Conversation
polymorph-iroh: #40 (their jco host retired), #43 (parking-kernel adoption), #44 (event-driven endpoint wakeups — the jco-era polling latency workaround is gone); no WIT changes. Nested websocket pin moves past their #43/#45 (TCP_NODELAY on the wasmtime socket). deltic: past the always-on parking kernel (#71), A7 fault-loudness (#76), and the three fixes this bump shook loose or reviewed out: #78 (timer pollable re-arm + ceiling chunking), #79 (smoke-c0 follows our rename), #82 (jspi entry-hop lift atomicity + poisoned late-settle retirement — found by our M2 gate within minutes: the marked wasi imports flip the plain sync engine into jspi mode, and the entry-hop window let a concurrent feed-keys turn reuse tick's un-lifted list<list<u8>> return area; 'Trap: list too long' + assert cascade). wosh-side changes: - client-core drive_ssh: drain to quiescence on exit-status before concluding — exit can beat the final stdout through the engine's goroutine buffers when one flight carries data + exit together (the new upstream arrival coalescing does exactly that); MOSH CONNECT sat undrained while the glue reported failure. Strictly convergent: no further input exists after exit. - deno.json: URL-prefix import-map keys collapse the four sibling-pinned raw-URL deltic trees onto .deps/deltic — the sibling packages' standalone configs were pulling MULTIPLE deltic runtimes into the graph and the browser bundle (esbuild WitError2 renames = a latent instanceof/module-identity break; first sibling error would have trapped as 'unbranded throw'). Bundle 563 -> 490 KB, zero remote content. Rule documented in the header; upstream convergence note in TASK.md. - browser-smoke: pageerror handler prints stacks (kept from the debugging session; strictly better gate diagnostics). - setup.sh: their SKIP_NODE env is gone with their jco host. Gates (full sweep at final pins): proto 6/6; m1 c+go; spikes 6/6; m2; m3; m4; m5 = web + client-deno + browser-e2e; m6; m7.
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.
Updates both pins to current upstream mains and fixes everything the bump shook loose. Full narrative: README finding 26 (+ addendum).
Pins
f46a80d→d8fdd039: their jco host retired (ssh: authenticate-auto -- offer every method, let the server steer #40), deltic parking-kernel adoption (client: the dial is a silent black box until iroh's timeout #43), event-driven endpoint wakeups (page: dying tunnel reads 'connected' and swallows input until the transport timeout #44 — the jco-era polling latency workaround finally gone). No WIT changes; nested websocket pin gains TCP_NODELAY on the wasmtime socket.d25551c2→a18be734: past the always-on parking kernel (site: survive the component's bad moments, and its death #71) and A7 fault-loudness (site: chrome refinements from review #76), plus three upstream changes from this session's review/debugging: #78 (timer pollable re-arm + setTimeout-ceiling chunking), #79 (smoke-c0 follows our rename), #82 (jspi entry-hop lift atomicity + poisoned late-settle retirement).What the bump found (and where it was fixed)
just m2within minutes): the parking kernel's marked wasi imports flip our plain sync engine into jspi mode; the entry-hop window (core returned, lift pending, bracket released) admitted a concurrentfeed-keysturn that reusedtick's un-liftedlist<list<u8>>return area →Trap: list too long→ poison → assert cascade. Fixed upstream with a deterministic tick/clobber wat fixture; wosh reproducers (bundled, real traffic) 3/3 green after.drive_sshrace (M7 gate, deterministic post-bump): exit-status can beat the final stdout through the engine's goroutine buffers when one flight carries data+exit together —MOSH CONNECTsat undrained while the glue reported failure. Now drains to quiescence on exit (strictly convergent).deno.jsonpins@deltic/*to its own prerelease URL, and Deno applies package-local configs — the graph and browser bundle carried multiple deltic runtimes (esbuildWitError2renames; latentinstanceofbreak — first sibling error would have trapped as "unbranded throw"). Fixed with URL-prefix import-map keys collapsing all four pinned trees onto.deps/deltic; bundle 563→490 KB, zero remote content. Upstream convergence note in TASK.md.Gates (full sweep at final pins)
proto 6/6 · m1 (c+go) · spikes 6/6 · m2 · m3 · m4 · m5 (web + client-deno + browser-e2e) · m6 · m7 — all OK.