7a592d1c34
core: better Poller.__repr__
2018-05-24 23:30:56 +01:00
b0ce6eecd7
fork: support on_start= argument.
2018-05-23 16:25:17 +01:00
00edf0d66d
core: have ExternalContext accept a config dict rather than kwargs.
...
The parameter lists had gotten out of control.
2018-05-23 16:24:38 +01:00
eac4cc7afe
tests: speed things up
2018-05-18 10:58:25 +01:00
863a95e860
docs: update contributors
2018-05-18 10:58:25 +01:00
b0aa413173
tests: import benchmark reproduction setup playbook
...
Incomplete, also needs to mess around with Gcloud routing and handle box
setup/teardown, because that's another "well engineered" waste of time.
2018-05-18 10:58:25 +01:00
a99c7a12f9
tests: output split groups in gcloud.py
2018-05-15 23:41:02 +00:00
1a0e630a80
tests: add debops to requirements
2018-05-15 22:15:49 +01:00
dbcee4041a
tests: refactor gcloud.py to be dynamic inventory.
2018-05-15 22:10:36 +01:00
a3995f8e5f
ansible: remove hard-coded dw username.
2018-05-15 21:34:13 +01:00
1eb5c20f57
ansible: add dummy init.pys so setup.py find_packages() DTRT.
2018-05-15 21:29:00 +01:00
2c58591129
Bump Ansible version in dev_requirements.txt
2018-05-15 21:17:48 +01:00
7d0209d8de
issue #249 : have upgrade_router() upgrade the poller too.
...
Now when a child becomes a parent, it gets a new poller suitable for
many more children than was possible using select().
2018-05-15 09:28:03 +01:00
55fff54774
core: make try/catch logic a little clearer in Latch.get()
2018-05-15 09:22:35 +01:00
05a5f2b6e5
core: if Poller.poll() fails, TimeoutError would be raised.
...
We must check whether poller threw an exception both in the case that we
weren't woken and the case that we were.
2018-05-15 04:18:55 +01:00
5bdc1719c5
issue #249 : epoll() raises IOError for EINTR, not select.error.
2018-05-15 04:15:32 +01:00
70376d861a
issue #217 : docs: remove limitation
2018-05-15 01:57:09 +01:00
6d18ce81d8
issue #249 : restore duplex behaviour for epoll
...
With epoll() there is only one kernel-side object per file descriptor,
which is why _control() is such a pain. Since we merge receive/transmit
watching into that single object, we must always test the mask for both
conditions when reading results.
Kqueue isn't/doesn't appear to be like this. The identity of a Kqueue
event is keyed on (fd, filter), and we register a separate event for
both transmit and receive, so the 'elif' in KqueuePoller.poll() does not
appear to need to change.
Previously, a FD marked for read+write would not indicate writeability
until it was no longer readable.
2018-05-15 01:49:03 +01:00
6b98818046
issue #249 : epoll distinguishes between hangup and disconnect
...
..typical Linux, for certain descriptor types only. So our receive mask
must match both, and normalize it into a read event like every other
poller.
2018-05-15 01:32:46 +01:00
07056b0dd1
issue #249 : fix ordering bug masked by previous implementation
2018-05-15 00:27:41 +01:00
4cd9e09130
issue #249 : docs: remove limitation
2018-05-15 00:22:02 +01:00
36a1024861
issue #249 : port Latch to poller too.
...
This is probably going to suck for perf :/
2018-05-15 00:21:02 +01:00
1070dfae72
issue #249 : fix iter_read() regression.
2018-05-15 00:17:38 +01:00
dcf0aa351e
issue #249 : whoops, fix new poller timeouts.
2018-05-15 00:15:32 +01:00
aa8f786413
issue #249 : fix Poller.from_existing() for v2 API
2018-05-15 00:11:08 +01:00
5645629e5d
issue #249 : the new pollers must handle syscall restarts too.
2018-05-15 00:06:31 +01:00
4df020827d
issue #249 : explicitly close pollers when done.
2018-05-15 00:01:01 +01:00
9905f6d8b4
issue #249 : make write_all() and iter_read() use PREFERRED_POLLER.
2018-05-14 23:56:11 +01:00
b6124f8396
issue #249 : EpollPoller v2.
2018-05-14 22:50:35 +00:00
9abcf63155
issue #249 : Poller API v2 (BSD only).
...
Now it's BasicStream/Side-agnostic, so it can be reused for Latch and
iter_read().
2018-05-14 23:32:26 +01:00
11c2e4ab3e
core: set _v and _vv to True in enable_debug_logging().
...
router.enable_debug() has been broken for ages.
2018-05-14 22:58:56 +01:00
7320c542df
issue #249 : EpollPoller() for Linux.
2018-05-14 17:52:08 +00:00
bc7be1879d
issue #249 : initial poller implementation (BSD only)
2018-05-14 18:31:38 +01:00
b7ab473343
issue #186 : split handle list up so it makes sense
2018-05-13 17:05:09 +01:00
d1a22cb5d4
issue #186 : parent: implement FORWARD_MODULE.
...
To support detach, we must be able to preload the target with every
module it will need prior to detachment. This implements the
intermediary part of the process (i.e. the Ansible fork parent) --
receiving LOAD_MODULE/FORWARD_MODULE pairs and ensuring they reach the
child.
2018-05-13 16:58:47 +01:00
8a089e975d
docs: Document Router.unidirectional.
2018-05-13 16:13:50 +01:00
bb61745a1a
issue #217 : pass through non-custom module utils to regular importer.
...
This may come back to bite later, but in the meantime it avoids shipping
up to 12KiB of junk metadata for every single task invocation.
For detachment (aka. async), we must ensure the target has two types of
preloads completed (modules and module_utils files) before detaching.
2018-05-13 04:40:07 +01:00
30034877a5
issue #217 : ansible: working, if extremely inefficient implementation
2018-05-13 01:47:11 +01:00
81b62d9a1a
issue #217 : ansible: beginnings of ModuleDepService.
2018-05-12 15:12:39 +01:00
0f8190eff6
issue #217 : ansible: working module_finder.py
2018-05-12 15:12:35 +01:00
6a17042020
issue #217 : ansible: merge duplicate ReplacerPlanner.
2018-05-12 15:12:29 +01:00
78d375e0c5
core: CallError should handle any exception type.
...
Previously SystemExit would be pickled.
2018-05-12 00:51:50 +01:00
faaac43a78
Disable SSH variables test on vanilla Ansible.
2018-05-12 00:51:50 +01:00
68d6e6bf34
ansible: tests for all SSH variables.
...
Closes #184 .
2018-05-12 00:51:50 +01:00
f01f237e6d
Merge pull request #247 from dw/issue179
...
issue #179 : don't reap first stage until core_src_fd is drained.
2018-05-11 22:18:31 +01:00
def22c226f
issue #179 : don't reap first stage until core_src_fd is drained.
...
Bootstrap would hang if (as of writing) a pipe sufficient to hold 42,006
bytes was not handed out by the kernel to the first stage. It was luck
that this didn't manifest before, as first stage could write the full
source and exit completely before reading begun.
It is not clear under which circumstances this could previously occur,
but at least since Linux 4.5, it can be triggered if
/proc/sys/fs/pipe-max-size is reduced from the default of 1MiB, which
can have the effect of capping the default pipe buffer size of 64KiB to
something lower.
Suspicion is that buffer pipe size could also be reduced under memory
pressure, as reference to busy machines appeared a few times in the bug
report.
2018-05-11 21:58:38 +01:00
3ff37113e2
Merge pull request #246 from dw/issue179
...
issue #179 : ec0_receive() and connect_bootstrap() should use deadline.
2018-05-11 18:33:13 +01:00
d65e047b12
issue #179 : ec0_receive() and connect_bootstrap() should use deadline.
...
Now there is a single global deadline derived from ansible.cfg timeout=
value.
2018-05-11 18:19:27 +01:00
2c309a378a
Merge pull request #245 from dw/dmw
...
Log masking, unlinked contributors page, resolv.conf fix, unidirectional routing
2018-05-08 16:24:06 +01:00
bd2cc0830c
Enable unidirectional routing in Ansible; closes #132 .
2018-05-08 16:10:44 +01:00
David Wilson