chore(deps): carry nova from git until the websocket frame-list fix releases - #349
Merged
Conversation
…eleases Hex nova 0.15.1 crashes the connection process when a websocket controller replies with a list of frames, which nova_websocket:call_result/0 documents as valid: nova_basic_handler:handle_ws/2 consed the payload onto the command list as one element, cowboy reads one command per element, so the whole list reached cow_ws:frame/2 as a single frame and died with function_clause. That is why asobi_ws_handler emits exactly one frame per websocket_info/2 return, and why #330 put the producing extension in the payload rather than dual-emitting game.message and module.message as two frames. Fixed upstream in novaframework/nova#400, merged and tagged v0.15.3 but not yet on Hex. Pinned to the merge commit rather than {branch, "master"} so the ref cannot drift underneath a build. The delta from hex 0.15.1 is exactly two commits - that fix and a cowboy bump which changes nothing here, since cowboy/cowlib/ranch resolve to the same 2.18.0/2.19.0/2.2.1 either way. Verified through the real handler chain: a {reply, [F1, F2], State} now yields two commands, both encodable by cow_ws:frame/2. Revert to plain hex nova once it releases - asobi#347.
🟡 Code Coverage — 72.9%5595 of 7679 lines covered. |
Taure
added a commit
that referenced
this pull request
Aug 4, 2026
…error (#354) * feat(ws): complete the S6 frame-type rename to module.message/module.error S6 put one extension (Lua) in the wire type, where no second extension could ever reuse it. #330 shipped the mechanism - the producing extension travels in the payload's `module` key - but not the rename, because nova 0.15.1 crashed the connection process on a list-valued reply, so dual-emitting old and new was impossible. novaframework/nova#400 fixed that and asobi carries the fixed nova by git ref (#349), so both frames can now go out on one reply. Extension pushes are `module.message` and `module.error`. `game.message` and `game.error` are emitted alongside them with identical payloads, so every SDK built before the rename keeps working, and are removed at the 1.0 wire break. `asobi.ws_legacy_game_frames` (default true) drops the legacy pair. `game.message` is asobi_lua's `game.send/2`, which a script may call per player per tick, so the compat frame doubles asobi's hottest extension-produced egress path. An operator whose clients all dispatch `module.*` gets that back without waiting for 1.0. Fixtures for both new types; the old two keep theirs. asobi_protocol_coverage_tests learns to read extension_frames/3, which is now the only emit site for either pair. * style: ASCII hyphen in the S6 test comment * docs: pin the wire-history note to v0.54.0
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Hex nova 0.15.1 crashes the websocket connection process when a controller replies with a list of frames, which
nova_websocket:call_result/0documents as valid.nova_basic_handler:handle_ws/2consed the payload onto the command list as a single element; cowboy reads one command per element, so the whole list reachedcow_ws:frame/2as one frame and died withfunction_clause, taking the connection with it.That is why
asobi_ws_handleremits exactly one frame perwebsocket_info/2return, and why #330 put the producing extension in the payload rather than dual-emittinggame.messageandmodule.messageas two frames.Fixed upstream in novaframework/nova#400, merged and auto-tagged v0.15.3, but nova's Hex is still on 0.15.1 and publishing is manual.
The pin
{nova, {git, "https://github.com/novaframework/nova.git", {ref, "a334a6db890d7ed1f1a3f69c4203ac1a84f3db56"}}},Pinned to the merge commit rather than
{branch, "master"}so the ref cannot drift underneath a build. This repo has been bitten by a moving pin before.The delta from Hex 0.15.1 is exactly two commits: the fix, and a cowboy 2.15 -> 2.18 bump that changes nothing here — asobi already forces
{cowboy, "~> 2.16"}andcowboy/cowlib/ranchresolve to the same2.18.0/2.19.0/2.2.1with and without this change. I checked rather than assumed, because a git override silently carrying unrelated commits has already caused a boot hang in this fleet once.Relocked from scratch, not
rebar3 upgrade nova.Verified
Through the real handler chain, not just the unit level:
Two frames, two commands, both encodable by
cow_ws:frame/2.eunit 1126/1126, xref, dialyzer,
fmt --checkall clean. CT deferred to CI: several agents were running Common Test against the same local Docker Postgres, which produces connection-closed failures unrelated to any change.What this unblocks
The S6 type rename (
game.message->module.message) can now be a clean two-frame transition instead of waiting for the 1.0 wire break.Follow-up
#347 tracks reverting to plain Hex
novaonce it releases. Note there that asobi_saas's nova override points at a branch commit for nova#390 (binary pubsub channels) which is not on master — the two overrides cannot be collapsed into one ref until #390 also lands.Also adds
*.coverdatato.gitignore; the new Coverage job leaves them in the working tree.