ocaml-h2/test/ at main · gazagnaire.org/ocaml-h2 · Tangled

gazagnaire.org / ocaml-h2

0

HTTP/2 framing, HPACK and connection state, I/O-free

0

ocaml-h2 / test

1 folder 18 files

at main

Thomas Gazagnaire nox-h2: let a 1xx be interim, not the answer 1d ago

nox-h2: let a 1xx be interim, not the answer on_headers took the first header block on a stream to be the response whatever its status, so an interim 1xx filled the response slot and the real response that followed was filed as trailers. The Eio adapter resolves a call on the first Response event, so a server sending 103 Early Hints -- an ordinary deployed thing, not a hostile peer -- answered the call with 103 and dropped the 200 carrying the result. RFC 9113 sec 8.1 lets a server send any number of interim responses before the final one, so the two are different events, not one event with a status to squint at. Client.event gains Informational; interim blocks arrive as that and leave the response slot empty, so Response now means the one final response and a caller awaiting it cannot latch an interim block. The adapter ignores Informational and leaves the promise pending. That also closes the other half of the same sentence: a HEADERS bearing END_STREAM with an informational status is malformed, which sec 8.1.1 makes a PROTOCOL_ERROR stream error. An interim response promises a final one, so ending the stream promises a response that can never arrive, and a caller would wait on a stream nobody will write to again. The check runs through a new ROLE.malformed hook, applied after the block is decoded rather than instead of it, so the HPACK context stays in step on a connection the stream error leaves alive.

1 day ago

nox-h2: let a 1xx be interim, not the answer on_headers took the first header block on a stream to be the response whatever its status, so an interim 1xx filled the response slot and the real response that followed was filed as trailers. The Eio adapter resolves a call on the first Response event, so a server sending 103 Early Hints -- an ordinary deployed thing, not a hostile peer -- answered the call with 103 and dropped the 200 carrying the result. RFC 9113 sec 8.1 lets a server send any number of interim responses before the final one, so the two are different events, not one event with a status to squint at. Client.event gains Informational; interim blocks arrive as that and leave the response slot empty, so Response now means the one final response and a caller awaiting it cannot latch an interim block. The adapter ignores Informational and leaves the promise pending. That also closes the other half of the same sentence: a HEADERS bearing END_STREAM with an informational status is malformed, which sec 8.1.1 makes a PROTOCOL_ERROR stream error. An interim response promises a final one, so ending the stream promises a response that can never arrive, and a caller would wait on a stream nobody will write to again. The check runs through a new ROLE.malformed hook, applied after the block is decoded rather than instead of it, so the HPACK context stays in step on a connection the stream error leaves alive.

1 day ago

h2: add the canonical field-block mapping

1 week ago

h2: add the canonical field-block mapping

1 week ago

h2: add the canonical field-block mapping

1 week ago

nox-h2: preserve typed client errors

1 day ago

test_client.mli

ocaml-h2: credit the upstream in each source file's copyright header Add the copyright header the fork inherited -- the upstream holder(s) from LICENSE.md and the matching SPDX identifier -- to every source file that lacked one, so the attribution is per-file, not only at the package level. Machine-generated code (EverParse C) is left untouched.

2 weeks ago

test_connection.ml

h2: scale driver for the connection stream map test_connection.ml gains a scalability section: the HTTP/2 analog of "millions of concurrent connections" is one connection holding a huge number of streams, so the stream map must be O(log n) to look up, O(n) in memory with a small per-entry constant, and self-reaping. O(log n) lookup: Connection keys streams with Map.Make (Int32); a structurally identical map over a comparison-counting Int32 measures the key comparisons a lookup costs, deterministically (no wall-clock). Doubling n adds a small constant, not a factor. Teeth: a List-backed connection scans O(n) keys -- pointing the measurement at an assoc list makes the O(log n) bound fail (100000 comparisons > 114). O(n) memory: a Stream.t record plus its Map node is a small, bounded per-entry cost. Reap to baseline: n exchanges driven to completion (END_STREAM both ways) leave active_stream_count at 0 and the map back at its baseline footprint, exercising the terminal-stream reap. Committed n is CI-sized (100_000). Measured at n = 1_000_000: lookup 35 key comparisons (bound 132), doubling to 2_000_000 gives 36; memory 24 words/stream (24_000_015 words for 1M streams); reap returns to baseline (retained ~0 words) with active 0 -- before the terminal-stream reap this retained ~24M words.

1 week ago

test_connection.mli

ocaml-h2: credit the upstream in each source file's copyright header Add the copyright header the fork inherited -- the upstream holder(s) from LICENSE.md and the matching SPDX identifier -- to every source file that lacked one, so the attribution is per-file, not only at the package level. Machine-generated code (EverParse C) is left untouched.

2 weeks ago

tree: reformat for the upgraded ocamlformat Comment reflow and header reindent produced by the new ocamlformat; no code changes.

1 month ago

ocaml-h2: credit the upstream in each source file's copyright header Add the copyright header the fork inherited -- the upstream holder(s) from LICENSE.md and the matching SPDX identifier -- to every source file that lacked one, so the attribution is per-file, not only at the package level. Machine-generated code (EverParse C) is left untouched.

2 weeks ago

docs: reflow comments across the tree

1 month ago

ocaml-h2: credit the upstream in each source file's copyright header Add the copyright header the fork inherited -- the upstream holder(s) from LICENSE.md and the matching SPDX identifier -- to every source file that lacked one, so the attribution is per-file, not only at the package level. Machine-generated code (EverParse C) is left untouched.

2 weeks ago

test_hpack_tables.ml

tree: reformat for the upgraded ocamlformat Comment reflow and header reindent produced by the new ocamlformat; no code changes.

1 month ago

test_hpack_tables.mli

ocaml-h2: credit the upstream in each source file's copyright header Add the copyright header the fork inherited -- the upstream holder(s) from LICENSE.md and the matching SPDX identifier -- to every source file that lacked one, so the attribution is per-file, not only at the package level. Machine-generated code (EverParse C) is left untouched.

2 weeks ago

h2: spec test that a GOAWAY names the last processed peer stream RFC 9113 6.8: the GOAWAY last-stream-id is "the highest-numbered stream identifier for which the sender of the GOAWAY frame might have taken some action on or might yet take action on". After the server processes a peer request on stream 3, both a connection-error GOAWAY (triggered here by an oversized frame, unrelated to any stream id) and a graceful Server.goaway must advertise last-stream-id 3, not 0. Advertising 0 tells the client none of its requests were acted upon, so it may safely retry a non-idempotent request the server already handled. Both tests fail today: the emitted GOAWAY carries 0.

1 week ago

test_server.mli

ocaml-h2: credit the upstream in each source file's copyright header Add the copyright header the fork inherited -- the upstream holder(s) from LICENSE.md and the matching SPDX identifier -- to every source file that lacked one, so the attribution is per-file, not only at the package level. Machine-generated code (EverParse C) is left untouched.

2 weeks ago

nox-h2: convert the connection core to a pure immutable state machine Connection.t, Stream.t and the Endpoint were a web of mutable fields and a Hashtbl of streams mutated in place. Make them immutable values whose transitions return a new state, per the ocaml-protocols pure-state-machine model: Stream/Connection mutators return the advanced value; streams move to an immutable Int32 Map; the HPACK codec (mutable dynamic tables + scratch) is lifted out of the connection into a borrowed Connection.hpack resource the adapter owns and passes into handle/respond/trailers; Server/Client v returns (t, hpack, hello) and every operation returns the advanced t. The frame loop threads the connection through a function-local ref so the public functions stay pure. Tests and fuzz thread state and pass the hpack resource; an adversarial WINDOW_UPDATE-flood fuzz checks the send window stays within the RFC 9113 sec 6.9.1 bound. All 173 unit tests and the fuzz suite pass.

1 month ago

test_stream.mli

ocaml-h2: credit the upstream in each source file's copyright header Add the copyright header the fork inherited -- the upstream holder(s) from LICENSE.md and the matching SPDX identifier -- to every source file that lacked one, so the attribution is per-file, not only at the package level. Machine-generated code (EverParse C) is left untouched.

2 weeks ago