Commit graph

23 commits

Author SHA1 Message Date
Matthew Chen
4837a9d371 Pin YapDatabase to v2.9.3 to avoid v.3.x.
// FREEBIE
2017-05-19 17:56:14 -04:00
Michael Kirk
78515377b1 Censorship circumvention in Egypt and UAE
* domain fronting
* non-websocket message fetching
* alternate pinning policy for google hosted reflector server

// FREEBIE
2016-12-21 09:56:48 -06:00
Michael Kirk
5ccbd4ca6d Bail if we can't build a database.
* Something is irrecoverably wrong at this point.
* Constantize error names

// FREEBIE
2016-12-16 14:49:39 -06:00
Michael Kirk
7036c63392 Compatible with libphonenumber 0.9.1
// FREEBIE
2016-12-02 19:13:34 -05:00
Michael Kirk
df756423f2 Ignore unknown group messages
otherwise when (e.g.) android adds a new group type they'll be shown as
empty bubbles.

// FREEBIE
2016-11-22 12:14:39 -05:00
Michael Kirk
1ba082356a Explicitly include newlines in numeric fingerprint
More straight forward than fudging / verifying different layout widths.

// FREEBIE
2016-11-14 17:38:26 -05:00
Michael Kirk
60a39f93c6 Remove phone numbers from scannable QR Code
Initially they were there to offer richer error messages when the wrong
code was scanned.

However, we found that people were posting them publicly,
misunderstanding the QR codes to be a kind of personal identifier.

For one, this isn't useful because the QR codes, like safety numbers,
are unique *per* conversation. So they aren't useful in a generic sense.
And secondly this is bad because the QR code contains both parties phone
numbers.

// FREEBIE
2016-11-14 08:43:24 -05:00
Michael Kirk
70e536ca8f Privacy preferences for blocking identity change
// FREEBIE
2016-11-10 16:28:36 -05:00
Michael Kirk
725153307e Add some nullability annotations (#62)
* fix compiler warnings about info message initializers

// FREEBIE
2016-11-10 15:21:56 -05:00
Michael Kirk
b0343ee1d1 Only fetch PreKey once.
Previously we were retrying as if it might succeed, and running into
rate-limit errors.

Also, added a specific rate limit error message.

// FREEBIE
2016-11-10 09:59:07 -05:00
Michael Kirk
027fa10730 Only dispatch async at top level.
Previously each method was dispatching it's callbacks async. This is a
waste of code, processing, and makes crash logs impenetrable.

// FREEBIE
2016-11-07 09:48:05 -05:00
Michael Kirk
b4c504f61d EndSessionMessage notifies remote side of reset session
Otherwise they'll send us messages we can't decrypt.

Also: Log and fail when message doesn't send due to no thread being specified.

// FREEBIE
2016-11-04 12:24:39 -04:00
Michael Kirk
47cad611e5 Fix register w/o background fetch & stale push tokens
* Separate account registration from push token registration
* Provide better errors when validation fails (e.g. numbers don't match, numbers blank)
* More logging during registration
* Call success after setting phone number to avoid any future race condition

  This isn't currently causing problems, but it's unexpected that we'd
  mutate the state *after* calling a callback which might inuitively rely
  on that state.

* Don't throw exception off thread when device keys 404's
* Better async startup handling
  - move processing off main thread
  - reduce code duplication
  - don't wrap it in a transaction in the future case where we want to
    further access the DB

// FREEBIE
2016-11-03 16:10:06 -04:00
Michael Kirk
4ba1e86ec1 Explain send failures for text and media messages
Motivation
----------
We were often swallowing errors or yielding generic errors when it would
be better to provide specific errors.

We also didn't create an attachment when attachments failed to send,
making it impossible to show the user what was happening with an
in-progress or failed attachment.

Primary Changes
---------------
- Funnel all message sending through MessageSender, and remove message sending
  from MessagesManager.
  - Record most recent sending error so we can expose it in the UI
  - Can resend attachments.
  - Update message status for attachments, just like text messages
- Extracted UploadingService from MessagesManager
  - Saving attachment stream before uploading gives uniform API for send vs.
    resend
  - update status for downloading transcript attachments
- TSAttachments have a local id, separate from the server allocated id
  This allows us to save the attachment before the allocation request. Which is
  is good because:
  1. can show feedback to user faster.
  2. allows us to show an error when allocation fails.

Code Cleanup
------------
- Replaced a lot of global singleton access with injected dependencies to make
  for easier testing.
- Never save group meta messages. Rather than checking before (hopefully) every
  save, do it in the save method.
- Don't use callbacks for sync code.
- Handle errors on writing attachment data
- Fix old long broken tests that weren't even running. =(
- Removed dead code
- Use constants vs define
- Port flaky travis fixes from Signal-iOS

// FREEBIE
2016-10-24 09:36:51 -04:00
Michael Kirk
9fe0ca000a bump version
// FREEBIE
2016-09-02 12:28:57 -04:00
Michael Kirk
1d0b645fc9 Update to new protocol (#23)
* new protobufs from libsignal-service-java@a1c93edc40e3ce201fa4e70339a3e6f4e197e319
* Makefile to build ObjC classes from *.proto
* refactored some touched code to minimize duplication
* removed unused protos
* deleted some dead code
* renamed method for clarity

// FREEBIE
2016-08-22 16:09:58 -04:00
Michael Kirk
f3a91c2629 Avoid collision with iOS10 SSKeychain framework (#24)
Our pod SSKeychain was renamed to -> SAMKeychain to avoid collision with
the iOS10 library SSKeychain.

* log failure to write keychain (this seems to only happen on simulator)
* ensure we exit if we fail to set DB cipher key

// FREEBIE
2016-08-18 13:40:48 -04:00
Michael Kirk
0f9a3334c1 Ensure interactions removed when thread is deleted
In theory, this should have already been handled by the
YapDatabaseRelationship extension via edges.

However, in practice, there were situations (cause unknown) where
interactions would exist without an edge to their corresponding thread.

Rather than being clever with the edge/callback machinery, now threads
explicitly delete all their interactions, and interactions delete all
their attachments (when applicable).

Also, a class to clean up spurious interactions / attachments

In the process:
- refactored TSYapDatabaseObject init to specify designated initializer
- added some testing niceties to TSYapDatabaseObject

// FREEBIE
2016-07-31 08:49:01 -07:00
Michael Kirk
664162fe24 Use SocketRocket pluggable policies (#15)
* Use SocketRocket pluggable policies

Pluggable policies are currently only in our fork, but pending upstream
PR: https://github.com/facebook/SocketRocket/pull/429

Also:
* rebased SocketRocket against latest upstream to incorporate bug fixes.
* adapt to new upstream API for error handling (vs previous exception throwing)
* renamed AFSecurityOWSPolicy -> OWSHTTPSecurityPolicy to differentiate
  it from OSWWebSocketSecurityPolicy
* Follow conventional singleton pattern with onceToken
* bump xcode version to play nice with SWIFT_NAME in SocketRocket

// FREEBIE
2016-06-30 20:39:01 -07:00
Michael Kirk
80671b247f Extract phone formatting method (#14)
* refactor and backfill tests for phone number parsing
* bump version to 0.0.5
2016-06-28 10:22:36 -07:00
Michael Kirk
8058951b08 Adapt to updated SocketRocket pod refactorings. (#11)
While trying to pull in upstream changes to our fork of SocketRocket I
noticed we're publishing it/installing it from CocoaPods trunk, but
there's no need to do that.

Also, due to upstream refactoring, the actual included file has
changed.

// FREEBIE
2016-06-14 12:19:19 -07:00
Michael Kirk
3024225655 Get tests + ci running 2016-04-23 20:08:51 -07:00
Frederic Jacobs
a49d36d66a Renaming to SignalServiceKit.
TODOS: Rename Symbols and Xcode files.
2016-04-08 18:57:31 +02:00
Renamed from TextSecureKit.podspec (Browse further)