Matches the explicit `dict.pop(uid, None)` contract one
line above; same semantics as the prior truthy check.
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
(cherry picked from commit 0e3e008b0c)
(cherry picked from commit 13ed668512)
All 5 flagged items were valid (4 real bugs + 1 dead assert),
- fix an inverted `sys.version_info < (3, 14)` guard in
`ipc._linux` — the "`cffi` has no 3.14 support" import note now
fires on 3.14+ (where it applies) instead of on older pys.
- use `os.environ.get('PYTHON_COLORS')` in the `sync_bp` example
so it doesn't `KeyError` when run outside the test harness.
- correct `dump_task_tree()`'s docstring: the `/tmp` + `/dev/tty`
tee is gated on `write_file`/`write_tty`, not "unconditional".
- tidy the `ActorTooSlowError` message spacing in `cancel_actor`.
- replace a tautological `applied is True or applied is False` in
`test_patches` with `isinstance(applied, bool)` (the value is
order-dependent across the module).
Review: PR #462 (copilot-pull-request-reviewer)
https://github.com/goodboy/tractor/pull/462#pullrequestreview-4527179852
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Wrap `os.unlink()` in `close_listener()` with a `FileNotFoundError`
guard — under concurrent pytest sessions the sock-file can already be
reaped. Without this the raise aborts `_serve_ipc_eps`'s finally before
`_shutdown.set()`, deadlocking `wait_for_shutdown()` on
`actor.cancel()`.
Also,
- close each endpoint independently in the finally so one raise doesn't
strand the rest.
- always signal `_shutdown.set()` regardless of remaining ep count.
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
(cherry picked from commit 2ee44a6fdd)
Implements the resolution described in c99d475d's
`subint_forkserver_mp_shared_memory_issue.md` (now
updated with the resolution post-mortem). Two-part
fix that side-steps `mp.resource_tracker` entirely
rather than try to make it fork-safe — turns out
that's both simpler AND more correct given tractor
already SC-manages allocation lifetimes.
Deats,
- `tractor/ipc/_mp_bs.py::disable_mantracker()`: drop the
`platform.python_version_tuple()[:-1] >= ('3', '13')` branch — patches
now run unconditionally:
* monkey-patch `mp.resource_tracker. _resource_tracker` to a no-op
`ManTracker` subclass (empty `register` / `unregister`
/ `ensure_running`).
* return `partial(SharedMemory, track=False)` for the per-allocation
opt-out.
* belt + suspenders: even if something dodges the wrapper, the
singleton can't talk to the inherited (broken) parent fd.
- `tractor/ipc/_shm.py::open_shm_list()`: drop the 3.13+ conditional
skip of the unlink-callback; install a `try_unlink()` wrapper that
swallows `FileNotFoundError` (sibling-already-cleaned race in
shared-key setups). Without `mp.resource_tracker` doing it for us, we
own the unlink — `actor. lifetime_stack` is the right place since
tractor already controls actor lifecycle.
- `tests/test_shm.py`: uncomment-out `subint_forkserver` from the
module-level skip- list (tests pass now). Inline comment cross-refs
the two `_mp_bs` / `_shm` workarounds.
- `ai/conc-anal/subint_forkserver_mp_shared_memory_ issue.md`: heavy
rewrite — flips status from "open / unresolvable in tractor" to
"resolved, kept as decision record". Adds Resolution section, "Why
this is the right call" rationale (mp tracker is widely criticized;
tractor already owns lifecycle), trade-offs (crash-leaked segments,
lost mp leak warning), verification (7 passed under both
`subint_forkserver` and `trio` backends), and upstream issue links
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
(cherry picked from commit aa3e230926)
(factored: dropped subint_forkserver conc-anal doc update)
Since we're devving subints we require the 3.14+ stdlib API
and a couple compiled libs don't support it yet, namely:
- `cffi`, which we're only using for the `.ipc._linux` eventfd
stuff (now factored into `hotbaud` anyway).
- `greenback`, which requires `greenlet` which doesn't seem to be
wheeled yet
* on nixos the sdist build was failing due to lack of `g++` which
i don't care to figure out rn since we don't need `.devx` stuff
immediately for this subints prototype.
* [ ] we still need to adjust any dependent suites to skip.
Adjust `test_ringbuf` to skip on import failure.
Also project wide,
- pin us to py 3.13+ in prep for last-2-minor-version policy.
- drop `msgspec>=0.20.0`, the first release with py3.14 support.
(cherry picked from commit d2ea8aa2de)
A backlog of 1 caused `ECONNREFUSED` when multiple
sub-actors simultaneously connect to deregister from
a remote-daemon registrar. Now matches the TCP
transport's default backlog (~128).
Also,
- add cross-ref comments between
`_uds.close_listener()` and `async_main()`'s
`parent_is_reg` deregistration path explaining
the UDS socket-file lifecycle
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Adjust all imports to match.
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Drop the NIH (notinventedhere) custom parser (`parse_maddr()`,
`iter_prot_layers()`, `prots`/`prot_params` tables) which was never
called anywhere in the codebase.
Replace with a thin `mk_maddr()` factory that wraps the upstream
`multiaddr.Multiaddr` type, dispatching on `Address.proto_key` to build
spec-compliant paths.
Deats,
- `'tcp'` addrs detect ipv4 vs ipv6 via stdlib
`ipaddress` (resolves existing TODO)
- `'uds'` addrs map to `/unix/{path}` per the
multiformats protocol registry (code 400)
- fix UDS `.maddr` to include full sockpath
(previously only used `filedir`, dropped filename)
- standardize protocol names: `ipv4`->`ip4`,
`uds`->`unix`
- `.maddr` properties now return `Multiaddr` objs
(`__str__()` gives the canonical path form so all
existing f-string/log consumers work unchanged)
- update `MsgTransport` protocol hint accordingly
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Move the `Arbiter` class out of `runtime._runtime` into its
logical home at `discovery._registry` as `Registrar(Actor)`.
This completes the long-standing terminology migration from
"arbiter" to "registrar/registry" throughout the codebase.
Deats,
- add new `discovery/_registry.py` mod with `Registrar`
class + backward-compat `Arbiter = Registrar` alias.
- rename `Actor.is_arbiter` attr -> `.is_registrar`;
old attr now a `@property` with `DeprecationWarning`.
- `_root.py` imports `Registrar` directly for
root-actor instantiation.
- export `Registrar` + `Arbiter` from `tractor.__init__`.
- `_runtime.py` re-imports from `discovery._registry`
for backward compat.
Also,
- update all test files to use `.is_registrar`
(`test_local`, `test_rpc`, `test_spawning`,
`test_discovery`, `test_multi_program`).
- update "arbiter" -> "registrar" in comments/docstrings
across `_discovery.py`, `_server.py`, `_transport.py`,
`_testing/pytest.py`, and examples.
- drop resolved TODOs from `_runtime.py` and `_root.py`.
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Restructure the flat `tractor/` top-level private mods
into (more nested) subpackages:
- `runtime/`: `_runtime`, `_portal`, `_rpc`, `_state`,
`_supervise`
- `spawn/`: `_spawn`, `_entry`, `_forkserver_override`,
`_mp_fixup_main`
- `discovery/`: `_addr`, `_discovery`, `_multiaddr`
Each subpkg `__init__.py` is kept lazy (no eager
imports) to avoid circular import issues.
Also,
- update all intra-pkg imports across ~35 mods to use
the new subpkg paths (e.g. `from .runtime._state`
instead of `from ._state`)
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
- add `is_valid` and `sockpath.resolve()` asserts in
`get_rando_addr()` for the `'uds'` case plus an
explicit `UDSAddress` type annotation.
- rename no-runtime sockname prefixes from
`'<unknown-actor>'`/`'root'` to
`'no_runtime_root'`/`'no_runtime_actor'` with a proper
if/else branch in `UDSAddress.get_random()`.
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
I started getting annoyed by all the warnings from `pytest` during work
on macos suport in CI, so this replaces all `Actor.uid`/`Channel.uid`
accesses with `.aid.uid` (or `.aid.reprol()` for log msgs) across the
core runtime and IPC subsystems to avoid the noise.
This also provides incentive to start the adjustment to all
`.uid`-holding/tracking internal `dict`-tables/data-structures to
instead use `.msg.types.Aid`. Hopefully that will come a (vibed?) follow
up shortly B)
Deats,
- `._context`: swap all `self._actor.uid`, `self.chan.uid`,
and `portal.actor.uid` refs to `.aid.uid`; use
`.aid.reprol()` for log/error formatting.
- `._rpc`: same treatment for `actor.uid`, `chan.uid` in
log msgs and cancel-scope handling; fix `str(err)` typo
in `ContextCancelled` log.
- `._runtime`: update `chan.uid` -> `chan.aid.uid` in ctx
cache lookups, RPC `Start` msg, registration and
cancel-request handling; improve ctxc log formatting.
- `._spawn`: replace all `subactor.uid` with
`.aid.uid` for child-proc tracking, IPC peer waiting,
debug-lock acquisition, and nursery child dict ops.
- `._supervise`: same for `subactor.uid` in cancel and
portal-wait paths; use `actor.aid.uid` for error dict.
- `._state`: fix `last.uid` -> `last.aid.uid` in
`current_actor()` error msg.
Also,
- `._chan`: make `Channel.aid` a proper `@property` backed
by `._aid` so we can add validation/typing later.
- `.log`: use `current_actor().aid.uuid` instead of
`.uid[1]` for actor-uid log field.
- `.msg.types`: add TODO comment for `Start.aid` field
conversion.
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Per the questionable `copilot` review which is detailed for follow up in
https://github.com/goodboy/tractor/issues/418. These constants are
directly linked from the kernel sources fwiw.
Though it was a good (vibed) try by @dnks, the previous "fix" was not
actually adding unix socket support but merely sidestepping a crash due
to `get_peer_info()`'s impl never going to work on MacOS (and it was
never intended to).
This patch instead solves the underlying issue by implementing a new
`get_peer_pid()` helper which does in fact retrieve the peer's PID in
a more generic/cross-platform way (:fingers_crossed:); much thanks to
the linked SO answer for this solution!
Impl deats,
- add `get_peer_pid()` and call it from
`MsgpackUDSStream.get_stream_addrs()` when we detect a non-'linux'
platform, OW use the original soln: `get_stream_addrs()`.
- add a new case for the `match (peername, sockname)` with a
`case (str(), str()):` which seems to at least work on macos.
- drop all the `LOCAL_PEERCRED` dynamic import branching since it was
never needed and was never going to work.
Same problem as for the `ShmArray` tokens, so tweak and reuse
the `_shorten_key_for_macos()` helper and call it from
`open_shm_list()` similarly.
Some tweaks/updates to the various helpers,
- support `prefix/suffix` inputs and if provided take their lengths and
subtract them from the known *macOS shm_open() has a 31 char limit
(PSHMNAMLEN)* when generating and using the `hashlib.sha256()` value
which overrides (for now..) wtv `key` is passed by the caller.
- pass the appropriate `suffix='_first/_last'` values for the `ShmArray`
token generators case.
- add a `prefix: str = 'shml_'` param to `open_shm_list()`.
- better log formatting with `!r` to report any key shortening.
Adapt the `PSHMNAMLEN` fix from `piker.data._sharedmem` (orig commit
96fb79ec thx @dnks!) to `tractor.ipc._shm` accounting for the
module-local differences:
- Add `hashlib` import for sha256 key hashing
- Add `key: str|None` field to `NDToken` for storing
the original descriptive key separate from the
(possibly shortened) OS-level `shm_name`
- Add `__eq__()`/`__hash__()` to `NDToken` excluding
the `key` field from identity comparison
- Add `_shorten_key_for_macos()` using `t_` prefix
(vs piker's `p_`) with 16 hex chars of sha256
- Use `platform.system() == 'Darwin'` in `_make_token()`
(tractor already imports the `platform` module vs
piker's `sys.platform`)
- Wrap `shm_unlink()` in `ShmArray.destroy()` with
`try/except FileNotFoundError` for teardown races
(was already done in `SharedInt.destroy()`)
- Move token creation before `SharedMemory()` alloc in
`open_shm_ndarray()` so `token.shm_name` is used
as the OS-level name
- Use `lookup_key` pattern in `attach_shm_ndarray()`
to decouple `_known_tokens` dict key from OS name
(this patch was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Move the multi-platorm-supporting conditional/dynamic `socket` constant
imports to *after* the main cross-platform ones.
Also add constant typing and reformat comments a bit for the macOS case.
Make socket credential imports platform-conditional in `.ipc._uds`.
- Linux: use `SO_PASSCRED`/`SO_PEERCRED` from socket module
- macOS: use `LOCAL_PEERCRED` (0x0001) instead, no need for `SO_PASSCRED`
- Conditionally call `setsockopt(SO_PASSCRED)` only on Linux
Fixes AttributeError on macOS where SO_PASSCRED doesn't exist.
🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-Authored-By: Claude <noreply@anthropic.com>
Expand and clarify the comment for the default `case _`
block in the `.send()` error matcher, noting that we
console-error and raise-thru for unexpected disconnect
conditions.
(this patch was suggested by copilot in,
https://github.com/goodboy/tractor/pull/411)
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Flesh out missing method doc-strings, improve log msg formatting and
assert -> `RuntimeError` for un-inited tpt layer.
Deats,
- add doc-string to `.send()` noting `TransportClosed` raise
on comms failures.
- add doc-string to `.recv()`.
- expand `._aiter_msgs()` doc-string, line-len reflow.
- add doc-string to `.connected()`.
- convert `assert self._transport` -> `RuntimeError` raise
in `._aiter_msgs()` for more explicit crashing.
- expand `_connect_chan()` doc-string, note it's lowlevel
and suggest `.open_portal()` to user instead.
- factor out `src_exc_str` in `TransportClosed` log handler
to avoid double-call
- use multiline style for `.connected()` return expr.
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Refine tpt-error reporting to include closure attribution (`'locally'`
vs `'by peer'`), tighten match conditions and reduce needless newlines
in exc reprs.
Deats,
- factor out `trans_err_msg: str` and `by_whom: str` into a `dict`
lookup before the `match:` block to pair specific err msgs to closure
attribution strings.
- use `by_whom` directly as `CRE` case guard condition
(truthy when msg matches known underlying CRE msg content).
- conveniently include `by_whom!r` in `TransportClosed` message.
- fix `'locally ?'` -> `'locally?'` in send-side `CRE`
handler (drop errant space).
- add masked `maybe_pause_bp()` calls at both `CRE` sites (from when
i was tracing a test harness issue where the UDS socket path wasn't
being cleaned up on teardown).
- drop trailing `\n` from `body=` args to `TransportClosed`.
- reuse `trans_err_msg` for the `BRE`/broken-pipe guard.
Also adjust testing, namely `test_ctxep_pauses_n_maybe_ipc_breaks`'s
expected patts-set for new msg formats to be raised out of
`.ipc._transport`.
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Unmask the CRE case block for peer-closed socket errors which already
had a TODO about reproducing the condition. It appears this case can
happen during inter-actor comms teardowns in `piker`, but i haven't been
able to figure out exactly what reproduces it yet..
So activate the block again for that 'socket already closed'-msg case,
and add a TODO questioning how to reproduce it.
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Use new implicit module-name detection throughout codebase to simplify
logger creation and leverage auto-naming from caller mod .
Main changes,
- drop `name=__name__` arg from all `get_logger()` calls
(across 29 modules).
- update `get_console_log()` calls to include `name='tractor'` for
enabling root logger in test harness and entry points; this ensures
logic in `get_logger()` triggers so that **all** `tractor`-internal
logging emits to console.
- add info log msg in test `conftest.py` showing test-harness
log level
Also,
- fix `.actor.uid` ref to `.actor.aid.uid` in `._trace`.
- adjust a `._context` log msg formatting for clarity.
- add TODO comments in `._addr`, `._uds` for when we mv to
using `multiaddr`.
- add todo for `RuntimeVars` type hint TODO in `.msg.types` (once we
eventually get that all going obvi!)
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Turns out we weren't despite the optional `stream_handler_nursery` input
to `Server.listen_on()`; fail over to the `Server._stream_handler_tn`
allocated during server setup in those cases.
So i need to either adjust the tests or figure out if/why this is needed
to avoid the crashing in `pikerd` i found when killin the chart during
a long backfill with `binance` backend..
Just like in the BRE case (for UDS) it seems when a peer closes the
(UDS?) socket `trio` instead raises a `ClosedResourceError` which we now
catch and re-raise as a `TransportClosed`. This again results in
`tpt.send()` calls from the rpc-runtime **not** raising when it's known
that the IPC channel is disconnected.
Shift around comments and expressions for better reading, assign
`tpt_closed` for easier introspection from REPL during debug oh and fix
the `MsgpackTransport.pformat()` to render '|_peers: 1' .. XD
Since it's merely a local-file-sys subdirectory and there should be no
reason file creation conflicts with other bind spaces.
Also add 2 test suites to match,
- `tests/ipc/test_each_tpt::test_uds_bindspace_created_implicitly` to
verify the dir creation when DNE.
- `..test_uds_double_listen_raises_connerr` to ensure a double bind
raises a `ConnectionError` from the src `OSError`.
Using `.get_stream_addrs()` such that we always (*can*) assign the peer
end's PID in the `._raddr`.
Also factor common `ConnectionError` re-raising into
a `_reraise_as_connerr()`-@cm.
As mentioned in prior testing commit, it can cause the worst kind of
hangs, the SIGINT ignoring kind.. Pretty sure there was never any reason
outside some esoteric multi-actor debugging case, and pretty sure that
already was solved?
Mostly adjusting indentation, noise level, and clarity via `.pformat()`
tweaks more general use of `.devx.pformat.nest_from_op()`.
Specific impl deats,
- use `pformat.ppfmt()/`nest_from_op()` more seriously throughout
`._server`.
- add a `._server.Endpoint.pformat()`.
- add `._server.Server.len_peers()` and `.repr_state()`.
- polish `Server.pformat()`.
- drop some redundant `log.runtime()`s from `._serve_ipc_eps()` instead
leaving-them-only/putting-them in the caller pub meth.
- `._tcp.start_listener()` log the bound addr, not the input (which may
be the 0-port.
- only generate a repr in `.from_addr()` when log level is >= 'runtime'.
|_ add a todo about supporting this optimization more generally on our
adapter.
- fix `Channel.pformat()` to show unknown peer field line fmt correctly.
- add a `Channel.maddr: str` which just delegates directly to the
`._transport` like other pass-thru property fields.
Such as the `MsgTransport.stream` and `.drain` attrs since they're
rarely that important at the chan level. Also start adopting
a `.<attr>=` style for actual attrs of the type versus a `<name>:
` style for meta-field info lines.
We already have the `.ipc` sub-pkg name so it seems a bit
redundant/noisy for a namespace path Bp
Leave an alias for the `Server` rn since it's already used in a few
other internal mods.. will likely rename later if everyone is cool with
it..
As much as is possible given we currently do some graceful
cancellation join-waiting on any connected sub-actors whenever an active
`local_nursery: AcrtorNursery` in the post-rpc teardown sequence of
`handle_stream_from_peer()` is detected. In such cases we try to allow
the higher level inter-actor (task) context(s) to fully cancelled-ack
before conducting IPC machinery shutdown.
The main immediate motivation for all this is to support unit testing
the `.ipc._server` APIs but in the future may be useful for anyone
wanting to use our modular IPC transport layer sin-"actors".
Impl deats,
- drop passing an `actor: Actor` ref from as many routines in
`.ipc._server` as possible instead opting to use
`._state.current_actor()` where abs needed; thus the fns dropping an
`actor` input param are:
- `open_ipc_server()`
- `IPCServer.listen_on()`
- `._serve_ipc_eps()`
- `.handle_stream_from_peer()`
- factor the above mentioned graceful remote-cancel-ack waiting into
a new `maybe_wait_on_canced_subs()` which is called from
`handle_stream_from_peer()` and delivers a
maybe-`local_nursery: ActorNursery` for downstream logic; it's this
new fn which primarily still needs to call `current_actor()`.
- in `handle_stream_from_peer()` also use `current_actor()` to check if
a handshake is needed (or if it was called as part of some
actor-runtime-less operation like our unit test suite!).
- also don't pass an `actor` to `._rpc.process_messages()` see how-n-why
below..
Surrounding ipc-server client/caller adjustments,
- `._rpc.process_messages()` no longer takes an `actor` input and
now calls `current_actor()` instead.
- `._portal.open_portal()` is adjusted to ^.
- `._runtime.async_main()` is adjusted to the `.ipc._server`'s removal
of `actor` ref passing.
Also,
- drop some server `log.info()`s to `.runtime()`
Namely transferring the `Actor` peer-`Channel` tracking attrs,
- `._peers` which maps the uids to client channels (with duplicates
apparently..)
- the `._peer_connected: dict[tuple[str, str], trio.Event]` child-peer
syncing table mostly used by parent actors to wait on sub's to connect
back during spawn.
- the `._no_more_peers = trio.Event()` level triggered state signal.
Further we move over with some minor reworks,
- `.wait_for_peer()` verbatim (adjusting all dependants).
- factor the no-more-peers shielded wait branch-block out of
the end of `async_main()` into 2 new server meths,
* `.has_peers()` with optional chan-connected checking flag.
* `.wait_for_no_more_peers()` which *just* does the
maybe-shielded `._no_more_peers.wait()`
Call it `handle_stream_from_peer()` and bind in the `actor: Actor` via
a `handler=partial()` to `trio.serve_listeners()`.
With this (minus the `Actor._peers/._peer_connected/._no_more_peers`
attrs ofc) we get nearly full separation of IPC-connection-processing
(concerns) from `Actor` state. Thus it's a first look at modularizing
the low-level runtime into isolated subsystems which will hopefully
improve the entire code base's grok-ability and ease any new feature
design discussions especially pertaining to introducing and/or
composing-together any new transport protocols.
Gud Boi
wygud
Tyler Goodlet