| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| ... | |
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
We already successfully handled links of the forms:
* `[RustId]`
* `[text](https://...)`
* `[text][label]`
But we didn't handle links of the form:
* `[text](RustId)`
These latter links aren't correct markdown on their own; but they're
simple enough to fix up after a little hacking on the marko backend
representation.
|
| | | |
|
| | |
| |
| |
| | |
(Current nightly works fine, so we can stop recommending a downgrade.)
|
| | |
| |
| |
| | |
Fortunately, the new API has simplified matters greatly!
|
| |\ \
| | |
| | |
| | |
| | | |
rpc: Tolerate NotFound on configured connect point directory.
See merge request tpo/core/arti!2735
|
| | | | |
|
| | |/
| |
| |
| |
| |
| |
| |
| | |
If Arti tries to bind to a directory of connect points, and that
directory isn't there, the right behavior is to treat the directory
as if it were empty. Otherwise, the user would have to create
the default connect point directory (as empty) before Arti would
start.
|
| |\ \
| | |
| | |
| | |
| | | |
rpclib: Avoid overuse of execute_internal_ok.
See merge request tpo/core/arti!2730
|
| | | | |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Also change corresponding message.
This way, it will be clear that we're just reporting that Arti
gave us an error—not that the error necessarily means that the
authentication itself was "wrong".
|
| | |/
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The execute_internal_ok method converts every error response into an
internal error; as such, it's only appropriate when there is no way
for a well-behaved Arti instance to give an error response.
But we had been using it in a few places where errors were possible
under other circumstances.
This commit fixes that behavior, and adds documentation to help
avoid it.
|
| |\ \
| | |
| | |
| | |
| | | |
rpc: Resolve a couple of dead code TODOs
See merge request tpo/core/arti!2731
|
| | | | |
|
| | | | |
|
| |\ \ \
| |_|/
|/| |
| | |
| | | |
Fix several 'TODO RPC' notes in the arti crate.
See merge request tpo/core/arti!2737
|
| | | |
| | |
| | |
| | |
| | |
| | | |
Formerly this was a conditional method argument, which is a huge
antipattern. Now it is unconditionally present, as `Option<T>` for
a type that is uninhabited when RPC isn't supported.
|
| | | |
| | |
| | |
| | |
| | | |
rust-analyzer keeps re-wrapping this piece for me, even though
rustfmt doesn't complain.
|
| | | |
| | |
| | |
| | | |
Information _is_ passed to the RpcMgr, via the argument to new_connection.
|
| | | |
| | |
| | |
| | |
| | | |
The RpcMgr does indirectly hold a reference to the client,
via its make_session argument.
|
| | | |
| | |
| | |
| | | |
We _do_ have error detection from this function, and have for ages.
|
| |/ /
| |
| |
| |
| | |
This was necessary before we had support for implementing
RPC methods on generic types.
|
| |\ \
| | |
| | |
| | |
| | | |
tor-config: Improve mistrust documentation
See merge request tpo/core/arti!2727
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
This commit improves the documentation for
`ConfigurationSources::set_mistrust`, by explaining that this option is
unrelated to the paths defined within the configuration file itself,
referring to the `storage.permissions.dangerously_trust_everyone`
option.
|
| |\ \ \
| | | |
| | | |
| | | |
| | | | |
arti-rpc-client-core: Use EmptyReply instead of EmptyResponse.
See merge request tpo/core/arti!2732
|
| |/ / /
| | |
| | |
| | |
| | |
| | | |
It looks like !2729 and !2722 raced with each other, because we're still
using the old name for `EmptyReply` (and so `arti-rpc-client-core`
doesn't currently compile on `main`).
|
| |\ \ \
| |_|/
|/| |
| | |
| | |
| | |
| | | |
rpc: Implement request cancellation
Closes #818
See merge request tpo/core/arti!2722
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
The current cancel code is prone to deadlock, so the easiest way to
solve it appears to be making cancel requests themselves
uncancellable.
I've included a test to verify the behavior; previously, this test
caused a deadlock.
|
| | | | |
|
| | | | |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | | |
Nothing used it, and it has some semantic complexity.
(see
https://gitlab.torproject.org/tpo/core/arti/-/merge_requests/2722#note_3149591
)
|
| | | |
| | |
| | |
| | |
| | |
| | | |
The Waker::clone_from implementation uses Waker::will_wake
to avoid unnecessarily cloning a Waker that it already
has a copy of.
|
| | | | |
|
| | | | |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | | |
I hope that this limitation is acceptable;
the alternative involves some significant refactoring to give
Request a Weak reference to RpcConn -- but RpcConn isn't currently
kept in an Arc<> at all, and so we'd need some fairly heavy hacking.
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | | |
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Cancellation is now fallable, which allows us to detect attempts
to cancel which cannot work.
We now guarantee that when you try to cancel a `Cancel<F>` future,
either the cancel operation will succeed, or the future will return
(or will have already returned) Ok(), but not both, and not neither.
Closes #818.
|
| |\ \ \
| |_|/
|/| |
| | |
| | |
| | |
| | | |
rpclib: Rename params/reply structs for consistency.
Closes #1586
See merge request tpo/core/arti!2729
|
| |/ /
| |
| |
| |
| |
| |
| |
| | |
Our now convention here in rpclib is that a struct holding a
request's parameters is called `FooParams`, and a struct holding
that request's reply is called `FooReply`.
Closes #1586
|
| |\ \
| | |
| | |
| | |
| | | |
Better instructions for handling new MPL dependencies
See merge request tpo/core/arti!2726
|
| | | | |
|
