The old code was unnecessarily complicated, full of unused and/or
duplicated functions making it hard to understand what will actually
happen and for whom at runtime.
Since we only support a single HTTP backend this can be greatly simplified.
Now everything gets default options from a single place and only
functions to modify parts actually difffering across calls are exposed.
No HTTP3/QUIC support yet.
Note, allowing both here means we don't actually profit from HTTP2 multiplexing
due to Finch(? or maybe a dependency of Finch?) limitations. But it means we can
now interact with HTTP2-only instances (if such exist) and still may get minor
gains from header compression etc
Adventurous admins can change the config to allow only HTTP2,
thus profiting from multiplexing (but breaking federation with
HTTP1-only instances which are in fact observed to exist).
A pool timeout shorter than the receive timeout
makes race conditions leading to active connections
being killed more likely and laso just doesn’t make
much sense in general.
See: https://github.com/sneako/finch/pull/292
OTP builds to 1.15
Changelog entry
Ensure policies are fully loaded
Fix :warn
use main branch for linkify
Fix warn in tests
Migrations for phoenix 1.17
Revert "Migrations for phoenix 1.17"
This reverts commit 6a3b2f15b74ea5e33150529385215b7a531f3999.
Oban upgrade
Add default empty whitelist
mix format
limit test to amd64
OTP 26 tests for 1.15
use OTP_VERSION tag
baka
just 1.15
Massive deps update
Update locale, deps
Mix format
shell????
multiline???
?
max cases 1
use assert_recieve
don't put_env in async tests
don't async conn/fs tests
mix format
FIx some uploader issues
Fix tests
Use a custom tesla middleware instead of adapter helper function +
custom redirect middleware.
This will also fix "Client died before releasing the connection"
messages when the request pool is overloaded. Since the checkout is
now done after passing ConcurrentLimiter.
This is technically less efficient, since the connection needs to be
checked in/out every time the middleware is left or entered respectively.
But I don't think the nanoseconds we might lose on redirects
to the same host are worth the complexity.
- fix for gun worker termination in some circumstances
- pool for http clients (ex_aws, tzdata)
- default pool timeouts for gun
- gun retries on gun_down messages
- s3 upload timeout if streaming enabled
- `verify_fun` is not useful now
- use `customize_check_hostname` (OTP 20+ so OK)
- `partial_chain` is useless as of OTP 21.1 (wasn't there, but hackney/..
uses it)
When the application restarts (which happens after certain config
changes), the limiters are not destroyed, so `ConcurrentLimiter.new`
will produce {:error, :existing}
See https://bugs.erlang.org/browse/ERL-1260 for more info.
The ssl match function is basically copied from mint, except
that `:string.lowercase/1` was replaced by `:string.casefold`.
It was a TODO in mint's code, so might as well do it since we don't need
to support OTP <20.
Closes#1834
This patch refactors gun pooling to use Elixir process registry and
simplifies adapter option insertion.
Having the pool use process registry instead of a GenServer has a number of advantages:
- Simpler code: the initial implementation adds about half the lines of code it deletes
- Concurrency: unlike a GenServer, ETS-based registry can handle multiple checkout/checkin
requests at the same time
- Precise and easy idle connection clousure: current proposal for closing idle connections in
the GenServer-based pool needs to filter through all connections once a minute and compare their
last active time with closing time. With Elixir process registry this can be done
by just using `Process.send_after`/`Process.cancel_timer` in the worker process.
- Lower memory footprint: In my tests `gun-memory-leak` branch uses about 290mb on peak load (250 connections)
and 235mb on idle (5-10 connections). Registry-based pool uses 210mb on idle and 240mb on peak load
