Eyre's clog logic was a tad inconsistent about "only facts" vs "not poke-acks".
This makes it consistently say "only facts" when it comes to clog-related logic.
Yes, in theory this means %watch-acks and %kicks can build up endlessly, but
those should take up negligible space compared to %facts.
Should fix any oddball cases of crashes here that #3835 didn't already catch.
This was a little bit too crummy. Instead, we put in a placeholder of ~,
which should be forwards-compatible with atomic session identifiers,
where ~ identifies the default session.
Additionally touches up the herm wires/paths to stick to the above more
closely.
This lets us support the "random userspace app sending dill belts".
Ultimately, we'll want to be able to specify a session identifier
alongside the belt, instead of strictly relying on the duct.
Adds a %view task, which opens a subscription on the output sent to the
specified session. %flee closes the same.
Whenever dill sends a blit to the session, any subscribers get the
output also.
The structures here will become more reasonable once we replace ducts
with proper dill session identifiers.
People using older runtimes might not support the %klr blit. It's not
uncommon for prompts without style to get passed in as %pom though, so
here we catch that case and turn it into a %pro, which gets rendered as
a traditional %lin.
Pretty-printing is expensive, yet we do it whenever we construct the cookie
string, at least once (but usually twice) per authenticated request.
Here we call out the the specific to-tape functions we need, instead of relying
on the pretty-printer for converting... tapes to tapes, among other things.
The primary gains come from the cookie-related instances, we update the others
mostly for good style.
For the "receive request and immediately send response" case, that is processed
synchronously within eyre (ie, client sends channel ack), speeds thing up by
roughly 55%.
Motivation for the change is performance improvements on the un-`^~`d uses of
ream. Parsing turns out to be slow, making ream slow in turn. So we construct
the hoon ast manually instead.
!, is arguably better style than ream, since it doesn't require a ^~ for static
input, and lets syntax highlighting function properly.
For the investigated case, in +get-cast's +grow flow, improves performance by
over 80%.
If the Forwarded header specifies the original connection is secure,
update the flag to reflect that, regardless of whether the connection
directly to the urbit was made securely.
When an application would send multiple facts during a single event, it
was possible for the first fact to trigger a clog, removing the
subscription and sending a quit, but then the second fact still getting
sent out at normal.
Here, we drop any facts for subscriptions we don't have registered in
state, which should only happen in the described case.
Because storing in reverse order means producing in reverse reverse
order.
The tests didn't catch this because they, too, were infected with the
"reverse moves" meme.
In order to curb event queue growth when a client for whatever reason
isn't acking the events we send out, we implement a mechanism for
detecting such "clogging", and proactively kick subscriptions which are
adding too many events to the queue.
If the client hasn't sent an ack for ~s30, any subscription that accrues
more than 50 unacked %facts gets closed to prevent further buildup.
Upon reconnecting, the client will see %kick for the relevant
subscriptions and can open a new subscription as appropriate.
Includes a simple test for this behavior, and updates /app/dbug to be
able to display the newly tracked statistics.
By doing a %watch instead of %watch-as %json for channel subscriptions,
we can hopefully make better use of noun deduplication, when storing
events in a channel's event queue until they get acked.
Store the gall events from channel subscriptions as (vaseless) signs,
instead of serialized events. This should be smaller in memory, and
makes it more likely for noun deduplication to happen.
The cost is needing to reserialize upon channel reconnect, but this is
the less common case, and we don't expect it to be particularly slow.
In certain cases +find-merge-points was very slow. Specifically, the
`done` set was meant to avoid checking the same commit repeatedly, but
it didn't catch the case where a commit was added to the worklist that
was already in that worklist.
Secondly, the worklist was stored as a list but used as a queue, which
resulted in a lot of unnecessary welding. We change it to a qeu.
Fixes#3735
Instead of always providing a wildcard for the allowed methods and
headers, now echoes back the method and headers that the client asked
for, if any.
Fixes#3676.
