| Age | Commit message (Collapse) | Author |
|---|
|
the token is set when we get the srp token in the keymanager service,
but it wasn't been propagated properly to the internal nicknym object
that was refactored recently.
here I make the setter write the same token to both keymanager itself
and the nicknym.
some little additional refactors:
- move gpg initialization to a helper method
in this way it becomes easier to stub the soledad and keymanager
components, for testing.
- name nicknym token as publicly visible, since i'm accessing
it from the keymanager object.
- Resolves: #8651
|
|
the bundle was trying to get the gpg binary from the config folder
instead of the bundled binary.
|
|
- some pep8 problems
- some confusion with old and new code after merging
from old keymanager
|
|
disabled some tests that were failing, for some reason the twisted
logger doesn't want to be patched?
besides, I don't think it's a good idea to rely on log information for
tests. what is it really that we were trying to test there?
|
|
|
|
Port of the original commit:
8f1fe8dd4a54fd2bdda2fc78c339ce9b3d0fc331
by Zara Gebru that introduced updating keys in the background.
This was made in the legacy leapcode/keymanager repo, but was lost in
the merge to the unified bitmask-dev.
Original commit message follows:
--------------------------------
- refresh random key in random time
- add get key by fingerprint
- refactor nicknym methods to own file
- tests
- note this do not include a check for
revoked key, since that need some changes
in gnupg
- Related: #6089
|
|
|
|
this also means that these bundles can only be used against cdev for
now, since we need the latest soledad server.
i'm using my personal devpi repo because the development mode install
breaks the bundles for some reason (namespace probably).
latest version uploaded is: leap-soledad-client-0.9.0+58.g0d9e74e
which is today's shyba state but rebased on top of the 0.9.0 tag.
|
|
this should affect the name in the download dropdown in Gitlab
|
|
I think 1 week should be enough so that we don't clutter the machine too
much. It probably will give trouble when someone goes on vacation, or if
we stall the development of bitmask because we're moving development
towards other fronts, but...
|
|
just the one shipped with the ubuntu image used by the docker container
for now. I have to investigate how to properly use a docker container
only for building the gpg binary.
|
|
|
|
|
|
|
|
otherwise, the call from the rest API hangs forever because the
authentication has not happened and not all the configs have been downloaded.
- Resolves: #8576
|
|
This has been previously encapsulated in a dict, and the commands were
not modified accordingly.
I'm adding some very basic test for the KeymanagerService public api
contract.
- Resolves: #8577
|
|
this reverts the refactoring made at 596f1f8
it seems it is now allowed to call a function that yields from inside as
part of an inline deferred.
|
|
you need to configure onion = True in the services config.
for sure, this is gonna be more interesting when we ship pixelated.
but for now I thought it can be handy for testing ui changes: no need to
bundle, just use your tor browser :)
still need to try to serve imap/smtp over the onion service.
kudos to meejah for this super-usable library!
|
|
some hacks needed, it seems pyinstaller is choking again on namespace
packages for some reason that is beyond my current comprenhension.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
in the variant that was merged, we call a failure method to get the
exception, so it's better to test a real failure initialized with an
exception.
besides, I'm starting to hate mocks.
|
|
|
|
|
|
This commit originall contained a fix for a wrong variable name that
avoided the outgoing service to be launched. This had been fixed in a
concurrent mr that got merged before this one. However, this MR
contained a couple of tests for the OutgoingService, and I'm merging
those tests now.
|
|
this takes too long.
if we want to test the ui npm build, we probably should do it only when
we touch anything under ui/*
|
|
otherwise, we were appending a list as the last element of the list.
|
|
Missing yield from _insert_key_docs, which returns a Deferred, was
causing tests to fail.
|
|
Twisted 16.5 sends a BytesIO as message, which was unexpected in this
types list.
|
|
|
|
|
|
basically, I give up trying to compile pyqt5 for python2. not worthy.
fortunately, the api changes are minimal so we can support two.
the way forward should be Py3.
still some bugs with rendering the viewport.
trying to help that, I introduced the web inspector, use DEBUG=1 flag to
have it visible.
|
|
this looks like a careless migration to twisted failures.
who knows how long this was like this, this should be covered by tests.
|
|
|
|
scrypt raises a "unable to get salt from /dev/urandom" error when using
encrypt method in windows.
this is a quick workaround for obfuscating the uuid map, but it needs to
be carefully reviewed as per the security implications (i'm using the
hostname to salt the key hashing).
doing this, I contemplated the possibility that the scrypt encryption we
were using could be flawed, since the payloads can be easily guessed.
so, take this with a pinch of salt (haha), and let's review it for the
future. as always, this was in a rush trying to bring windows support.
|
|
currently windows 7 support is broken with gnupg.
|
|
|
|
|
|
|
|
|
|
|
|
|
