Push notifications #582

ricardopereira · 2017-02-09T00:08:31Z

Installation

You can install Ably for iOS through CocoaPods, Carthage or manually.

Installing through CocoaPods (recommended)

Add this line to your application's Podfile:

# For Xcode 7.3 and newer
pod 'Ably', :git => 'https://github.com/ably/ably-ios.git', :branch => 'push-beta-legacy'

And then install the dependency:

$ pod install

Then you can:

 // On Objective-C
#import <Ably/Ably.h>

// On Swift
import Ably

Installing through Carthage

Add this line to your application's Cartfile:

# For Xcode 7.3 and newer
github "ably/ably-ios" "push-beta-legacy"

And then run carthage update to build the framework and drag the built Ably.framework into your Xcode project.

Manual installation

Clone the push branch: git clone -b push [email protected]:ably/ably-ios.git
Drag the directory ably-ios/ably-ios into your project as a group.
Ably depends on SocketRocket 0.5.1; get it from the releases page and follow its manual installation instructions.
Ably also depends on msgpack 0.1.8; get it from the releases page and link it into your project.

ricardopereira · 2017-02-09T00:14:45Z

Source/ARTPush.m

+    }
+
+    if (registerCallback) {
+        self.device.updateToken = registerCallback(deviceDetails, nil);


@tcard maybe this should be async and not waiting for a return.

In which situation you want to pass an error to the registerCallback? I don't see any case.

In which situation you want to pass an error to the registerCallback? I don't see any case.

In none that I can think of, really. Maybe we'll end up removing it.

maybe this should be async and not waiting for a return.

Yes, it should. It returns io T. In iOS, ) -> io T translates roughly into , callback: (T, Error)). Just like authCallback.

Fixed 08f9d01.

tcard

Looking pretty good! Some things need changing though.

tcard · 2017-02-13T16:54:06Z

Source/ARTDeviceDetails.h

+
+extern NSString *const ARTDevicePlatform;
+
+typedef NS_ENUM(NSUInteger, ARTDeviceFormFactor) {


I believe we talked about those enums being plain strings in the client libraries. You did it for platform so probably I wasn't clear that it affects formFactor and transportType too.

The point is that, if we add new form factors in the future, the client libraries don't need to be updated.

Fixed 64e6dab.

tcard · 2017-02-13T16:57:39Z

Source/ARTPush.h

+@interface ARTDeviceToken : NSData
+@end
+
+@interface ARTUpdateToken : NSString


I don't think this is type-safe? You can't ask for type Dog and get Animals. Don't you get a compile error or at least warning when passing a NSString as ARTUpdateToken?

(Same of all other classes above.)

Yes, that was an initial solution but I changed it to a typedef (581e76f#diff-e040b62ac54206ef54da9e6409623314R18). Now it's working fine. 👍

tcard · 2017-02-13T17:01:32Z

Source/ARTDeviceDetails.m

+}
+
+- (instancetype)init {
+    return [self initWithId:[[NSUUID new] UUIDString]];


OK for now, but I did mention ULID, right? It's not crucial but it's also straightforward to use instead. It's what I use in Java so it'd be nice for IDs to be uniform.

@tcard Yes, you mention it but there is no standard ULID generator. I will look into the paper of ULID and try to implement it.

BTW, does the Java has it already implemented in the language itself?

I will look into the paper of ULID and try to implement it.

I don't think we should be writing our own implementation if we can help it. Is that really necessary if we can't find a lib that does it for us?

No, I used a library. Sorry, I assumed there was a library for Obj-C! It doesn't seem to be the case so you can be the first to make a CocoaPod :) Or I will if you don't want to!

There is an C++ library, so I created an Objective-C wrapper of it: https://github.com/whitesmith/ulid. It still need some work but it's working. I'll improve it later to continue this work.

tcard · 2017-02-13T17:01:59Z

Source/ARTDeviceDetails.m

+@implementation ARTDeviceDetails
+
+ (instancetype)fromLocalDevice {
+    return [[ARTDeviceDetails alloc] init];


This is missing all device info, so this is TODO, right?

@tcard I forgot about it, 👍 . So I should initialise it with the deviceId that is stored and the current updateToken, if it's in memory?

tcard · 2017-02-13T17:15:17Z

Source/ARTJsonLikeEncoder.m

+    dictionary[@"transportType"] = devicePushDetails.transportType;
+
+    if (devicePushDetails.deviceToken) {
+        // Normalizing token by removing symbols and spaces


Why? In tests I've used the device token as-is just fine.

@tcard You tested it converting it into a string type? The token is a Hex string, something like <12ce7dda 8032c423 8f8bd40f 3484e5bb f4698da5 8b7fdf8d 5c55e0a2 XXXXXXXX>. So I should let it be as it is.

Fixed d67ba78.

Ah, sorry, I was wrong, this is what I did: [deviceToken hexString] using some random hexString method I found googling. So yes, removing symbols and spaces should do the same. Sorry!

Done 4ab1d74.

tcard · 2017-02-13T17:20:42Z

Source/ARTPush.h

+/// Register a device, including the information necessary to deliver push notifications to it.
+- (void)activate;
+- (void)activate:(ARTDeviceDetails *)deviceDetails;
+- (void)activate:(ARTDeviceDetails *)deviceDetails registerCallback:(nullable ARTUpdateToken* (^)(ARTDeviceDetails * _Nullable, ARTErrorInfo * _Nullable))registerCallback;


So a problem I've encountered in Java is that this callback doesn't survive closing the app. So if you call activate, then the app closes, then the device token arrives (starting the app, I guess?) and the activation completes, the callback won't exist at that point, will it?

In Java I'm solving this with an Android's BroadcastReceiver. I don't know about iOS.

I need to check that on a real device but from what I know is that the app remains in the background when the token is getting retrieved from APNS.

tcard · 2017-02-13T17:24:40Z

Source/ARTPush.m

+        _device = [ARTDeviceDetails fromLocalDevice];
+        _state = ARTPushStateDeactivated;
+        _deviceTokenEmitter = [[ARTEventEmitter alloc] init];
+        _jsonEncoder = [[ARTJsonLikeEncoder alloc] init];


This should respect rest.options.useBinaryProtocol and use MsgPack if true.

Fixed bd73596.

tcard · 2017-02-13T17:25:44Z

Source/ARTPush.m

+    NSData *deviceToken = [[NSUserDefaults standardUserDefaults] dataForKey:ARTDeviceTokenKey];
+    if (!deviceToken) {
+        // Waiting for device token
+        [_deviceTokenEmitter once:^(ARTDeviceToken *deviceToken) {


As I said about registerCallback, this won't survive closing the app.

tcard · 2017-02-13T17:29:06Z

Source/ARTPush.m

+    }
+
+    if (registerCallback) {
+        self.device.updateToken = registerCallback(deviceDetails, nil);


In which situation you want to pass an error to the registerCallback? I don't see any case.

In none that I can think of, really. Maybe we'll end up removing it.

maybe this should be async and not waiting for a return.

Yes, it should. It returns io T. In iOS, ) -> io T translates roughly into , callback: (T, Error)). Just like authCallback.

tcard · 2017-02-13T17:31:30Z

Source/ARTPushRecipient.h

@@ -0,0 +1,47 @@
+//
+//  ARTPushRecipient.h


Sorry you took all the effort to do this, but we really want to keep this "stringly-typed" so that adding new recipient types on the server don't require changing anything in the client libraries.

So ARTPushRecipient should just be NSDictionary.

Done 58f7eab.

mattheworiordan

Nice progress!

mattheworiordan · 2017-02-14T11:40:58Z

Source/ARTDeviceDetails.m

+#import "ARTDevicePushDetails.h"
+
+NSString *const ARTDevicePlatform = @"ios";
+NSString *const ARTDeviceFormFactor = @"mobile";


Really hard-coded const? Given this could run on tablet or watch, I don't think this is right. Equally, we should really support MacOS at some point too.

Done 002c882.

mattheworiordan · 2017-02-14T11:41:56Z

Source/ARTDeviceDetails.m

+}
+
+- (instancetype)init {
+    return [self initWithId:[[NSUUID new] UUIDString]];


I will look into the paper of ULID and try to implement it.

I don't think we should be writing our own implementation if we can help it. Is that really necessary if we can't find a lib that does it for us?

mattheworiordan · 2017-02-14T11:44:53Z

Source/ARTPushChannel.m

+    }];
+}
+
+- (void)subscribeForClientId:(NSString *)clientId {


Would clientId not be implicit @tcard? Or do you think it's better to leave it as explicit as it's more flexible in future?

In the IDL it is implicit. But I don't know, I guess it can work like Presence#enterClient?

In the IDL it is implicit. But I don't know, I guess it can work like Presence#enterClient?

Not sure I see a use case for a client device to specify clientId other than the one implicit in the token. Servers on the other hand would do this frequently, is this API intended for server-side access as well?

Servers on the other hand would do this frequently, is this API intended for server-side access as well?

No, they would use the admin API most likely.

Well I would say that we shouldn't encourage people to try and register their device for a clientId that doesn't match their token, so if we want to offer this we should perhaps have an overloaded method where by default clientId is not required.

we should perhaps have an overloaded method where by default clientId is not required

Done 2c68fa9.

ricardopereira · 2017-02-20T20:40:15Z

@tcard PushAdmin is done. So, the final step is activation state machine? /cc @mattheworiordan

ricardopereira · 2017-02-22T21:04:20Z

Source/ARTPush.m

+    dispatch_once(&once, ^{
+        // Error: ARTPush.m:62:81: Instance variable '_httpExecutor' accessed in class method
+        //activationMachineInstance = [[ARTPushActivationStateMachine alloc] init:_httpExecutor];
+    });


@tcard ARTPushActivationStateMachine should be a singleton right? I need to have in mind which data the state machine needs to perform is work. Does it need more than the local device?

Yes, it should be a singleton. The persistent data it needs is the current state, the pending events, the device ID (which, once generated, doesn't change), the device's update token and the APNs device token. Also, whether activate/deactivate were called with custom registration or not.

@tcard I can do requests without the auth token or key, right? I don't really need the Rest instance. So, I can create the singleton with no dependencies.

@ricardopereira No, you can't, you need to authenticate in order to register, update the registration, and unregister. Unless the user provides the custom callback, of course.

ricardopereira · 2017-03-06T14:01:22Z

@tcard I think this is done.

tcard

Sorry I couldn't take a look at this sooner!

tcard · 2017-03-20T17:13:15Z

Source/ARTDeviceDetails.h

+@property (nullable, nonatomic) NSString *clientId;
+@property (nonatomic, readonly) NSString *platform;
+@property (nonatomic, readonly) NSString *formFactor;
+@property (nullable, nonatomic) NSDictionary<NSString *, NSString *> *metadata;


Non-nullable. It should be an empty map if not present.

tcard · 2017-03-20T17:15:59Z

Source/ARTDevicePushDetails.h

+extern NSString *const ARTDevicePushTransportType;
+
+typedef NS_ENUM(NSUInteger, ARTDevicePushState) {
+    ARTDevicePushStateInitialized,


There's no initialized state. It should default to active.

tcard · 2017-03-20T17:25:24Z

Source/ARTLocalDevice.h

+
+@interface ARTLocalDevice : ARTDeviceDetails
+
+@property (nonatomic, readonly) ARTDeviceToken *registrationToken;


This shouldn't be public.

tcard · 2017-03-20T17:26:31Z

Source/ARTLocalDevice.h

+
+@property (nonatomic, readonly) ARTDeviceToken *registrationToken;
+
+- (instancetype)init;


Not public.

tcard · 2017-03-20T17:26:48Z

Source/ARTLocalDevice.h

+@property (nonatomic, readonly) ARTDeviceToken *registrationToken;
+
+- (instancetype)init;
+ (ARTLocalDevice *)local;


Also not public. People should only get to the local device through rest.device().

tcard · 2017-03-20T19:07:27Z

Source/ARTPushChannel.h

+- (instancetype)init:(id<ARTHTTPAuthenticatedExecutor>)httpExecutor withChannel:(ARTChannel *)channel;
+
+- (void)subscribe;
+- (void)subscribeDevice:(ARTDeviceId *)deviceId;


This and unsubscribeDevice shouldn't take an argument but implicitly use the local device.

tcard · 2017-03-20T19:09:29Z

Source/ARTPushChannelSubscription.h

+@property (nonatomic, readonly) NSString *channelName;
+
+- (instancetype)init NS_UNAVAILABLE;
+- (instancetype)initWithDeviceId:(NSString *)deviceId andChannel:(NSString *)channelName;


In the spec, those are static methods named forDevice and forClientId, so we should stick to those.

tcard · 2017-03-20T19:09:36Z

Source/ARTPushChannelSubscription.h

+
+@property (nonatomic, readonly) NSString *deviceId;
+@property (nonatomic, readonly) NSString *clientId;
+@property (nonatomic, readonly) NSString *channelName;


Should be just channel.

tcard · 2017-03-20T19:09:55Z

Source/ARTPushChannelSubscription.h

+
+@interface ARTPushChannelSubscription : NSObject
+
+@property (nonatomic, readonly) NSString *deviceId;


Nullable, and clientId too.

tcard · 2017-03-20T19:13:02Z

Source/ARTPushDeviceRegistrations.m

+    NSString *tokenBase64 = [tokenData base64EncodedStringWithOptions:0];
+    [request setValue:[NSString stringWithFormat:@"Bearer %@", tokenBase64] forHTTPHeaderField:@"Authorization"];
+    request.HTTPMethod = @"PUT";
+    request.HTTPBody = [[_httpExecutor defaultEncoder] encode:@{


This will just save the device token. It should send the whole thing.

tcard · 2017-03-21T13:00:46Z

Source/ARTDeviceDetails.m

+#elif TARGET_OS_WATCH
+NSString *const ARTDeviceFormFactor = @"watch";
+#elif TARGET_OS_SIMULATOR
+NSString *const ARTDeviceFormFactor = @"simulator";


Doesn't the simulator simulate other form factors?

tcard · 2017-04-10T09:28:21Z

@ricardopereira This was getting urgent last week so I had to take over this, so don't worry about it. I'll push my changes (hopefully) shortly.

ricardopereira · 2017-04-10T09:32:39Z

@tcard What about tests? Should I add them?

tcard · 2017-04-10T09:43:12Z

@ricardopereira Not yet, I think. The code we want to test is still in flux.

mattheworiordan · 2017-04-11T02:42:18Z

The code we want to test is still in flux.

@tcard could we not add test coverage over the REST API at least? That's not changing much is it? See https://github.com/ably/ably-ruby/pull/115/files for a reasonable set of tests over this functionality or of course ably/ably-js@6294597

ricardopereira · 2017-04-25T18:02:17Z

@tcard I will change the way how the user adds his custom register/deregister callback because there is no way to test it nicely using the AppDelegate: https://github.com/ably/ably-ios/blob/push/Source/ARTPushActivationStateMachine.m#L207-L224. I didn't thought about it while I was coding 😒

I will change it and use the NSNotificationCenter. The user can implement it like:

NSNotificationCenter.default.addObserver(
    UIApplication.shared,
    selector: #selector(AppNotifications.refreshNotification(_:)), //<-- Custom callback
    name: ARTCustomRegisterCallbackNotificationName, //<-- identifier
    object: nil
)

It gives the user more flexibility and a clean AppDelegate.

tcard · 2017-04-25T18:21:51Z

@ricardopereira Does that allow you to check whether the user added an observer or?

In any case, I don't think the former way is so hard to test. Instead of hardcoding [UIApplication sharedApplication].delegate, you can define a method that returns it and you can replace for testing using the hooks thing.

ricardopereira · 2017-04-25T18:42:56Z

Nevermind, I can't detect if an ARTCustomRegisterCallbackNotificationName as registered observers.

ARTPushNotifications interface Realtime: integrate Push type Rest: integrate Push type Push.activate Push.publish - add PushRecipient protocol PushChannel subscribe and unsubscribe Fix registerCallback Remove warnings - unused methods Fix FormFactor Fix: use device token as Hex string Fix: should respect the default encoder - useBinaryProtocol Remove ARTPushRecipient protocol and types Add ULID pod Use ULID as DeviceId Push.deactivate Load local device Push: update device PushChannel.subscriptions Update ULID pod PushAdmin Update ULID pod Remove warnings JSON encoder: normalize deviceToken hex string Fix: Push.publish request body Update device forms Activation State Machine: init Activation State Machine: registration Removed ablyPushAuthKey, ablyPushAuthToken and ablyPushClientId from push delegate Add Rest argument on Push delegate Fix things for push, and adapt to latest API. Push: add registering delegate methods for Realtime too. Add new private headers to modulemap. Fix rebase merge conflicts. Recalculate Podfile.lock checksum.

With a static method, the constructed class is always the parent abstract class ARTPushActivationState, instead of the desired child class. Fixes #628.

* MockHTTPExecutor * PushAdmin * Move `publish` method to Admin * Add `get` method in PushDeviceRegistrations * Should use response encoder type * Fix save channel subscription * Fix save device registration * Better errors information * Rename publish notification argument to data to conform specs * Add equatable to DeviceDetails * Fix DeviceRegistrations.get * Add validations to PushAdmin.publish * Fix RHS1a * Fix RHS1b1 * Fix PushAdmin implementation * Fix ambiguous reference to member 'dataTask(with:completionHandler:)' * fixup! Fix PushAdmin implementation * CI test * fixup! Fix PushAdmin implementation

ricardopereira · 2018-01-17T22:54:33Z

Test suite (https://travis-ci.org/ably/ably-ios/builds/325283018#L3028) is failing because of #669.

* Initial Push types * Update project * ARTPushNotifications interface * Realtime: integrate Push type * Rest: integrate Push type * Push.activate * fixup! Push.activate * Push.publish - add PushRecipient protocol * fixup! Push.publish * PushChannel subscribe and unsubscribe * Fix registerCallback * Remove warnings - unused methods * Fix FormFactor * Fix: use device token as Hex string * Fix: should respect the default encoder - useBinaryProtocol * Remove ARTPushRecipient protocol and types * Add ULID pod * Use ULID as DeviceId * Push.deactivate * Load local device * Push: update device * PushChannel.subscriptions * Update ULID pod * PushAdmin * Update ULID pod * Remove warnings * JSON encoder: normalize deviceToken hex string * Fix: Push.publish request body * Update device forms * Activation State Machine: init * Activation State Machine: registration * Removed ablyPushAuthKey, ablyPushAuthToken and ablyPushClientId from push delegate * Add Rest argument on Push delegate * tmp * Fix things for push, and adapt to latest API. * Initial State Machine tests * ARTLocalDeviceStorage * Fix: ARTPushActivationState * Storage should synchronize UserDefaults * Storage Keys accessible on tests * State Machine tests * ActivationStateMachine: custom delegate and other fixes * PushActivationStateMachine * Implement Device storage * Fix headers * Fix tests

New changes has been made.

funkyboy · 2018-02-01T09:32:59Z

@ricardopereira Please let me know if tests pass also on iOS10 and 11.
Working on this now, so hopefully things will be more streamlined soon.

ricardopereira · 2018-02-01T10:18:32Z

Done.

funkyboy · 2018-02-01T11:16:35Z

Source/ARTPushActivationStateMachine.m

@@ -101,7 +101,7 @@ - (void)persist {

 - (void)deviceRegistration:(ARTErrorInfo *)error {
    #ifdef TARGET_OS_IOS
-    ARTLocalDevice *local = _rest.device;
+    ARTLocalDevice *local = _rest.device_nosync;

    if (![[UIApplication sharedApplication].delegate conformsToProtocol:@protocol(ARTPushRegistererDelegate)]) {
        [NSException raise:@"ARTPushRegistererDelegate must be implemented on AppDelegate" format:@""];


Wondering: can this fix or impact on this? #673

@funkyboy If the user is using the push branch, then yes.

That was a fix for #654.

ricardopereira requested a review from tcard February 9, 2017 00:08

ricardopereira self-assigned this Feb 9, 2017

ricardopereira commented Feb 9, 2017

View reviewed changes

ricardopereira force-pushed the push branch from 666cada to bb829d8 Compare February 13, 2017 11:53

tcard suggested changes Feb 13, 2017

View reviewed changes

mattheworiordan reviewed Feb 14, 2017

View reviewed changes

ricardopereira force-pushed the push branch 4 times, most recently from f56247e to 4d2d686 Compare February 20, 2017 15:16

ricardopereira force-pushed the push branch 2 times, most recently from 077f20b to 002c882 Compare February 22, 2017 14:24

ricardopereira commented Feb 22, 2017

View reviewed changes

ricardopereira changed the title ~~WIP: Push notifications~~ Push notifications Mar 8, 2017

tcard previously requested changes Mar 20, 2017

View reviewed changes

tcard reviewed Mar 21, 2017

View reviewed changes

shantamg mentioned this pull request Apr 27, 2017

Use of unresolved identifier 'ARTPush' #595

Closed

tcard force-pushed the push branch from 26d46d8 to 1134f88 Compare May 29, 2017 16:43

tcard force-pushed the push branch from 1134f88 to 9a746ab Compare July 10, 2017 10:14

ricardopereira force-pushed the master branch from 7f92795 to 7497259 Compare July 24, 2017 14:42

tcard force-pushed the push branch from 9a746ab to 2967616 Compare August 23, 2017 18:17

tcard force-pushed the push branch from 2967616 to 05e9809 Compare August 31, 2017 21:51

tcard force-pushed the push branch from 5cd99f6 to 373f936 Compare September 15, 2017 11:25

ricardopereira and others added 6 commits October 2, 2017 11:55

ARTPushActivationState: use init instead of new.

0468ff2

With a static method, the constructed class is always the parent abstract class ARTPushActivationState, instead of the desired child class. Fixes #628.

Log push activation state machine in a single place.

aa9900a

Adapt push to thread-safety and Sentry.

2fa7fdf

Push activation state machine: fix restored states not having a machine.

ba2af3a

Push: fix double call to callback when using custom registerer.

02765c4

tcard force-pushed the push branch from 373f936 to 02765c4 Compare October 11, 2017 06:43

tcard and others added 7 commits October 13, 2017 15:07

Fix ARTRealtime.device getter name.

ce05f08

Fix deadlock in rest.device (#663)

accd5cb

Fix: PushChannel, use rest instance instead of unused var

32125c5

Fix: UIApplication.delegate is an UI API call

39f766f

Merge branch 'master' into push

b2ffb1f

Fix imports

b9c0951

ricardopereira and others added 2 commits February 1, 2018 01:04

Merge branch 'master' into push

1496561

ricardopereira mentioned this pull request Feb 1, 2018

Full test coverage of push functionality before GA release #632

Closed

funkyboy reviewed Feb 1, 2018

View reviewed changes

funkyboy merged commit 7012435 into master Feb 1, 2018


		extern NSString *const ARTDevicePlatform;

		typedef NS_ENUM(NSUInteger, ARTDeviceFormFactor) {


		@interface ARTLocalDevice : ARTDeviceDetails

		@property (nonatomic, readonly) ARTDeviceToken *registrationToken;


		@property (nonatomic, readonly) ARTDeviceToken *registrationToken;

		- (instancetype)init;


		@interface ARTPushChannelSubscription : NSObject

		@property (nonatomic, readonly) NSString *deviceId;

Push notifications #582

Push notifications #582

Conversation

ricardopereira commented Feb 9, 2017 • edited by funkyboy Loading

Installation

Installing through CocoaPods (recommended)

Installing through Carthage

Manual installation

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tcard left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

mattheworiordan left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ricardopereira commented Feb 20, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

ricardopereira commented Mar 6, 2017

tcard left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tcard commented Apr 10, 2017

ricardopereira commented Apr 10, 2017

tcard commented Apr 10, 2017

mattheworiordan commented Apr 11, 2017

ricardopereira commented Apr 25, 2017

tcard commented Apr 25, 2017

ricardopereira commented Apr 25, 2017

ricardopereira commented Jan 17, 2018 • edited Loading

funkyboy commented Feb 1, 2018

ricardopereira commented Feb 9, 2017 •

edited by funkyboy

Loading

ricardopereira commented Jan 17, 2018 •

edited

Loading

ricardopereira commented Feb 1, 2018 •

edited

Loading