| Age | Commit message (Collapse) | Author |
|---|
|
Before this commit frames could "cheat" by compressing data
below the limit which would get expanded above the limit.
Now Cowboy will stop decompressing data when the limit is
reached.
|
|
|
|
This can be used to limit the maximum frame size before
some authentication or other validation is completed.
|
|
|
|
Following the same strategy as Websocket described in
commit cbed21c383e4cebb7df5a0a8b81f18c1738bef3e
Gains are comparable as far as Websocket over HTTP/2
is concerned.
|
|
`perf` has shown that Cowboy spends a lot of time
cancelling and starting this timer. Instead of resetting
for every data received, we now only reset a field in the
state.
Before it was working like this:
- start idle timeout timer
- on trigger, close the connection
- on data, cancel and start again
Now it's working like this:
- start idle timeout timer for a tenth of its duration, with tick number = 0
- on trigger, if tick number != 10
- start the timer again, again for a tenth of its duration
- increment tick number
- on trigger, if tick number = 10
- close the connection
- on data, set tick number to 0
|
|
It benchmarks binary, ascii, mixed and japanese data
using Websocket and Websocket over HTTP/2.
HTTP/2 options get set to ensure that performance is
better than the default HTTP/2 options.
It switches to Gun and Ranch branches that include
fixes that are required for tests to complete successfully.
|
|
|
|
This includes Websocket over HTTP/3.
Since quicer, which provides the QUIC implementation,
is a NIF, Cowboy cannot depend directly on it. In order
to enable QUIC and HTTP/3, users have to set the
COWBOY_QUICER environment variable:
export COWBOY_QUICER=1
In order to run the test suites, the same must be done
for Gun:
export GUN_QUICER=1
HTTP/3 support is currently not available on Windows
due to compilation issues of quicer which have yet to
be looked at or resolved.
HTTP/3 support is also unavailable on the upcoming
OTP-27 due to compilation errors in quicer dependencies.
Once resolved HTTP/3 should work on OTP-27.
Because of how QUIC currently works, it's possible
that streams that get reset after sending a response
do not receive that response. The test suite was
modified to accomodate for that. A future extension
to QUIC will allow us to gracefully reset streams.
This also updates Erlang.mk.
|
|
|
|
|
|
Original fix by Ali Farhadi <[email protected]>.
|
|
To avoid having the connection get closed due to us taking
too long on unreliable environments like GitHub Actions.
|
|
This is caused by the timeout being 1s after the period.
When the CI environment is overloaded, sometimes the
timeout will trigger. We retry, knowing that the
timetrap will catch us if we retry too much.
|
|
GitHub Actions runners are not as good as self-hosted BuildKite
so some adjustments need to be made to timeouts and such.
|
|
This allows conditionally generating an etag.
|
|
|
|
|
|
LH: The tests received a lot of fixes and tweaking.
I also reworded the error message to be more concise.
|
|
Added many tests to ensure the right timeout is picked in
the appropriate situation. Should there be any issues
remaining we can add more tests.
|
|
Also crash if trying to push after a reply was sent.
|
|
|
|
It is now tested both via cowboy_req:read_body and
via cowboy_req:cast.
Removes a bad example from the guide of body reading
with period of infinity, which does not work.
|
|
While we are identified as a supervisor in the tree,
we no longer manage children processes at that point,
so do not need to trap exit signals. Users can still
enable trap_exit if they prefer to.
|
|
We must add it even if we don't end up compressing because
it indicates that we might. This indication doesn't mean
that the user agent's accept-encoding values will ever
result in content encoding being applied.
|
|
The previous behavior was to accept them and drop the
content-length header as per the RFC recommendation.
But since this behavior is not normal it is safer to
just reject such requests than risk security issues.
|
|
In the cowboy_compress_h stream handler.
Otherwise this could cause issues with caching, with the
etag being the same for compressed/uncompressed content.
Users that wish to send etags AND compress will have to
do it manually for the time being.
|
|
|
|
The read buffer was changed into an iovec to avoid doing
too many binary concatenations and allocations.
Decompression happens transparently: when decoding gzip,
the content-encoding header is removed (we only decode
when "gzip" is the only encoding so nothing remains).
We always add a content_decoded key to the Req object.
This key contains a list of codings that were decoded,
in the reverse order in which they were. Currently it
can only be empty or contain <<"gzip">> but future
improvements or user handlers may see it contain more
values.
The option to disable decompression was renamed to
decompress_enabled and defaults to true.
It is no longer possible to enable/disable decompression
in the middle of reading the body: this ensures that the
data we pass forward is always valid.
Various smaller improvements were made to the code,
tests and manual pages.
|
|
|
|
LH: Simplified the test a little.
|
|
A new option reset_idle_timeout_on_send has been added.
When set to 'true', the idle timeout is reset not only
when data is received, but also when data is sent.
This allows sending large responses without having to
worry about timeouts triggering.
The default is currently unchanged but might change in
a future release.
LH: Greatly reworked the implementation so that the
timeout gets reset on almost all socket writes.
This essentially completely supersets the original
work. Tests are mostly the same although I
refactored a bit to avoid test code duplication.
This commit also changes HTTP/2 behavior a little when
data is received: Cowboy will not attempt to update the
window before running stream handler commands to avoid
sending WINDOW_UPDATE frames twice. Now it has some
small heuristic to ensure they can only be sent once
at most.
|
|
|
|
The socket staying open meant that the graceful shut down
of the Cowboy listeners were waiting for the connections
to be closed gracefully (or a timeout). Closing explicitly
where it makes sense ensures we don't unnecessarily wait.
This commit removes a full minute in the run time of all
Cowboy test suites (minus examples).
|
|
Sending extra response prevented by terminating all streams except
the one currently executing.
LH: Reworded some variables to make what happens more obvious.
|
|
LH: Small tweaks and added an HTTP/1.0 test.
|
|
It seems that macOS GH runners don't do timeouts well.
|
|
LH: I have added a test that does both hibernate and timeout
and fixed a related issue. I also tweaked the docs and tests.
|
|
A future OTP release will use 'strict' code path by default.
This change ensures it works both for old and new OTP.
|
|
|
|
Doing so will let us notice when the connection is gone instead
of waiting for timeouts, at least in the cases where the remote
socket was closed properly. Timeouts are still needed in case
of TCP half-open problems.
This change means that the order of stream handler commands is
more important than before because socket errors may occur
during the processing of commands.
|
|
LH: I reworked the test a little and added the same test
for HTTP/2 so that both HTTP/1.1 and HTTP/2 get the issue
fixed.
|
|
|
|
LH: I have fixed issues in the PR and incorporated changes
from a sister PR by @djankovic (git author: Dom J). I also
made sure the UTF-8 files were readable without Chinese
fonts and added their downloading to the examples test suite.
|
|
|
|
Co-authored-by: Björn Svensson <[email protected]>
|
|
|
|
|
|
This makes req_SUITE's cert_undefined test work again.
|
|
The key order of serialised maps changed in OTP-26.
|
