Commit graph

3,191 commits

Author SHA1 Message Date
Meta Vens
0bd1bac963 Python: pin what the unreadable-format and extra-slot refusals say and what the runtime reads
Mutation survivors on the merge's refusal text (the two readable formats, the
unrecognised-header description, the record-slot refusal) had no test pinning them.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 20:02:15 +02:00
Meta Vens
c9bd003bb3 CultMath intervals: comments name the gradient without the snoise_grad/fbm_grad tokens
The cut's negative check is a pattern over math.Interval.cs and CultMath.Interval.hlsl: the
bounds use values only, so neither file names the gradient functions. Comments only.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:57:01 +02:00
Meta Vens
b8dddc90d9 CultMath design.md and README: the Intervals section
Representation, the op list, the one-ulp widening rule and its two cases, the Lipschitz
provenance and pin, what the ball bounds guarantee and what they do not, tightness, the
saving and its open fork, and the consumers. Every number is one a committed test prints.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:56:30 +02:00
Meta Vens
cd7913bfcc CultMath intervals: enclosure harness, one-ulp shrink mutants, noise bounds, Lipschitz pin, saving
IntervalTests runs every op over 10,000 seeded intervals x 16 points with no tolerance, and
kills each op's one-ulp-shrunk lo and hi; exp's widening is an equivalent mutant on this CPU,
and smoothstep's is killed by a constructed non-monotone step. NoiseBoundTests encloses snoise
and fbm over 2,000 balls x 64 points and the bounded domain warp, pins SNOISE_LIPSCHITZ against
1e5 points and the ascent from MeasureLipschitz's witness, prints tightness, and prints the
march saving. The saving is 0.93x, not 2x: that test skips with the fork named.
MeasureLipschitz and MeasureMonotonicity are explicit, slow tests.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:56:30 +02:00
Meta Vens
bbf5269751 CultMath intervals: cultmath_iv_* over float2(lo, hi) in C# and HLSL, with snoise and fbm ball bounds
math.Interval.cs and shaders/CultMath.Interval.hlsl (included from CultMath.hlsl after Phacelle;
both tracked copies) add iv_point, iv_add, iv_sub, iv_neg, iv_mul, iv_scale, iv_abs, iv_min,
iv_max, iv_sqr, iv_sqrt, iv_exp, iv_clamp, iv_saturate, iv_smoothstep, iv_lerp, iv_snoise_ball
and iv_fbm_ball, and SNOISE_LIPSCHITZ = 10.099261 (1.10 x the refined max |snoise_grad| that
NoiseBoundTests.MeasureLipschitz prints). iv_exp and iv_smoothstep widen by one ulp through
asfloat(uint), added beside asuint as the HLSL intrinsic. The mirror test now compares 52
functions; the Unity package and the NuGet content files carry the new include.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:52:33 +02:00
Meta Vens
c5d9728b05 Semver policy: refusals, first-release declaration, unmeasured changes list
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:52:17 +02:00
Meta Vens
83e3f81263 Semver check: measuring nothing refuses; reference assemblies; tool restored alone
- A missing or moved baseline path, a previous tag that does not track the DLLs, and a built DLL
  with no namesake at the tag (unless declared with --api-new) refuse, naming what is missing.
- First release is declared with --first-release, never inferred: no previous tag, or tags that
  cannot be read, refuse without it.
- The "already published" skip is removed; a local tag of the released version exempts nothing.
- ApiCompat and the pinned NETStandard.Library.Ref 2.1.0 live in scripts/api-gate and are restored
  there, so the root manifest (dotnet-stryker) is no longer needed by a release. Both sides resolve
  netstandard through the reference assemblies.
- Tests: refusals, netstandard false breaks, cross-assembly references, restore diagnostics.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:52:14 +02:00
Meta Vens
d350e150b0 Rust: bring back R1's leading-bytes readers for leading_header
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:49:01 +02:00
Meta Vens
f35867a232 Rust: a newer store cut short is still named unsupported; tests follow the single reader
R1's rule that the header alone decides an unreadable-format store (a newer store
cut short is not damaged) is kept as leading_header, asked before the whole-file
proof. A directory at the store path now stops a write at the read as an I/O
error, so the two tests that used it to reach a failed rename assert that instead,
and the two c2a header tests assert R1's UnsupportedFormat kind, not c2a's old text.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:48:34 +02:00
Meta Vens
ff0251bbf8 Fix the R1 merge: close the Rust store_header visitor; a whole-store replace through a dangling link keeps replacing it
The merge left store_header's visitor unclosed (does not build). R3's pin that a
whole-store flush replaces a dangling link survives the single reader: the reader
still refuses the link as an I/O error, and the flush, which has no content to
carry from a link that holds none, writes v1 over it (Rust write_all_unlocked, TS
pushAll).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:46:39 +02:00
Meta Vens
bad511f5ac Merge CultCache R1 (f789c8d1) into C2a: one reader per runtime, R1's failure taxonomy
R1's rules win where both sides decided the same question: a zero-byte file is
not a store (refused as unreadable in every runtime), only nothing at the path is
empty, a dangling link is an I/O error, refusals never echo stored values and echo
a header only as cultcache.store.v<digits>, writes report Rejected, NotReplaced
and ReplacedNotDurable. C2a's rules win on what it added: the v3 element-id
header, variants, whole-file proof, typed unreadable errors in C#, TypeScript and
Python. Rust keeps R1's CultCacheStoreUnreadable{path, kind}; C2a's
StoreUnreadableError is deleted. The readability manifest's zero-byte row is
refuses x4.

Not yet built or tested: verification follows in the next commits.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:41:35 +02:00
Meta Vens
d0ea37f8e4 Research: prior art on cross-process file locking across runtimes and OSes
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:41:10 +02:00
Meta Vens
dbc2a6295b Imagination pass 2: the ground moves, the CSS wash is sampled from it, stardust and DoF mapped
Folds in the operator's rulings: Q1 cultmath-tapes, Q2 samples then
resolution, Q3 the shader owns a dynamic ground and the CSS wash is sampled
from its result. The nebula is flow-warped (two-phase flow map) with one
flow texture owning motion for the warp, the TAA motion vectors, the hue and
the stardust. Interval culling encloses the warp by enlarging the ball;
interval-ops gains that test. New fourth cut site-stardust: stateless
instanced particles plus an opacity-driven gather on the stardust layer, the
web form of the raymarch-driven DoF. Budget re-checked: the full stack lands
a mid-range phone at S = 4-6, not 8. Docs only.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-02 19:31:13 +02:00
Meta Vens
49e0fefb69 Imagination: map the gamecult.org ground shader as the proving ground for CultMath intervals
Three cuts under a proposed cultmath-tapes campaign: a textual HLSL -> GLSL
lowering pinned by test, interval ops with enclosure tests and a committed
Lipschitz constant for snoise, and the site's per-pixel interval-culled,
TAA-accumulated cloud ground driven by body tokens with the CSS wash as the
fallback. Docs only.

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
2026-10-02 19:16:19 +02:00
Meta Vens
32f5700b2a Semver check tests: double trailing slash, and a process started outside the target repo
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 19:03:00 +02:00
Meta Vens
15ccdbb953 Test: a staged write a commit landed is no longer staged
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 18:52:42 +02:00
Meta Vens
c49c76ee50 Research: prior art on records a writer does not own (formats, embedded stores, literature)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 18:51:58 +02:00
Meta Vens
a9741dbf57 Tests: pin the staged-key rule, the multi-key refusal and the carried record's stamp; a pull no longer clears staged keys
A pull runs only on a clean store, and a store is clean exactly when nothing is staged, so the pull's clear was redundant and is gone.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 18:47:47 +02:00
Meta Vens
84a34b8735 Semver check tests: pin the broken-manifest failure text
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 18:35:27 +02:00
Meta Vens
86eea40bce One owner for the foreign-record refusal: it reaches the cache, names every key, and refuses only what the cache wrote
CacheBackingStore.RefuseForeign now decides, for both store kinds, which written keys a foreign record refuses, drops the
store's staged change and the cache's copy of each through the Loaded path a pull uses, and throws one conflict naming every
refused key. The single-file store tracks the keys staged since its last load or write, as the directory store already did, so
a clean copy of a record another writer made foreign is carried, not refused.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 18:34:09 +02:00
Meta Vens
f40069dd88 Semver check tests: scope the cleanup assertion to the checker's own temp dirs, cover a broken manifest
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 18:34:05 +02:00
Meta Vens
6a3dae8954 Semver check: run tests from outside the repo, pin cleanup, usage and restore failure
The checker now runs with the target repo as its working directory in tests,
so the tool manifest is found from the script's own location as it is in the
release scripts. Adds tests for usage errors, a missing changelog, backslash
and trailing-slash baseline paths, temp-file cleanup, a restore failure and a
missing dotnet failing closed, the five-line failure excerpt and the
five-diagnostic cap. Drops two unreachable defaults and an unreachable
spawn-error branch from the checker.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 18:32:38 +02:00
Meta Vens
ed31a97689 Semver check tests: cache built fixture assemblies by content
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 18:01:37 +02:00
Meta Vens
bc15dba81b Test: the owner's-id case of the moved-record lay-back publishes the owner's catalog entry
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 18:00:52 +02:00
Meta Vens
0a98004bc4 Tests: pin the Stryker survivors in the lay-back and the directory refusal
The file's record moves when only its storedAt or only its schema id differs from the loaded one; a
staged removal is refused and dropped with the rest; a refusal leaves staged what it did not refuse.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 17:58:50 +02:00
Meta Vens
65bc144da0 Semver policy: describe the measured input and its failure rule
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 17:08:36 +02:00
Meta Vens
b8e1f7e0f6 Unity release scripts: run the semver check after the build, with the built DLLs
The check moves from before the build to once the assemblies under judgement
exist, and receives every shipped GameCult.* DLL (cultlib) or the fresh
CultMath.dll with the baseline path of the previous tag's tracked plugins.
The early calls are deleted; each script keeps exactly one.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 17:08:36 +02:00
Meta Vens
8f2eab7651 Semver check: behavioural tests for the measured public API
Real assemblies built in the test and compared through the CLI: removal
without ### Breaking, removal under too small a bump (post-1.0 and 0.y.z),
additions alone, a declared break with nothing measured, a removed GameCult.*
assembly, tool failure failing closed, inherited-member removals, an
already-published version, the baseline coming from the tag's blob, and a
first release.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 17:08:36 +02:00
Meta Vens
716ef9341c Semver check: measure the public API against the previous tag's DLLs
check-changelog-semver.mjs takes --api-baseline-path, --api-built and
--api-refs, extracts the previous tag's tracked DLLs as git blobs, and runs
the pinned Microsoft.DotNet.ApiCompat.Tool on each built assembly. Measured
breaks make the release breaking for the lane check and must be declared
under ### Breaking; a tool failure without a diagnostic refuses the release;
a GameCult.* assembly the tag tracked and the build lacks is a break.
Behavioural tests follow in the next commit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 17:03:09 +02:00
Meta Vens
614fd44584 Merge CultMath 0.3.0 release (Asura 2a-ii-c) cultmath-unity-v0.3.0
snoise moves to the upstream kernel (breaking values, minor bump for 0.y),
smin_grad, cellular, snoise/fbm/ridged gradients, Phacelle, BoundedLeastSquares.
Unity package licence is MIT AND MPL-2.0. Soul holds: asura:verdict:cut-2a-ii-c.s3.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 16:58:28 +02:00
Meta Vens
9e0f0ca36d CultMath 0.3.0: notices say the Unity package is MIT AND MPL-2.0, not MPL-2.0 as a whole
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 16:56:55 +02:00
Meta Vens
603d972f57 Single-file store: an unowned variant has no resolved document at load, so its in-place check is skipped
The previous commit serialized every unowned record's document at load, which a variant cannot give
until the cache resolves it (five tests failed on it). A variant persists its delta, never its
resolved document, so there is no in-place change to detect.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 16:54:41 +02:00
Meta Vens
b0f7d59c96 Directory store: a refused foreign overwrite leaves the pending set with its failed write
WriteGeneration left the refused key in the dirty and deleted sets, so every later generation
refused it. The refusal is now one check before any staging, shared by flush and commit, that
removes the refused keys from the pending sets. Tests cover both store kinds.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 16:51:54 +02:00
Meta Vens
c246e0c5ae Single-file store: lay back an unowned record only while the file still holds it as loaded
A whole-view write laid back the loaded bytes of an unclaimed record over whatever the file held,
reverting its owner's newer write and moving storedAt backwards. It now lays back only while the
file holds the loaded (schemaId, storedAt); a newer record is carried as the file has it and a
removed one is not written back. The v3 header is decided by records that are written as the cache
holds them, so a laid-back id-less record no longer marks the store. A document changed in place is
written as a write of the record, not dropped. A refusal drops the store's copy of the refused
records, so it no longer refuses every later write.

Does not build with the directory store untouched? It builds; the directory twin follows.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 16:51:54 +02:00
Meta Vens
8ab384448c CultMath 0.3.0: scope snoise constants to 3D in the changelog; Unity package is MIT AND MPL-2.0
2D snoise keeps r^2 0.5 and scale 130; only the permutation (+10) changed.
package.json license and LICENSE.md now match CultMath.csproj: MIT overall,
Phacelle files MPL-2.0 (operator ruling cultmath-license-mit-and-mpl).

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 16:48:16 +02:00
Meta Vens
98e3b91389 CultMath 0.3.0: rebuilt tracked DLL and pdb (Windows, Starfire)
Output of packages/cultmath/scripts/build-unity-package.ps1; reruns of the
script exit 0 with a clean tree.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 16:05:23 +02:00
Meta Vens
32ff786aed CultMath 0.3.0: version bump and changelog (snoise values break, minor bump)
Does not carry the rebuilt DLL yet; the Unity package script's version
parity check fails until the binary follows in the next commit.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 16:03:56 +02:00
Meta Vens
c2cd4202d2 Tape campaign: meshing defaults to surface nets plus the tile pass, not dual contouring
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 15:40:05 +02:00
Meta Vens
f7d4b63940 Park the CultMath tape campaign: interval-pruned field evaluation, option c
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-02 15:30:54 +02:00
Meta Vens
35ff3fbe42 Variants map: Soul holds C2a F2 (entry bytes, stale lay-back, stuck flush; main merge is its own cut)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-01 00:53:45 +02:00
Meta Vens
7007a17059 Variants map: Soul holds the cultcache-ts push identity fix (label match survives; Python and Rust share the bug)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-01 00:29:44 +02:00
Meta Vens
0a26f96ede Variants map: F2 fork A ruled (forbid slot reuse); fork B deferred as the next parity cut
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-10-01 00:08:21 +02:00
Meta Vens
9dcb3aa154 Merge put-serve bound: a put the server could never serve is refused, and refusals say why
CultMesh document servers (Rust, TS, Python) refuse a put whose served snapshot reply
could never fit the response bound, before storing it. Refusals carry fixed text per
variant, acknowledge nothing, go unreliable and unordered, and never echo peer input or
OS text; publishers and the snapshot client report a refusal instead of timing out. Send
failures follow Cut D; a reply the queue cannot take is answered, not dropped; a failed
send's goodbye names a fixed error code in every runtime. Python validates options as
Rust does and survives any unstorable put or malformed TCP message.

Soul passes 1-4 plus merge verdict; the final batch closes it. Verified on Yggdrasil at
8b8f8436 (main merged): every CultNet/CultMesh suite in Rust, TS, Python and C#, interop 27/27.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 23:40:50 +02:00
Meta Vens
8b8f8436cc cultmesh-ts: a snapshot refusal carries fixed text and acknowledges nothing
The snapshot fallback sent the caught error's message to the peer, and
errors from getCache or the registry went the same way, so a message the
application or a peer supplied could reach the wire. Its ack fields also
covered the refused request. The snapshot path now refuses per case with
fixed text (the snapshot source failed / the snapshot response is too
large / the snapshot response could not be queued) through refuse(),
which sends unreliable and unordered, so it needs no room in a full
queue, with ack fields that acknowledge nothing, as every runtime's
refusal does. onError keeps the detail.

Tests, with a raw peer that acknowledges nothing: a queue of 8 filled
exactly by an unacknowledged 8-fragment reply, then a refusal that still
arrives and acknowledges nothing; and a getCache failure carrying a
canary, refused with "the snapshot source failed".

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 23:06:03 +02:00
Meta Vens
e1bdf92321 C#: an unsendable goodbye names the failure by a fixed code
EndUnsendable put the SocketException's message on the wire, the
platform's wording. SendErrorCode names a permanent failure EMSGSIZE,
EINVAL or EAFNOSUPPORT, anything else UNKNOWN, so every runtime's
unsendable goodbye reads "packet could not be sent: CODE" with the same
codes.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 23:04:40 +02:00
Meta Vens
d541e9eb35 cultmesh-py: a refusal acknowledges nothing and names its own send failure; TCP handling refuses instead of dying; the daemon checks its port
Queue-full refusal (findings 2 and 6). The refusal was built in
_send_rudp_schema_frame before _end_refused_peer reset the session, so
its ack fields covered the refused request (ack=102 mask=0x3). A reply
the queue cannot take now raises _ReplyUnqueueable; the loop ends the
peer through _end_refused_peer, which builds the refusal before the
reset, gives it ack fields built after it (acknowledging nothing, like
the goodbye and like Rust's refusal), sends it unreliable and unordered,
then the goodbye. A refusal fragment that can never be sent is not
dropped: the goodbye names its failure ("packet could not be sent:
EMSGSIZE"), where it used to say the session refused a packet.

Queue full to the brim (item 1). The tests fill a queue of 8 exactly with
a reply the peer never acknowledges, then show the next reply is answered
with response_unqueueable, acknowledging nothing, followed by the
goodbye; and that a refusal whose fragment cannot be sent ends the
session with a goodbye naming EMSGSIZE.

TCP (finding 8). A message whose handling raised (snapshot recordKeys as
an int, a shard-log afterSequence that is not a number, subscribe,
schema-catalog and verse requests with int fields) killed the
connection's thread with a traceback that could quote the peer. The TCP
handler now refuses such a message with fixed text (message_unhandled),
logs the error type only, and serves the connection's next message.

Daemon (finding 10). --port 99999 is an argparse error (exit 2), not a
traceback.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 23:04:07 +02:00
Meta Vens
c3cc2b49c4 cultnet-py: an unsendable goodbye names the failure by a fixed code
end_unsendable_session put the OSError's text on the wire, which is the
platform's wording and can quote an address. send_error_code names a
permanent failure EMSGSIZE, EINVAL or EAFNOSUPPORT (Winsock equivalents
included), anything else UNKNOWN, as the Rust and TypeScript goodbyes
do; is_permanent_send_error is now "has a known code".

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 23:04:07 +02:00
Meta Vens
a71f62e997 cultmesh-rs: pin the refusal at a queue full to the brim, and a real EMSGSIZE
The queue-full test used an empty queue, so a reliable refusal fitted and
the reliable-refusal mutant lived. The new test fills a queue of 8
exactly with a reply the peer never acknowledges; the next reply cannot
be queued, and its refusal must still arrive, acknowledging nothing the
peer sent. A RawPeer helper drives the peer packet by packet; the gap
test uses it too.

A response too large for any datagram fails with the kernel's EMSGSIZE:
the rejection is ResponseSendFailed("EMSGSIZE"), the goodbye is exactly
"packet could not be sent: EMSGSIZE", and no refusal is sent.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 23:01:33 +02:00
Meta Vens
b8f754c898 cultmesh-rs: the snapshot client quotes only a refusal it recognises; the publisher doc says what holds
request_raw_snapshot_from_rudp_catalog put a catalog's error text into
its own error unfiltered. It now reports "refused the snapshot", naming
the refusal only when it is one of the fixed CultMesh refusal texts, as
the publisher does; both share recognised_refusal.

The publisher's doc claimed any refusing catalog is reported as a
refusal. That holds against a Rust catalog, which never acknowledges a
message it refuses; a catalog that acknowledges on receipt and refuses
afterwards (the TypeScript document server) is reported as having
admitted it. The doc now says so.

A fake catalog refusing with a canary is shared by the publisher test
and a new snapshot-client test.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 23:01:33 +02:00
Meta Vens
5ef988b1d0 cultnet-rs, cultmesh-rs: an unsendable goodbye names the failure by a fixed code
end_unsendable_session put the io::Error's Display on the wire: the
platform's wording, which differs by OS and by which packet failed.
cultnet-rs gains send_error_code, the fixed name of a failed send's
cause (EMSGSIZE, EINVAL, EAFNOSUPPORT, or UNKNOWN), matching the codes
the TypeScript goodbye already carries; is_permanent_send_error is now
"has a known code". The goodbye reads "packet could not be sent: CODE".

ResponseSendFailed carried an ErrorKind, which lost the cause (EMSGSIZE
has no kind and read as Uncategorized). It now carries the same fixed
code, and the session ends where the send failed, over the real error,
so the goodbye and the caller's rejection name the same failure.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
2026-09-30 23:01:33 +02:00