- The configuration and crypto model is now entirely independent
of Android classes other than Nullable and TextUtils.
- Model classes are immutable and use builders that enforce the
appropriate optional/required attributes.
- The Android config proxies (for Parcelable and databinding) are
moved to the Android side of the codebase, and are designed to be
safe for two-way databinding. This allows proper observability in
TunnelDetailFragment.
- Various robustness fixes and documentation updates to helper classes.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
When a tunnel is running, saving the tunnel's config with an IPv6
address endpoint like [::1]:42 would result in the wrong format ::1:42.
This patch fixes it.
For endpoints with an IPv6 address(e.g. [::1]:42). Since the default
endpoint InetSocketAddress is created unresolved, getEndpointString()
returns "[::1]:42" (InetSocketAddress.getHostString() returns the
literal hostname). After the endpoint is resolved, getEndpointString()
returns "::1:42" (InetSocketAddress.getHostString() returns the IPv6
address without the square brackets). This inconsistent return values
caused the above mentioned bug.
With this patch, function getEndpointString would return the right
format string whether the endpoint is resolved or not.
Signed-off-by: Zhao Gang <gang.zhao.42@gmail.com>
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
The configurations are supposed to be in a very specific
format which is not user-facing and hence doesn't have to
be adjusted for locale avoiding both the redundancy as well
as potential breakages in the configuration file format from
different locales.
Fixes: 71c67aa24ae2 ("config: Minor cleanup")
Reported-by: Samuel Holland <samuel@sholland.org>
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
Use a canonically-named utility class to tack on methods to the existing
InetAddress class. Rename IPCidr to InetNetwork so it better matches
InetAddress and is more pronouceable :) While here, simplify the
constructor and toString() functions, and properly implement hashCode().
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
I know we decided to ditch the idea of shutting up "Exception
thrown with empty param" warnings but this pesters me too
much and we can instead just treat this as a weird future proofing
thing if and when we end up needing the exception messages.
Signed-off-by: Harsh Shandilya <me@msfjarvis.dev>
This is insane, but it appears to be working. We essentially store
things in a separate class for editing, and then commit it back at a
given time.
This business with onViewStateRestored in both TunnelEditorFragment and
in TunnelDetailFragment is buggy and likely wrong.
In general TunnelEditorFragment should probably be rewritten. The
relationship with the changed name is not clear.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
This will make the two way data binding crash more, but it will improve
the robustness of the config file parser, which deals with exceptions
gracefully, and when we move to one way data binding, it will help with
that too.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
Since the amount of mind numbing boiler plate has been increased, this
must be the proper way to do things.
Signed-off-by: Jason A. Donenfeld <Jason@zx2c4.com>
