Turns out the nursery doesn't have to care about allocating a per task
`CancelScope` since the user can just do that in the
`@task_scope_manager` if desired B) So just mask all the nursery cs
allocating with the intention of removal.
Also add a test for per-task-cancellation by starting the crash task as
a `trio.sleep_forever()` but then cancel it via the user allocated cs
and ensure the crash propagates as expected 💥
As was listed in the many todos, this changes the `.start_soon()` impl
to instead (manually) `.send()` into the user defined
`@task_scope_manager` an `Outcome` from the spawned task. In this case
the task manager wraps that in a user defined (and renamed)
`TaskOutcome` and delivers that + a containing `trio.CancelScope` to the
`.start_soon()` caller. Here the user defined `TaskOutcome` defines
a `.wait_for_result()` method that can be used to await the task's exit
and handle it's underlying returned value or raised error; the
implementation could be different and subject to the user's own whims.
Note that by default, if this was added to `trio`'s core, the
`@task_scope_manager` would simply be implemented as either a `None`
yielding single-yield-generator but more likely just entirely ignored
by the runtime (as in no manual task outcome collecting, generator
calling and sending is done at all) by default if the user does not provide
the `task_scope_manager` to the nursery at open time.
Namely splitting the handles for each in 2 separate tables and adding
a `.cancel_service_task()`.
Also,
- move `_open_and_supervise_service_ctx()` to mod level.
- rename `target` -> `ctx_fn` params througout.
- fill out method doc strings.
Namely distinguishing service "IPC contexts" (opened in a
subactor via a `Portal`) from just local `trio.Task`s started
and managed under the `.service_n` (more or less wrapping in the
interface of a "task-manager" style nursery - aka a one-cancels-one
supervision start).
API changes from original (`piker`) impl,
- mk `.start_service_task()` do ONLY that, start a task with a wrapping
cancel-scope and completion event.
|_ ideally this gets factored-out/re-implemented using the
task-manager/OCO-style-nursery from GH #363.
- change what was the impl of `.start_service_task()` to `.start_service_ctx()`
since it more explicitly defines the functionality of entering
`Portal.open_context()` with a wrapping cs and completion event inside
a bg task (which syncs the ctx's lifetime with termination of the
remote actor runtime).
- factor out what was a `.start_service_ctx()` closure to a new
`_open_and_supervise_service_ctx()` mod-func holding the meat of
the supervision logic.
`ServiceMngr` API brief,
- use `open_service_mngr()` and `get_service_mngr()` to acquire the
actor-global singleton.
- `ServiceMngr.start_service()` and `.cancel_service()` which allow for
straight forward mgmt of "service subactor daemons".
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
Improve the `spawn` fixture teardown logic in
`tests/devx/conftest.py` fixing the while-else bug, and fix
`test_advanced_faults` genexp for `TransportClosed` exc type
checking.
Deats,
- replace broken `while-else` pattern with direct
`if ptyproc.isalive()` check after the SIGINT loop.
- fix undefined `spawned` ref -> `ptyproc.isalive()` in
while condition.
- improve walrus expr formatting in timeout check (multiline
style).
Also fix `test_ipc_channel_break_during_stream()` assertion,
- wrap genexp in `all()` call so it actually checks all excs
are `TransportClosed` instead of just creating an unused
generator.
(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
Adjust `basic_echo_server()` default sequence len to avoid the race
where the 'tell_little_bro()` finished streaming **before** the
echo-server sub is cancelled by its peer subactor (which is the whole
thing we're testing!).
Deats,
- bump `rng_seed` default from 50 -> 100 to ensure peer
cancel req arrives before echo dialog completes on fast hw.
- add `trio.sleep(0.001)` between send/receive in msg loop on the
"client" streamer side to give cancel request transit more time to
arrive.
Also,
- add more native `tractor`-type hints.
- reflow `basic_echo_server()` doc-string for 67 char limit
- add masked `pause()` call with comment about unreachable
code path
- alphabetize imports: mv `current_actor` and `open_nursery`
below typed imports
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Brief descriptions for both fns in `._discovery` clarifying
what each delivers and under what conditions.
(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
Remove the `trio.ClosedResourceError` and `trio.BrokenResourceError`
handling that should now be subsumed by `TransportClosed` re-raising out
of the `.ipc` stack.
Deats,
- drop CRE and BRE from `._streaming.MsgStream.aclose()/.send()` blocks.
- similarly rm from `._context.open_context_from_portal()`.
- also from `._portal.Portal.cancel_actor()` and drop the
(now-completed-todo) comment about this exact thing.
Also add comment in `._rpc.try_ship_error_to_remote()` noting the
remaining `trio` catches there are bc the `.ipc` layers *should* be
wrapping them; thus `log.critical()` use is warranted.
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Add `TransportClosed` to except clauses where `trio`'s own
resource-closed errors are already caught, ensuring our
higher-level tpt exc is also tolerated in those same spots.
Likely i will follow up with a removal of the `trio` variants since most
*should be* caught and re-raised as tpt-closed out of the `.ipc` stack
now?
Add `TransportClosed` to various handler blocks,
- `._streaming.MsgStream.aclose()/.send()` except blocks.
- the broken-channel except in `._context.open_context_from_portal()`.
- obvi import it where necessary in those ^ mods.
Adjust `test_advanced_faults` suite + exs-script to match,
- update `ipc_failure_during_stream.py` example to catch
`TransportClosed` alongside `trio.ClosedResourceError`
in both the break and send-check paths.
- shield the `trio.sleep(0.01)` after tpt close in example to avoid
taskc-raise/masking on that checkpoint since we want to simulate
waiting for a user to send a KBI.
- loosen `ExceptionGroup` assertion to `len(excs) <= 2` and ensure all
excs are `TransportClosed`.
- improve multi-line formatting, minor style/formatting fixes in
condition expressions.
(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
Internal helper which falls back to sync `pdb` when the
child actor can't reach root to acquire the TTY lock.
Useful when debugging tpt layer failures (intentional or
otherwise) where a sub-actor can no longer IPC-contact the
root to coordinate REPL access; root uses `.pause()` as
normal while non-root falls back to `mk_pdb().set_trace()`.
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Convert `spawn` fixture to a generator and add post-test graceful
subproc cleanup via `SIGINT`/`SIGKILL` to avoid leaving stale `pexpect`
child procs around between test runs as well as any UDS-tpt socket files
under the system runtime-dir.
Deats,
- convert `return _spawn` -> `yield _spawn` to enable
post-yield teardown logic.
- add a new `nonlocal spawned` ref so teardown logic can access the last
spawned child from outside the delivered spawner fn-closure.
- add `SIGINT`-loop after yield with 5s timeout, then
`SIGKILL` if proc still alive.
- add masked `breakpoint()` and TODO about UDS path cleanup
(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
A partial revert of commit c05d08e426 since it seem we already
suppress tpt-closed errors lower down in `.ipc.Channel.send()`; given
that i'm pretty sure this new handler code should basically never run?
Left in a todo to remove the masked content once i'm done more
thoroughly testing under `piker`.
For IPC-disconnects-during-teardown edge cases, augment some `._rpc`
machinery,
- in `._invoke()` around the `await chan.send(return_msg)` where we
suppress if the underlying `Channel` already disconnected.
- add a disjoint handler in `_errors_relayed_via_ipc()` which just
reports-n-reraises the exc (same as prior behaviour).
* originally i thought it needed to be handled specially (to avoid
being crash handled) but turns out that isn't necessary?
* hence the also-added-bu-masked-out `debug_filter` / guard expression
around the `await debug._maybe_enter_pm()` line.
- show the `._invoke()` frame for the moment.
Turns out we aren't clearing the `._state._runtime_vars` entries in
between `open_root_actor` calls.. This test refinement catches that by
adding runtime-vars asserts on the expected root-addrs value; ensure
`_runtime_vars['_root_addrs'] ONLY match the values provided by the
test's CURRENT root actor.
This causes a failure when the (just added)
`test_non_registrar_spawns_child` is run as part of the module suite,
it's fine when run standalone.
Move `_root_addrs` assignment to after `async_main()` unblocks (via
`.started()`) which now delivers the bind addrs , ensuring correct
`UnwrappedAddress` propagation into `._state._runtime_vars` for
non-registar root actors..
Previously for non-registrar root actors the `._state._runtime_vars`
entries were being set as `Address` values which ofc IPC serialize
incorrectly rn vs. the unwrapped versions, (well until we add a msgspec
for their structs anyway) and thus are passed in incorrect form to
children/subactors during spawning..
This fixes the issue by waiting for the `.ipc.*` stack to
bind-and-resolve any randomly allocated addrs (by the OS) until after
the initial `Actor` startup is complete.
Deats,
- primarily, mv `_root_addrs` assignment from before `root_tn.start()`
to after, using started(-ed) `accept_addrs` now delivered from
`._runtime.async_main()`..
- update `task_status` type hints to match.
- unpack and set the `(accept_addrs, reg_addrs)` tuple from
`root_tn.start()` call into `._state._runtime_vars` entries.
- improve and embolden comments distinguishing registrar vs non-registrar
init paths, ensure typing reflects wrapped vs. unwrapped addrs.
Also,
- add a masked `mk_pdb().set_trace()` for debugging `raddrs` values
being "off".
- add TODO about using UDS on linux for root mailbox
- rename `trans_bind_addrs` -> `tpt_bind_addrs` for clarity.
- expand comment about random port allocation for
non-registrar case
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Ensure non-registrar root actors can spawn children and that
those children receive correct parent contact info. This test
catches the bug reported in,
https://github.com/goodboy/tractor/issues/410
Add new `test_non_registrar_spawns_child()` which spawns a sub-actor
from a non-registrar root and verifies the child can manually connect
back to its parent using `get_root()` API, auditing
`._state._runtime_vars` addr propagation from rent to child.
Also,
- improve type hints throughout test suites
(`subprocess.Popen`, `UnwrappedAddress`, `Aid` etc.)
- rename `n` -> `an` for actor nursery vars
- use multiline style for function signatures
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Test pkg-level init module and sub-pkg module logger naming
to better validate auto-naming logic.
Deats,
- create `pkg_init_mod` and write `mod_code` to it for
testing pkg-level `__init__.py` logger instance creation.
* assert `snakelib.__init__` logger name is `proj_name`.
- write `mod_code` to `subpkg/__init__.py`` as well and check the same.
Also,
- rename some vars,
* `pkg_mod` -> `pkg_submod`,
* `pkgmod` -> `subpkgmod`
- add `ModuleType` import for type hints
- improve comments explaining pkg init vs first-level
sub-module naming expectations.
- drop trailing whitespace and unused TODO comment
- remove masked `breakpoint()` call
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
Add assertions and comments to better test the reworked
implicit module-name detection in `get_logger()`.
Deats,
- add `assert not tractor.current_actor()` check to verify
no runtime is active during test.
- import `.log` submod directly for use.
- add masked `breakpoint()` for debugging mod loading.
- add comment about using `ranger` to inspect `testdir` layout
of auto-generated py pkg + module-files.
- improve comments explaining pkg-root-log creation.
- add TODO for testing `get_logger()` call from pkg
`__init__.py`
- add comment about first-pkg-level module naming.
(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
Overhaul of the automatic-calling-module-name detection and sub-log
creation logic to avoid (at least warn) on duplication(s) and still
handle the common usage of a call with `name=__name__` from a mod's top
level scope.
Main `get_logger()` changes,
- refactor auto-naming logic for implicit `name=None` case such that we
handle at least `tractor` internal "bare" calls from internal submods.
- factor out the `get_caller_mod()` closure (still inside
`get_logger()`)for introspecting caller's module with configurable
frame depth.
- use `.removeprefix()` instead of `.lstrip()` for stripping pkg-name
from mod paths
- mv root-logger creation before sub-logger name processing
- improve duplicate detection for `pkg_name` in `name`
- add `_strict_debug=True`-only-emitted warnings for duplicate
pkg/leaf-mod names.
- use `print()` fallback for warnings when no actor runtime is up at
call time.
Surrounding tweaks,
- add `.level` property to `StackLevelAdapter` for getting
current emit level as lowercase `str`.
- mv `_proj_name` def to just above `get_logger()`
- use `_curr_actor_no_exc` partial in `_conc_name_getters`
to avoid runtime errors
- improve comments/doc-strings throughout
- keep some masked `breakpoint()` calls for future debugging
(this commit msg was generated in some part by [`claude-code`][claude-code-gh])
[claude-code-gh]: https://github.com/anthropics/claude-code
That is when no `name` is passed to `get_logger()`, try to introspect
the caller's `module.__name__` and use it to infer/get the "namespace
path" to that module the same as if using `name=__name__` as in the most
common usage.
Further, change the `_root_name` to be `pkg_name: str`, a public and
more obvious param name, and deprecate the former. This obviously adds
the necessary impl to make the new
`test_sys_log::test_implicit_mod_name_applied_for_child` test pass.
Impl detalles for `get_logger()`,
- add `pkg_name` and deprecate `_root_name`, include failover logic
and a warning.
- implement calling module introspection using
`inspect.stack()/getmodule()` to get both the `.__name__` and
`.__package__` info alongside adjusted logic to set the `name`
when not provided but only when a new `mk_sublog: bool` is set.
- tweak the `name` processing for implicitly set case,
- rename `sub_name` -> `pkg_path: str` which is the path
to the calling module minus that module's name component.
- only partition `name` if `pkg_name` is `in` it.
- use the `_root_log` for `pkg_name` duplication warnings.
Other/related,
- add types to various public mod vars missing them.
- rename `.log.log` -> `.log._root_log`.
A bit of test driven dev to anticipate support of `.log.get_logger()`
usage such that it can be called from arbitrary sub-modules, themselves
embedded in arbitrary sub-pkgs, of some project; the when not provided,
the `sub_name` passed to the `Logger.getChild(<sub_name>)` will be set
as the sub-pkg path "down to" the calling module.
IOW if you call something like,
`log = tractor.log.get_logger(pkg_name='mypylib')`
from some `submod.py` in a project-dir that looks like,
mypylib/
mod.py
subpkg/
submod.py <- calling module
the `log: StackLevelAdapter` child-`Logger` instance will have a
`.name: str = 'mypylib.subpkg'`, discluding the `submod` part since this
already rendered as the `{filename}` header in `log.LOG_FORMAT`.
Previously similar behaviour would be obtained by passing
`get_logger(name=__name__)` in the calling module and so much so it
motivated me to make this the default, presuming we can introspect for
the info.
Impl deats,
- duplicated a `load_module_from_path()` from `modden` to load the
`testdir` rendered py project dir from its path.
|_should prolly factor it down to this lib anyway bc we're going to
need it for hot code reload? (well that and `watchfiles` Bp)
- in each of `mod.py` and `submod.py` render the `get_logger()` code
sin `name`, expecting the (coming shortly) implicit introspection
feat to do this.
- do `.name` and `.parent` checks against expected sub-logger values
from `StackLevelAdapter.logger.getChildren()`.
To start ensuring that when `name=__name__` is passed we try to
de-duplicate the `_root_name` and any `leaf_mod: str` since it's already
included in the headers as `{filename}`.
Deats,
- heavily document the de-duplication `str.partition()`s in
`.log.get_logger()` and provide the end fix by changing the predicate,
`if rname == 'tractor':` -> `if rname == _root_name`.
* also toss in some warnings for when we still detect duplicates.
- add todo comments around logging "filters" (vs. our "adapter").
- create the new `test_log_sys.test_root_pkg_not_duplicated()` which
runs green with the fixes from ^.
- add a ton of test-suite todos both for existing and anticipated
logging sys feats in the new mod.
Based on the impure template from `pyproject.nix` and providing
a dev-shell for easy bypass-n-hack on nix(os) using `uv`.
Deats,
- include bash completion pkgs for devx/happiness.
- pull `ruff` from <nixpkgs> to avoid wheel (build) issues.
- pin to py313 `cpython` for now.
Tyler Goodlet
Bd
Gud Boi