If no session management is performed, txi2p starts a process-wide session the
first time a connection (client or server) is opened; all subsequent connections
use that session and its configuration properties.
This commit ensures that the same properties are passed to both client and
server endpoints, so that the correct I2P Destination is started regardless of
whether the first connection made by Tahoe-LAFS is for a client or server.
Closes#2858.
Previously this looked at sys.platform to decide what dependencies to
include. The problem with that approach is that wheels built on a unix box
won't work on windows (and vice versa), when the (pure-python) Tahoe wheels
aren't supposed to be platform-dependent.
setup.py provides a syntax to express this properly, so wheels created on
either platform will include the pypiwin32 dependency in the metadata, but
marked as only being relevant when installing on a win32 platform.
closes ticket:2763
Also adds a --poll-interval option to both 'magic-folder join'
and 'magic-folder create' so that the integration tests can pass
something "very short".
* replace "last_details" with "non_connected_statuses" dict
* rename "last_connection_summary" to just "summary"
* for connected servers, show other hints in a tooltip
* for not-yet-connected servers, show all hints in a list
* build the list (in STAN) on the server side, not using IContainer
This shows current-connection info, and provides per-hint status details in a
tooltip.
The "Connection" section no longer shows seconds-since-loss when the server
was not connected (previously it showed seconds-since-connect when connected,
and flipped to seconds-since-loss when disconnected). We already have the
"Last RX" column, which is arguably more meaningful (and I can't think of a
good case when these would differ), so we don't really need
seconds-since-loss, and the new ConnectionStatus doesn't track it anyways.
So now the "Connection" timestamp for non-connected servers is just
"N/A" (both the main text and the tooltip). The "Introducers" section was
changed the same way.
This moves the per-server connection timestamp out of the nickname/serverid
box and over into the Connection box. It also right-floats all timestamps,
regardless of which box they're in, which makes them share the box with
connection_status more politely.
Internally, this adds code to create ConnectionStatus objects when necessary.
Previously, "tahoe create-node" without an --introducer= argument would
result in the literal string "None" being written into tahoe.cfg:
[client]
introducer.furl = None
We were using config.get("introducer",""), but that didn't suffice because
the key was actually present: it just had a value of None, which then got
stringified into "None" when writing out tahoe.cfg.
This briefly caused test/cli/test_create to fail, as the startup code tried
to parse "None" as a FURL. This only happened against a development version
of Foolscap which accidentally became sensitive to unparseable FURLs in
started Reconnectors. I fixed that in the final foolscap-0.12.5 release, so
we shouldn't hit this bug, but I wanted to fix it properly in the tahoe-side
source.
As discussed in this week's meeting, since we don't yet know why some
flavors of linux have slightly different inotify behavior than others,
and since we believe the actual functionality is not significantly
impacted, and since the red buildbot is reducing our confidence that the
other tests are passing, and since we have a release coming up: we're
marking the one troublesome test as ".todo". We expect that the test
will be fixed soon (perhaps to accept either 3 or 4 events), but not
necessarily before the 1.12 release.
refs ticket:2834
This forces the Uploader and Downloader to implement a _scan_delay
method and makes the naming more consistent with what's actually
happening. Also, fix a few "bugs" in the names of args in the
mocks for some tests.
We use it for two things: to create the foolscap connection handler, and
to possibly start an .onion listener at startup.
This also updates node._common_config_sections to accept the new tor
settings written by create-node/create-introducer.
This adds tor-related CLI arguments to "create-node" and
"create-introducer", to control exactly how we should be using Tor.
* --tor-launch
* --tor-executable=
* --tor-control-port=
I went with "--tor-launch" instead of "--launch-tor" for consistency. I
don't particularly like the grammatical flow of it, and it doesn't
actually put all the tor-related arguments next to each other in the
--help output (the flags are put in one block, then the parameters in
the next). But it seems slightly more consistent to start all the
tor-related argument names with a "--tor*" prefix.
This uses a unix-domain control port, and includes test coverage.
create_onion() displays pacifier messages, since the allocate-onion step
takes around 35 seconds
All server-like nodes (storage servers and introducers both) will need
this for the tor state directory and .onion private key file, and it
needs to exist before the config is written, so tor onion-service
private keys can be placed there.
Also remove a redundant import.
This puts the right inlineCallbacks in place to allow
write_node_config() to return a Deferred. The upcoming Tor support will
need this (since it must wait for an .onion address to be allocated
before it can write tahoe.cfg's tub.port and tub.location lines).
which uses SHA1 to combine the file's storage index (known as "peer
selection index" in this context) and each server's "server permutation
seed". This is the only thing in tahoe that uses SHA1.
With this change, we stop importing sha1 from anywhere else.
I think the preferred way to listen on both IPv4 and IPv6 will be to use
"--port=tcp:PORT,tcp6:PORT". This is now reflected in the docs.
refs ticket:867
This enables an I2P-only node, which disables TCP entirely (instead of
mapping TCP to Tor, which was the only other option that
reveal-IP-address=False would allow).
closes ticket:2824
parse_cli() got added during the async-CLI-dispatch work
assertRaises/assertFailure have been in Twisted for a while, but I only
learned about them recently. Over time I'm looking forward to changing
all tahoe tests to use them (and getting rid of ShouldFailMixin/etc).
In addition, CLI functions are allowed to use sys.exit() instead of
always needing to return the exit code as an integer.
runner.py now knows about the blocking httplib calls in scripts/cli and
scripts/magic_folder, and uses deferToThread() to invoke them. Those
functions cannot return a Deferred: when rewrite them to use twisted.web
or treq, we'll remove this deferToThread call.
Option parsing was split out to a separate function for testing. We now
use twisted.internet.task.react() to start the reactor, which required
changing the way runner.py is tested.
closes ticket:2826
The main part of CLITestMixin.do_cli() was split into a standalone
function named run_cli(), leaving do_cli() as a method which includes a
nodedir in the arguments (for use by GridTestMixin tests which do a lot
of CLI operations against one of their client nodes, for which adding
the extra --nodedir argument would be ugly).
