feat: TAP Agent #83

aasseman · 2023-10-26T21:07:35Z

Sorry, it's a big one. Perhaps we could set up a call to discuss this PR.

The TAP Agent is a new daemon that indexers will have to run alongside the indexer-service instances. It:

Monitors receipts being saved by the indexer-service instances into the DB (through pg_notify).
Keeps track of the unaggregated fees for each Account, where an Account is (indexer, sender, allocation).
Handles the receipt validation checks (because the indexer-service is designed to do barely any checks).
Handles RAV (Receipt Aggregate Vouchers) requests whenever the total unaggregated fees surpass a trigger amount.
Automatically does one last RAV request when an Account relationship ends (Allocation closed, for example).

What it should, but does not do yet:

Check the receipt values against the Agora system
Signal the indexer-service that a sender should be banned (not enough in Escrow, too many invalid receipts, etc)
A lot of other, smaller, but important things, that are marked as TODO in the code...

I'm not particularly proud of the overall design, and it also shone some light onto design shortcomings in tap_core that we did not foresee while designing it "in a vacuum". That's why there is some weirdness in the TAP adapters implementation such as https://github.com/graphops/indexer-service-rs/blob/20e48b1191a8d39f02e32200ef75fc065470cdd6/tap_agent/src/tap/receipt_checks_adapter.rs#L64-L76.

My goal here is to get something through the door so that we can continue improving this incrementally together. I propose that we try moving as many review issues into actual issues that we tackle later.

github-actions · 2023-10-26T21:46:07Z

Pull Request Test Coverage Report for Build 6698506386

867 of 1235 (70.2%) changed or added relevant lines in 13 files are covered.
No unchanged relevant lines lost coverage.
Overall coverage increased (+8.4%) to 58.656%

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
tap_agent/src/tap/test_utils.rs	66	68	97.06%
tap_agent/src/tap/rav_storage_adapter.rs	39	45	86.67%
tap_agent/src/main.rs	0	15	0.0%
tap_agent/src/tap/escrow_adapter.rs	87	102	85.29%
tap_agent/src/database.rs	0	22	0.0%
tap_agent/src/tap/receipt_storage_adapter.rs	182	211	86.26%
tap_agent/src/tap/receipt_checks_adapter.rs	107	149	71.81%
tap_agent/src/tap/account.rs	222	265	83.77%
tap_agent/src/tap/accounts_manager.rs	127	183	69.4%
tap_agent/src/config.rs	8	69	11.59%

Totals
Change from base Build 6644835579:	8.4%
Covered Lines:	2348
Relevant Lines:	4003

💛 - Coveralls

common/src/tap_manager.rs

hopeyen

Great work! still need to digest the tap details, would you mind sharing the flow diagram somewhere? 🙏

hopeyen · 2023-10-27T10:15:55Z

migrations/20230912220523_tap_receipts.up.sql

should there be a ...tap_receipts.down.sql migration?

Yes, it's there, I just didn't change the version from main: https://github.com/graphops/indexer-service-rs/blob/407c17d3ecf766c8e37a2b7e4b9b1132136c4a33/migrations/20230912220523_tap_receipts.down.sql

Also, I modified the migrations in place instead of creating new ones. I thought at this stage of development we don't need to care about migrating the database. (tell me if that's the wrong thinking though)

hopeyen · 2023-10-27T10:18:12Z

tap_agent/src/agent.rs

+        SubgraphClient::new(
+            config
+                .network_subgraph
+                .network_subgraph_deployment


This seems complicated for making a subgraph client. Maybe we can later on make an easier creation helper?

It's the new API introduced in #81

I agree it looks complicated. Perhaps the DeploymentDetails I added into the API is making it a little verbose. Might be worth cleaning up a little at some point, but doesn't feel urgent.

hopeyen · 2023-10-31T10:21:08Z

tap_agent/src/config.rs

+    )]
+    pub rav_request_timestamp_buffer_ns: u64,
+
+    // TODO: Remove this whenever the the gateway registry is ready


how do we expect indexers to have this file? does senders register or host a discoverable aggregator endpoint somewhere? would you please provide an example file I could run the service with?

It will be in the form of a smart contract. That's the current plan/suggestion I got from @Theodus . But no work started on this yet I believe.

PS: the file is just to be able to get going for testing, basically.

Ah, ok, if you got this from @Theodus, then that answers my earlier question about which contract we're talking about - there's just one. 😄

I think it's fine to keep this file as a temporary approach.

hopeyen · 2023-10-31T10:21:54Z

tap_agent/src/config.rs

+    as Receipt Aggregate Voucher (RAV) requests for an indexer."
+)]
+#[command(author, version, about, long_about = None, arg_required_else_help = true)]
+pub struct Cli {


Some of these are duplicates of indexer-service CLI, would it make sense to refactor the duplicates to indexer-common?

Yup! Not sure there is an elegant way to do that though.

I'd suggest leaving these duplicates in for the moment. One thing I'm doing with the indexer-service framework is to have a common config file structure (TOML) and allow indexer service implementations to add their own structure to that (within the same TOML file).

But indexer-service and indexer-agent and tap-agent are all their own processes and I think it's better to keep their interfaces independent.

That said, from now on I'll advocate using as few CLI arguments as possible and moving everything to a config file. It is more secure (env vars are really bad for passing in private keys, mnemonics, passwords) and also easier to manage in systems like terraform, kubernetes etc.

Actually in k8s, the preferred way of passing secrets is using env vars. Unless you are willing to integrate your software directly with the k8s or your cloud's secrets API.
My favorite config implementation is that of graph-node. It uses TOML, but also does something like envsubst over the file before loading it.

https://www.cloudtruth.com/blog/the-pitfalls-of-using-environment-variables-for-config-and-secrets suggests that environment variables for secrets is insecure.

https://www.cloudtruth.com/blog/the-pitfalls-of-using-environment-variables-for-config-and-secrets suggests that environment variables for secrets is insecure.

I'm not sure this really applies to containers though. This is still one of the recommended ways in k8s for ex: https://kubernetes.io/docs/tasks/inject-data-application/distribute-credentials-secure/

aasseman · 2023-10-31T22:57:58Z

Great work! still need to digest the tap details, would you mind sharing the flow diagram somewhere? 🙏

Working on it!

aasseman · 2023-11-01T22:55:31Z

Sorry for the delay. Here's a simple diagram. Which part would you like more details for?

The green blocks are Eventuals

Jannis

Made it 30% through. Will keep going but wanted to drop my comments so far. 😉

migrations/20230915230734_tap_ravs.up.sql

tap_agent/Cargo.toml

Jannis · 2023-11-02T13:51:22Z

tap_agent/src/agent.rs

+    // TODO: replace with a proper implementation once the gateway registry contract is ready
+    let sender_aggregator_endpoints = aggregator_endpoints::load_aggregator_endpoints(
+        config.tap.sender_aggregator_endpoints_file.clone(),
+    );


Once gateways register somewhere (they will do this for subscriptions — perhaps that's the contract you're referring to here or is there another one for TAP?), this could make for a nice Eventual in the indexer service. Add the TAP agent gRPC API, pipe the gateways eventual into an eventual that polls the "need a RAV?" status for each gateway periodically and whenever the gateways change, pipe that into code that requests the RAVs and stores them back in the DB. Done? 😁

Sounds good!

Saved your comment in the appropriate issue: #84 (comment)

Jannis · 2023-11-02T13:56:40Z

tap_agent/src/config.rs

+    as Receipt Aggregate Voucher (RAV) requests for an indexer."
+)]
+#[command(author, version, about, long_about = None, arg_required_else_help = true)]
+pub struct Cli {


I'd suggest leaving these duplicates in for the moment. One thing I'm doing with the indexer-service framework is to have a common config file structure (TOML) and allow indexer service implementations to add their own structure to that (within the same TOML file).

But indexer-service and indexer-agent and tap-agent are all their own processes and I think it's better to keep their interfaces independent.

Jannis · 2023-11-02T13:59:04Z

tap_agent/src/config.rs

+    as Receipt Aggregate Voucher (RAV) requests for an indexer."
+)]
+#[command(author, version, about, long_about = None, arg_required_else_help = true)]
+pub struct Cli {


That said, from now on I'll advocate using as few CLI arguments as possible and moving everything to a config file. It is more secure (env vars are really bad for passing in private keys, mnemonics, passwords) and also easier to manage in systems like terraform, kubernetes etc.

Jannis · 2023-11-02T14:49:30Z

tap_agent/src/tap/account.rs

+    pub async fn start_last_rav_request(&self) {
+        *(self.inner.state.lock().await) = State::LastRavPending;
+        let mut rav_requester_task = self.rav_requester_task.lock().await;
+        if !Self::rav_requester_task_is_running(&rav_requester_task) {
+            *rav_requester_task = Some(tokio::spawn(Self::rav_requester(self.inner.clone())));
+        }
+    }


Not a reason to rework this PR but it would be nice to avoid the state machine and all the locking happening in Account. But I'd need to think how for a bit, not sure right now.

Yup, it's complex enough that we could have a risk of an unforeseen deadlock.

That was the best I could come up with to make sure that:

Only one RAV request happens at a time

The RAV request is async from all the rest

I can set the "last rav" state asynchronously

The RAV request can clear the state on its own (while living in its own async bubble)

Sorry, I guess I default back to Electrical Engineer mode in situations like this 😁

An mpsc::channel with a size of 1 could perhaps work to ensure there's only one RAV request processed and it's processed somewhere else. Buuuuuut, I'm not sure. I don't think it works, as reading the next RAV request from the receiver is enough to allow the next one in. And you want more coordination to happen there. I'll think about it some more, but this can be improved later if needed.

Jannis · 2023-11-02T14:53:53Z

tap_agent/src/tap/account.rs

+                    AND sender_address = $2
+            ) 
+            SELECT 
+                COUNT(*), 


How many receipts do we think this would at max count over? count(*) and the other aggregations too are pretty slow; I'd like to get an idea of how slow this query could get.

I estimate that for a regular indexer, should be rarely above ~1000s, as there should be only the unaggregated receipts in there.

Also, I believe things could be optimized such that this query would happen only once at TAP Agent start.

Jannis · 2023-11-02T15:02:33Z

tap_agent/src/tap/account.rs

+        // `COUNT(*)` will always return a value, so we don't need to check for `None`.
+        match res.count.unwrap() {


I actually think something like the below might be preferable to querying count(*) only to avoid the check for None:

unaggregated_fees.last_id = res.max.unwrap_or(0).try_into()?; unaggregated_fees.value = res.sum.unwrap_or(0).to_string().parse::<u128>()?;

The actual reason I wanted the count was to run those checks:

indexer-rs/tap_agent/src/tap/account.rs

Lines 263 to 270 in f7447f7

ensure!(

res.max.is_some(),

"MAX(id) is null but the receipt COUNT(*) is not zero"

);

ensure!(

res.sum.is_some(),

"SUM(value) is null, but the receipt COUNT(*) is not zero"

);

That's indeed a bit overzealous, and a check like this should be more than enough:

ensure!( res.sum.is_none() == res.max.is_none(), "Exactly one of SUM(value) and MAX(id) is null. This should not happen." );

Fixed in 8515c03

Nice! (And good point about the checks, I missed that in my oversimplified suggestion.)

Jannis · 2023-11-02T15:05:02Z

tap_agent/src/tap/account.rs

+    }
+
+    async fn rav_requester_try(inner: &Arc<Inner>) -> anyhow::Result<()> {
+        loop {


Related to this: do we make sure that we don't request multiple RAVs for the same receipts in parallel if a new receipt comes in before the RAV comes back and we still have unaggregated fees > threshold?

We do make sure that we don't do RAV requests in parallel for any given (sender,allocation), by checking Self::rav_requester_task_is_running before spawning a RAV request task. Could be made more elegant I think.

Each RAV request must contain the last RAV (unless it is the first RAV request). We basically aggregate the receipts into the latest RAV we got, and we get a new RAV out of that. That's why it is very important the the whole RAV request dance is serial.

The TapManager will only aggregate receipts that have a timestamp more recent than the last RAV (though note that we are OK with multiple receipts having the same timestamp, they also have a nonce). To make sure that there is no confusion with the new receipts coming in while we're requesting a RAV, the TapManager will only aggregate receipts up to timestamps that are more than timestamp_buffer old: https://github.com/semiotic-ai/timeline-aggregation-protocol/blob/656fc15700f6d29299d78f9c790539cbddb68d02/tap_core/src/tap_manager/manager.rs#L179

One issue I can see here is that we are not checking (in indexer-service) that new receipts have a reasonably accurate timestamp (let's say within 5 seconds of our own clock). So as it is right now it is not really safe :/.

Would it be useful to track locking this in better via an issue?

Yup, created and issue: #92

Jannis · 2023-11-02T15:08:55Z

tap_agent/src/tap/account.rs

+            // TODO: Request compression and response decompression. Also a fancy user agent?
+            let client = HttpClientBuilder::default().build(&inner.sender_aggregator_endpoint)?;


Curious: Say you receive 5000 receipts/s from a gateway, each for a value of $0.0001 and your threshold is $100. 5000 receipts/s make for $0.5/s. So you'd hit the threshold after 200s, at which point you'll have 1M receipts to aggregate.

My questions would be:

How long will the SQL query to calculate the sum of the 1M unaggregated receipts take roughly?

How many MBs of payload is this to send to the gateway?

How long will it take the gateway to aggregate 1M receipts?

We set some standards here: https://github.com/semiotic-ai/timeline-aggregation-protocol/blob/main/tap_aggregator/README.md

We recommend a max of 15,000 receipts / RAV request.

So here are the missing bits currently:

In tap_core's Manager, we should be able to specify a maximim number of receipts / RAV request

In TAP agent, we should trigger RAV requests not only a value, but also a number of receipts (< 15,000)

Sounds like it would be good to have an issue for this. 😄

Let's say we had 500 indexers all receiving 5,000 receipts/s from a single gateway (the numbers are higher than we'd see them right now, but one day...), this gateway would have 167 RAVs to generate for 15,000 receipts each, per second. Would it be able to keep up? 🤔

Created issue #96

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

Was `scalar_tap_latest_ravs` everywhere except in the down migration. Opted for the shorter of the 2 everywhere. Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

Instead of underscores. Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

In particular, simplify handling of NULL results from the DB. Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

aasseman · 2023-11-11T00:15:15Z

Rebased onto latest main (at least so that the tests don't fail)

github-actions · 2023-11-11T00:36:42Z

Pull Request Test Coverage Report for Build 6857909753

913 of 1288 (70.89%) changed or added relevant lines in 13 files are covered.
No unchanged relevant lines lost coverage.
Overall coverage increased (+8.6%) to 59.005%

Changes Missing Coverage	Covered Lines	Changed/Added Lines	%
tap-agent/src/tap/test_utils.rs	66	68	97.06%
tap-agent/src/tap/rav_storage_adapter.rs	39	45	86.67%
tap-agent/src/main.rs	0	15	0.0%
tap-agent/src/tap/escrow_adapter.rs	87	102	85.29%
tap-agent/src/database.rs	0	22	0.0%
tap-agent/src/tap/receipt_storage_adapter.rs	182	211	86.26%
tap-agent/src/tap/sender_allocation_relationship.rs	221	262	84.35%
tap-agent/src/tap/receipt_checks_adapter.rs	107	149	71.81%
tap-agent/src/tap/sender_allocation_relationships_manager.rs	134	197	68.02%
tap-agent/src/config.rs	48	113	42.48%

Totals
Change from base Build 6820857884:	8.6%
Covered Lines:	2408
Relevant Lines:	4081

💛 - Coveralls

hopeyen

The overall structure makes sense to me. Would you be able to provide a list (maybe in docs or something) that tracks todo/unimplemented items so we can have a better view of the progress? I think we should also provide a link to the docs describing the lifetime of a receipt/rav

tap-agent/src/agent.rs

tap-agent/src/tap/receipt_checks_adapter.rs

hopeyen · 2023-11-12T13:26:13Z

tap-agent/src/tap/receipt_storage_adapter.rs

+    type AdapterError = AdapterError;
+
+    /// We don't need this method in TAP Agent.
+    async fn store_receipt(&self, _receipt: ReceivedReceipt) -> Result<u64, Self::AdapterError> {


What component calls this function and update_receipt_by_id?

It the TapManager from tap_core: https://github.com/semiotic-ai/timeline-aggregation-protocol/blob/3d247c06380038d0a871a7427ce00e0406909160/tap_core/src/tap_manager/manager.rs#L80-L111.

But we don't need those since the storage of receipts is done only in service, and in a mush simplified way, so it doesn't even leverage the TapManager at all.

Jannis · 2023-11-02T15:43:57Z

tap_agent/src/tap/account.rs

+            let updated_rows = sqlx::query!(
+                r#"
+                    UPDATE scalar_tap_latest_ravs
+                    SET is_last = true


I wonder if we should call this "closing RAV" or the "final RAV" or simply "is_last_for_allocation_and_sender" to be very explicit. Something like that. "last" can have so many meanings: the last received, the last in some other ordering (I first wanted to comment that we can find the last easily and quickly via the timestamp, but then I realized last has a different meaning here), the last for the allocation/sender...

Fixed in 1178c63

Jannis · 2023-11-03T19:27:04Z

tap_agent/src/agent.rs

+        SubgraphClient::new(
+            config
+                .network_subgraph
+                .network_subgraph_deployment


I agree it looks complicated. Perhaps the DeploymentDetails I added into the API is making it a little verbose. Might be worth cleaning up a little at some point, but doesn't feel urgent.

Jannis · 2023-11-03T19:34:08Z

tap_agent/src/config.rs

+    as Receipt Aggregate Voucher (RAV) requests for an indexer."
+)]
+#[command(author, version, about, long_about = None, arg_required_else_help = true)]
+pub struct Cli {


https://www.cloudtruth.com/blog/the-pitfalls-of-using-environment-variables-for-config-and-secrets suggests that environment variables for secrets is insecure.

tap_agent/src/main.rs

Jannis · 2023-11-03T19:41:53Z

tap_agent/src/tap/account.rs

+            unaggregated_fees.last_id = new_receipt_notification.id;
+
+            let mut rav_requester_task = self.rav_requester_task.lock().await;
+            // TODO: consider making the trigger per sender, instead of per (sender, allocation).


Potentially. Now that you talk about it like it's a scheduling algorithm, I'm thinking we should perhaps merge the logic as is and play with different "algorithms" using realistic amounts of queries, receipts and thresholds. But that's more about optimizing RAV requests in the face of millons of receipts than it is about how an indexer would define a threshold. Hm...

Jannis · 2023-11-03T20:06:53Z

tap_agent/src/tap/account.rs

+            // TODO: Request compression and response decompression. Also a fancy user agent?
+            let client = HttpClientBuilder::default().build(&inner.sender_aggregator_endpoint)?;


Let's say we had 500 indexers all receiving 5,000 receipts/s from a single gateway (the numbers are higher than we'd see them right now, but one day...), this gateway would have 167 RAVs to generate for 15,000 receipts each, per second. Would it be able to keep up? 🤔

tap-agent/src/config.rs

tap-agent/src/tap/receipt_checks_adapter.rs

tap-agent/src/tap/escrow_adapter.rs

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

aasseman · 2023-11-14T01:21:44Z

The overall structure makes sense to me. Would you be able to provide a list (maybe in docs or something) that tracks todo/unimplemented items so we can have a better view of the progress? I think we should also provide a link to the docs describing the lifetime of a receipt/rav

For the progress, I created a milestone here: https://github.com/graphprotocol/indexer-rs/milestone/1. I'm going to add all the todos there.

Jannis

This is good to go from my point of view. Thanks for addressing the feedback so quickly!

aasseman · 2023-11-14T20:03:15Z

This is good to go from my point of view. Thanks for addressing the feedback so quickly!

Thanks for taking the time to go through the big PR 🙏 !

aasseman added size:x-large Very large p1 High priority type:feature New or enhanced functionality labels Oct 26, 2023

aasseman requested review from Jannis and hopeyen October 26, 2023 21:07

aasseman self-assigned this Oct 26, 2023

Jannis reviewed Oct 30, 2023

View reviewed changes

common/src/tap_manager.rs Outdated Show resolved Hide resolved

hopeyen reviewed Oct 31, 2023

View reviewed changes

Jannis reviewed Nov 2, 2023

View reviewed changes

aasseman added 14 commits November 10, 2023 16:13

feat: TAP Agent

9e4735a

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

chore: silence one clippy::too_many_arguments

755cad0

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

ci: add postgres/sqlx to clippy job

96d0f5b

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

refactor(tap_agent): trim_start_matches addresses

c66e5e5

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

fix: table name scalar_tap_ravs discrepancy

75c90dc

Was `scalar_tap_latest_ravs` everywhere except in the down migration. Opted for the shorter of the 2 everywhere. Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

refactor: dashes in bin and crate names

33782e9

Instead of underscores. Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

style: fix spaces in macro call

4389f35

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

refactor: rename Account to SenderAllocationRelationship

a53110b

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

refactor: simplify update_unaggregated_fees_static

8630749

In particular, simplify handling of NULL results from the DB. Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

refactor: cleanup config

906c79b

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

docs: importance of receipt's SERIAL id in DB

692bf46

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

feat: add timeout config for RAV request

01336a1

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

fix: state-based receipt notification rejection

4995082

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

refactor: cleanup

446bde2

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

aasseman force-pushed the aasseman/tap-agent branch from c734829 to 446bde2 Compare November 11, 2023 00:14

aasseman requested a review from Theodus November 11, 2023 00:18

hopeyen reviewed Nov 12, 2023

View reviewed changes

Jannis reviewed Nov 13, 2023

View reviewed changes

aasseman mentioned this pull request Nov 13, 2023

[Feat.Req] Check receipt timestamps early #92

Closed

aasseman added 3 commits November 13, 2023 11:15

fix: duration conversions

8e6460f

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

feat: rav trigger config in GRT instead of wei

05cac7c

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

refactor: rename "gateway" to "sender"

4f0ea80

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

This was referenced Nov 13, 2023

[Feat.Req] TAP Agent gRPC API #84

Closed

Make RAV value threshold per sender #93

Closed

refactor: rename is_last col to final in scalar_tap_ravs table

1178c63

Signed-off-by: Alexis Asseman <alexis@semiotic.ai>

Jannis approved these changes Nov 14, 2023

View reviewed changes

aasseman requested a review from hopeyen November 14, 2023 20:02

aasseman merged commit d2fad80 into main Nov 14, 2023
5 checks passed

aasseman deleted the aasseman/tap-agent branch November 14, 2023 20:08

		// `COUNT(*)` will always return a value, so we don't need to check for `None`.
		match res.count.unwrap() {

	ensure!(
	res.max.is_some(),
	"MAX(id) is null but the receipt COUNT(*) is not zero"
	);
	ensure!(
	res.sum.is_some(),
	"SUM(value) is null, but the receipt COUNT(*) is not zero"
	);

		// TODO: Request compression and response decompression. Also a fancy user agent?
		let client = HttpClientBuilder::default().build(&inner.sender_aggregator_endpoint)?;

feat: TAP Agent #83

feat: TAP Agent #83

Conversation

aasseman commented Oct 26, 2023 • edited Loading

github-actions bot commented Oct 26, 2023 • edited Loading

Pull Request Test Coverage Report for Build 6698506386

💛 - Coveralls

hopeyen 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

aasseman commented Oct 31, 2023

aasseman commented Nov 1, 2023 • edited Loading

Jannis 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

aasseman Nov 2, 2023 • edited Loading

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

aasseman Nov 14, 2023 • edited Loading

Choose a reason for hiding this comment

aasseman commented Nov 11, 2023

github-actions bot commented Nov 11, 2023 • edited Loading

Pull Request Test Coverage Report for Build 6857909753

💛 - Coveralls

hopeyen 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

aasseman commented Nov 14, 2023

Jannis left a comment

Choose a reason for hiding this comment

aasseman commented Nov 14, 2023

aasseman commented Oct 26, 2023 •

edited

Loading

github-actions bot commented Oct 26, 2023 •

edited

Loading

aasseman commented Nov 1, 2023 •

edited

Loading

aasseman Nov 2, 2023 •

edited

Loading

aasseman Nov 14, 2023 •

edited

Loading

github-actions bot commented Nov 11, 2023 •

edited

Loading