels, aside from the SSE framing, all
communication happens through @uw-encoded jammed nouns. This applies to
both outgoing channel events, as well as incoming channel requests.
We choose @uw-style encoding because raw bytestreams are fragile and
cannot work inside the SSE stream context.
Currently, a separate endpoint (/~/channel-jam/etc) is used to indicate
%jam as the desired mode for a channel. We will probably want to make
this a bit cleaner, not least because it's not currently implemented as
a formal standalone endpoint, but also to give it stronger aesthetic
equivalence with the existing channel endpoint. Putting the mode in the
file extension is a tempting option here, but semantically not quite
right.
Connecting to the same channel across multiple modes is currently
supported, but it's untested, and unclear whether this is desirable or
not.
Marks it as deprecated in lull with a comment, and removes the verbosity
toggle state from dill. Filtering should now happen at the edges where
%crud error messages get printed.
We don't remove it from lull just yet, because that would necessitate a
kelvin bump, even though the rest of this changeset doesn't.
The previous changes implementing the /~/name endpoint were breaking,
since we changed the type of `$action:eyre`. This commit keeps the /~/name
endpoint functional, but adds adapters to eyre scries that returns the old
`$action:eyre` type. These adapters and their associated intermediate types
can be removed the next time we burn a kelvin.
These haven't been in use for a long time (if ever), but are now fully
deprecated: if you want to receive system output, see dill's %logs task
instead.
A dill %logs task can be used to open and close subscriptions to "system
output". Whenever dill receives a %text, %talk, or %crud task, it
considers this "system output", and passes it along to all %logs
subscribers.
This removes the logic from cleaning up stale subscriptions in %gall,
leaving +ap-rake as it was, and moves it to the +on-kroc arm in %ames.
Failed subscriptions from nacking a %watch plea that were
not properly corked (fixed in https://github.com/urbit/urbit/pull/6102)
are a subset of the more general "stale re-subscription" issue, so
we take care of all stale flows at the same time, by focusing on the
current subscription—leaving all others to be corked automatically—and
checking if it received a nack, to subsequently cork it.
This modifies the %rake task in %gall, to select what kind of
subscriptions we try to close:
=mode %o: kill old pre-nonce subscriptions
=mode %z: kill old pre-nonce subscriptions, including sub-nonce = 0
=mode %r: kills all stale resubscription flows
It also adds a dry-run option to both tasks (%kroc in ames, %rake in gall)
During pill and install prop generation.
In autopill, we stop tracking a copy of the blob store, instead just
having the pill lib scry it out on-demand.
