refactor: channel handshake version improvements by seantking · Pull Request #1283 · cosmos/ibc-go

seantking · 2022-04-21T14:43:48Z

Description

closes: #722

Before we can merge this PR, please make sure that all the following items have been
checked off. If any of the checklist items are not applicable, please leave them but
write a little note why.

Targeted PR against correct branch (see CONTRIBUTING.md)
Linked to Github issue with discussion and accepted design OR link to spec that describes this work.
Code follows the module structure standards.
Wrote unit and integration tests
Updated relevant documentation (docs/) or specification (x/<module>/spec/)
Added relevant godoc comments.
Added a relevant changelog entry to the Unreleased section in CHANGELOG.md
Re-reviewed Files changed in the Github PR explorer
Review Codecov Report in the comment section below once CI passes

seantking · 2022-04-21T15:27:10Z


 	// Write channel into state
-	k.ChannelKeeper.WriteOpenInitChannel(ctx, msg.PortId, channelID, msg.Channel.Ordering, msg.Channel.ConnectionHops, msg.Channel.Counterparty, msg.Channel.Version)
+	k.ChannelKeeper.WriteOpenInitChannel(ctx, msg.PortId, channelID, msg.Channel.Ordering, msg.Channel.ConnectionHops, msg.Channel.Counterparty, version)


Changing this causes the fee tests to break. Looking into it 👀

Needed: 0178c09

codecov-commenter · 2022-04-26T11:12:20Z

Codecov Report

Merging #1283 (35f71cf) into main (bd08650) will increase coverage by 0.00%.
The diff coverage is 93.87%.

@@           Coverage Diff           @@
##             main    #1283   +/-   ##
=======================================
  Coverage   80.27%   80.28%           
=======================================
  Files         166      166           
  Lines       12023    12044   +21     
=======================================
+ Hits         9652     9670   +18     
- Misses       1916     1918    +2     
- Partials      455      456    +1

Impacted Files	Coverage Δ
modules/core/keeper/msg_server.go	`58.29% <75.00%> (+0.09%)`	⬆️
modules/apps/29-fee/ibc_middleware.go	`93.75% <92.85%> (-1.43%)`	⬇️
...7-interchain-accounts/controller/ibc_middleware.go	`87.03% <100.00%> (+0.49%)`	⬆️
...les/apps/27-interchain-accounts/host/ibc_module.go	`100.00% <100.00%> (ø)`
modules/apps/transfer/ibc_module.go	`63.88% <100.00%> (+0.76%)`	⬆️

seantking · 2022-04-26T11:16:13Z

+		return "", err
+	}
+
+	if strings.TrimSpace(version) == "" {


I can add a test case for this

…anOpenInit

Co-authored-by: colin axnér <25233464+colin-axner@users.noreply.github.com>

seantking · 2022-04-27T08:28:40Z

 ### API Breaking

 * (transfer) [\#1250](https://github.com/cosmos/ibc-go/pull/1250) Deprecate `GetTransferAccount` since the `transfer` module account is never used.
+* (channel) [\#1283](https://github.com/cosmos/ibc-go/pull/1283) `OnChanOpenInit` now returns a version string in line with the latest [spec changes](https://github.com/cosmos/ibc/pull/629).  


Open to improvements on the changelog message.

crodriguezvega

Nice work, @seantking!

I just noticed that in ICA we don't use any types.Version like we have for transfer (ics20-1) or fee. Was there a reason not use one there?

crodriguezvega · 2022-04-27T08:25:30Z

 	// call underlying app's OnChanOpenInit callback with the appVersion
-	return im.app.OnChanOpenInit(ctx, order, connectionHops, portID, channelID,
-		chanCap, counterparty, versionMetadata.AppVersion)
+	return string(versionBytes), nil


Why was the call to underlying app's OnChanOpenInit removed here?

Its moved to above on line 53 so that the returned application version (for example: ics20-1) can be inserted into the AppVersion field of types.Metadata, and re-marshalled to a JSON encoded version string. Otherwise we would return directly from the app callback (writing ics20-1 as the channel version into state), we would expect ics29 to be enabled but subsequent handshake callbacks on would fail as the channel version in state would not be the metadata object:

{ "AppVersion": "ics20-1", "FeeVersion": "ics29-1", }

Does that make sense?

crodriguezvega · 2022-04-27T08:37:12Z

+			"empty version string", func() {
+				channel.Version = ""
+			}, true,
+		},


Can you also add a test case when channel.Version is not empty but it's a string that does not match the expected types.Version?

We have this test case already, no?

oh, sorry. I missed it.

damiannolan

Nice work! LGTM today, my only concern is the ICA auth module handling of the version string and if it should be given the opportunity to modify it, my initial thinking is no.
Other than that just some minor nits

damiannolan · 2022-04-27T08:59:49Z

 ### API Breaking

 * (transfer) [\#1250](https://github.com/cosmos/ibc-go/pull/1250) Deprecate `GetTransferAccount` since the `transfer` module account is never used.
+* (channel) [\#1283](https://github.com/cosmos/ibc-go/pull/1283) `OnChanOpenInit` now returns a version string in line with the latest [spec changes](https://github.com/cosmos/ibc/pull/629).  


Some sugar coating?

Suggested change

* (channel) [\#1283](https://github.com/cosmos/ibc-go/pull/1283) `OnChanOpenInit` now returns a version string in line with the latest [spec changes](https://github.com/cosmos/ibc/pull/629).

* (channel) [\#1283](https://github.com/cosmos/ibc-go/pull/1283) The `OnChanOpenInit` application callback now supports version selection and returns a version string in line with the latest [spec changes](https://github.com/cosmos/ibc/pull/629).

Added the first part but left out the version selection. I found that term confusing personally.

damiannolan · 2022-04-27T09:09:25Z

+	// call underlying app's OnChanOpenInit callback with the passed in version
 	return im.app.OnChanOpenInit(ctx, order, connectionHops, portID, channelID,
 		chanCap, counterparty, version)


This gives an opportunity to the interchain accounts auth module to modify the version string, or perhaps mistakenly return the incorrect string, right?

Is this something we want to support? i.e. providing the auth module the ability to mutate this?
Currently the only expectation from this callback is the claiming of the channel capability.

The code below would complain about appVersion declared but not used...

Suggested change

// call underlying app's OnChanOpenInit callback with the passed in version

return im.app.OnChanOpenInit(ctx, order, connectionHops, portID, channelID,

chanCap, counterparty, version)

// call underlying app's OnChanOpenInit callback with the passed in version

appVersion, err := im.app.OnChanOpenInit(ctx, order, connectionHops, portID, channelID,

chanCap, counterparty, version)

if err != nil {

return "", err

}

return version, nil

Not really following your point with the code example.

What would be the alternative here?

Agreed if ibc-auth is expected not to change the version at all. Then we should ignore it's return value. And if relayer passes in empty string, then just return our default version

Sorry if that was confusing, like @AdityaSripal says, we should ignore the appVersion field from the code example above by discarding it, as we don't want to depend on ica-auth to return the correct string(version) or risk modifying it:

_, err := im.app.OnChanOpenInit(ctx, order, connectionHops, portID, channelID, chanCap, counterparty version) if err != nil { return "", err }

And if relayer passes in empty string, then just return our default version

But for ICA a relayer is never expected to initiate the channel handshake, it should be from some on-chain mechanism, correct?

But for ICA a relayer is never expected to initiate the channel handshake, it should be from some on-chain mechanism, correct?

In our implementation of the controller module, yes. The version string will always be the Metadata struct

Updated this. Good catch 👍

damiannolan · 2022-04-27T09:11:28Z

 			)

 			if tc.expPass {
+				expMetaData := icatypes.NewMetadata(


minor nit: casing

Suggested change

expMetaData := icatypes.NewMetadata(

expMetadata := icatypes.NewMetadata(

damiannolan · 2022-04-27T09:17:12Z

 	// call underlying app's OnChanOpenInit callback with the appVersion
-	return im.app.OnChanOpenInit(ctx, order, connectionHops, portID, channelID,
-		chanCap, counterparty, versionMetadata.AppVersion)
+	return string(versionBytes), nil


Its moved to above on line 53 so that the returned application version (for example: ics20-1) can be inserted into the AppVersion field of types.Metadata, and re-marshalled to a JSON encoded version string. Otherwise we would return directly from the app callback (writing ics20-1 as the channel version into state), we would expect ics29 to be enabled but subsequent handshake callbacks on would fail as the channel version in state would not be the metadata object:

{ "AppVersion": "ics20-1", "FeeVersion": "ics29-1", }

Does that make sense?

damiannolan · 2022-04-27T09:19:54Z

+
 				suite.Require().NoError(err, "unexpected error from version: %s", tc.version)
 			} else {
 				suite.Require().Error(err, "error not returned for version: %s", tc.version)


Could we also assert version is empty here too?

damiannolan · 2022-04-27T09:20:16Z

 				suite.Require().NoError(err)
+				suite.Require().Equal(types.Version, version)
 			} else {
 				suite.Require().Error(err)


ditto on empty version assertion

AdityaSripal

Looks good so far. The default version logic you have in transfer should be extended to the other applications

AdityaSripal · 2022-04-27T10:35:04Z

 ### API Breaking

 * (transfer) [\#1250](https://github.com/cosmos/ibc-go/pull/1250) Deprecate `GetTransferAccount` since the `transfer` module account is never used.
+* (channel) [\#1283](https://github.com/cosmos/ibc-go/pull/1283) `OnChanOpenInit` now returns a version string in line with the latest [spec changes](https://github.com/cosmos/ibc/pull/629).  


I would add that it may return a default version string if relayers pass in empty version

I left this out as I don't think the default is applicable to all modules.

Hence the word may, i think it could still be useful but maybe we want to leave that suggestion elsewhere (e.g. release notes or something)

AdityaSripal · 2022-04-27T10:39:04Z

+		return "", err
+	}
+
+	if strings.TrimSpace(version) == "" {


Shouldn't all of the applications/middleware have a similar code block?

responded below

AdityaSripal · 2022-04-27T10:41:41Z

-) error {
+) (string, error) {
 	var versionMetadata types.Metadata
 	if err := types.ModuleCdc.UnmarshalJSON([]byte(version), &versionMetadata); err != nil {


This will not enable fee by default. If provided string is empty then, get default version from underlying app and wrap with fee version.
If underlying app returns error with empty version then return error as well

Do we want to enable fees by default? Otherwise there's no way to only use the default version of the base application. I'd think middleware applications must be explicitly activated or, the versionMetadata should be specified with an empty fee version

Yes, this was my assumption also. What is the benefit of enabling the fee middleware by default with an empty version passed in?

I do think fees should be enabled by default. My preference is that relayers get the default version of the stack if they don't explicitly specify otherwise. Some middleware will want to be activated by default and some may want to be explicitly activated. For adoption purposes, I think it would be good to make incentivization the default rather than having to explicitly enable it. Because as I think the number of applications scale, most relayers will only be passing in empty strings and use non-empty strings only in very special cases (otherwise they would need to have context on every single port they relay for).

So a relayer that always defaults to what the application prefers will be opening incentivized channels.

Of course, if the relayer wants to explicitly disable fees then they can by passing in a version without the feeVersion. But we should design for a future where relayers are rarely passing in specific strings

Sounds good to me. In favor of @AdityaSripal's approach. It should just be well documented that ics29 is an exceptional middleware application that we believe should be enabled by default, but that other applications (who might be copy/pasting our code) should give good consideration to whether their application should be enabled by default

Thanks for the explanation @AdityaSripal. Makes sense to me!

AdityaSripal · 2022-04-27T10:44:27Z

+	// call underlying app's OnChanOpenInit callback with the passed in version
 	return im.app.OnChanOpenInit(ctx, order, connectionHops, portID, channelID,
 		chanCap, counterparty, version)


Agreed if ibc-auth is expected not to change the version at all. Then we should ignore it's return value. And if relayer passes in empty string, then just return our default version

seantking · 2022-04-27T15:59:43Z

Looks good so far. The default version logic you have in transfer should be extended to the other applications

I left the default only for ics20.

I think ics29 should not be enabled by default as @colin-axner suggests & relayers do not pass a version for ics-27. The controller for ics27 module always passes a metaData struct as a version to the OnChanOpenInit callback, it's not optional.

seantking · 2022-04-27T16:01:48Z

Nice work, @seantking!

I just noticed that in ICA we don't use any types.Version like we have for transfer (ics20-1) or fee. Was there a reason not use one there?

Here :)

damiannolan · 2022-04-28T08:00:33Z

+	// call underlying app's OnChanOpenInit callback with the passed in version
+	if _, err := im.app.OnChanOpenInit(ctx, order, connectionHops, portID, channelID, chanCap, counterparty, version); err != nil {
+		return "", err
 	}


Could be useful to have a small explanation of why this is ignored in the comment.

…re stack

colin-axner

Nice work! LGTM!

AdityaSripal

Some testing and documentation nits :)

Would appreciate having them fixed, but doesn't need to happen in this PR

AdityaSripal · 2022-05-04T11:21:47Z

+				isFeeEnabled := suite.chainA.GetSimApp().IBCFeeKeeper.IsFeeEnabled(suite.chainA.GetContext(), suite.path.EndpointA.ChannelConfig.PortID, suite.path.EndpointA.ChannelID)
+				if isFeeEnabled {


This is a strange way to test here because you're relying on the result of the callback execution to create your expected value. If the callback incorrectly creates a fee version and incorrectly sets fee enabled you won't catch it here. Similarly if the callback incorrectly doesn't set fee version and incorrectly doesn't set fee enabled you won't catch it.

I think we should add a boolean in the test cases that says feeEnabled. You then can add the expected value for each test case and then check if the feeEnabled flag is true and check if the version contains a fee version

Yup good catch. Thanks for the feedback 🤝

AdityaSripal · 2022-05-04T11:23:51Z

+	// If the provided version string is non-empty, OnChanOpenInit should return
+	// the version string if valid or an error if the provided version is invalid.


Do we require that Init returns the exact same version if version is non-empty? I don't believe that is the case. Applications may want to modify the version string even if it is non-empty

I took this directly from the spec: https://github.com/cosmos/ibc/pull/629/files#diff-508b3e7784a300436fbeb6940be22f31c74e958270a36bb851d06a8782ad7e7aR50

Should we update both? Agree with your assessment.

Yes I agree we should change the spec as well

AdityaSripal

Nice work! Thanks @seantking

Fixes cosmos#1283  ## Summary by CodeRabbit Refactor: - Improvement: Enhanced the transaction handling process in the `TxMempool` module. Transactions are now more accurately filtered based on their gas and data size limits, ensuring that only valid transactions are processed. This update improves the overall efficiency and reliability of transaction processing. Tests: - Test: Added a new test, `TestTxMempoolTxLargerThanMaxBytes`, to verify the correct handling of large transactions. This ensures the robustness of the system when dealing with transactions of varying sizes.  --------- Co-authored-by: Ganesha Upadhyaya <gupadhyaya@Ganeshas-MacBook-Pro-2.local>

seantking added 2 commits April 21, 2022 16:27

refactor: returning version from OnChanOpenInit

5b0bc50

refactor: update tests and add version to proto resp

b7088f8

seantking changed the title ~~refactor: channel handshake improvements~~ refactor: channel handshake version improvements Apr 21, 2022

refactor: adding version to msg server resp

6f20ad9

seantking commented Apr 21, 2022

View reviewed changes

colin-axner reviewed Apr 21, 2022

View reviewed changes

Comment thread modules/core/05-port/types/module.go

Comment thread modules/apps/transfer/ibc_module.go Outdated

Comment thread modules/apps/29-fee/ibc_module.go Outdated

seantking added 3 commits April 26, 2022 12:17

refactor: remove unncessary if & update version on Endpoint.Ack

31a2831

fix: ics29 OnChanOpenInit remake versionMetaData before returning

0178c09

chore: update godoc

8f8ce78

seantking marked this pull request as ready for review April 26, 2022 11:00

seantking requested review from AdityaSripal, charleenfei, crodriguezvega, damiannolan and fedekunze as code owners April 26, 2022 11:00

Merge branch 'main' into sean/issue#722-channel-handshake-improvements

f83b472

seantking commented Apr 26, 2022

View reviewed changes

colin-axner reviewed Apr 26, 2022

View reviewed changes

Comment thread modules/apps/27-interchain-accounts/controller/ibc_module.go Outdated

colin-axner reviewed Apr 26, 2022

View reviewed changes

Comment thread modules/apps/27-interchain-accounts/controller/ibc_module_test.go Outdated

Comment thread modules/apps/29-fee/ibc_module.go Outdated

Comment thread modules/apps/transfer/ibc_module_test.go Outdated

seantking and others added 5 commits April 26, 2022 19:21

test: adding check for expected version string

2b0cca4

test: adding test case for passing empty version string to ics20 onCh…

2d7f0b8

…anOpenInit

Update modules/apps/29-fee/ibc_module.go

534e00c

Co-authored-by: colin axnér <25233464+colin-axner@users.noreply.github.com>

chore: comment

8f08760

Merge branch 'main' into sean/issue#722-channel-handshake-improvements

7d6246c

seantking requested a review from colin-axner April 26, 2022 17:29

seantking and others added 2 commits April 27, 2022 10:27

chore: changelog

a03916d

Merge branch 'main' into sean/issue#722-channel-handshake-improvements

0db8c86

seantking commented Apr 27, 2022

View reviewed changes

crodriguezvega reviewed Apr 27, 2022

View reviewed changes

damiannolan reviewed Apr 27, 2022

View reviewed changes

AdityaSripal suggested changes Apr 27, 2022

View reviewed changes

seantking added 2 commits April 27, 2022 17:57

fix: ica now discards auth module version

ce656c4

chore: update changelog

092d3fe

seantking requested review from AdityaSripal, crodriguezvega and damiannolan April 27, 2022 16:02

damiannolan approved these changes Apr 28, 2022

View reviewed changes

adding default version for ics29

d30f887

seantking force-pushed the sean/issue#722-channel-handshake-improvements branch from 2f640f3 to d30f887 Compare April 29, 2022 16:42

colin-axner reviewed May 2, 2022

View reviewed changes

Comment thread modules/apps/transfer/ibc_module_test.go Outdated

seantking and others added 3 commits May 4, 2022 12:31

fix: using transfer module directly rather than calling full middlewa…

1a44c0b

…re stack

fix testing bug

15d7092

Merge branch 'main' into sean/issue#722-channel-handshake-improvements

35f71cf

colin-axner approved these changes May 4, 2022

View reviewed changes

Comment thread CHANGELOG.md

Comment thread modules/apps/27-interchain-accounts/controller/ibc_middleware.go

AdityaSripal reviewed May 4, 2022

View reviewed changes

seantking added 2 commits May 4, 2022 15:54

refactor: test now uses bool for isFeeEnabled rather than direct check

0b85196

docs: updating comment and migration docs

0286c64

seantking requested review from AdityaSripal and colin-axner May 4, 2022 14:07

AdityaSripal approved these changes May 10, 2022

View reviewed changes

Merge branch 'main' into sean/issue#722-channel-handshake-improvements

28d5622

seantking enabled auto-merge (squash) May 10, 2022 15:19

seantking merged commit dcd0681 into main May 10, 2022

seantking deleted the sean/issue#722-channel-handshake-improvements branch May 10, 2022 15:28

crodriguezvega mentioned this pull request May 16, 2022

Bump package to v4 #1359

Closed

crodriguezvega mentioned this pull request Jun 26, 2022

docs: document that version string can be empty in RegisterInterchainAccount #1582

Merged

9 tasks

	* (channel) [\#1283](https://github.com/cosmos/ibc-go/pull/1283) `OnChanOpenInit` now returns a version string in line with the latest [spec changes](https://github.com/cosmos/ibc/pull/629).
	* (channel) [\#1283](https://github.com/cosmos/ibc-go/pull/1283) The `OnChanOpenInit` application callback now supports version selection and returns a version string in line with the latest [spec changes](https://github.com/cosmos/ibc/pull/629).

	expMetaData := icatypes.NewMetadata(
	expMetadata := icatypes.NewMetadata(

		isFeeEnabled := suite.chainA.GetSimApp().IBCFeeKeeper.IsFeeEnabled(suite.chainA.GetContext(), suite.path.EndpointA.ChannelConfig.PortID, suite.path.EndpointA.ChannelID)
		if isFeeEnabled {

		// If the provided version string is non-empty, OnChanOpenInit should return
		// the version string if valid or an error if the provided version is invalid.

Conversation

seantking commented Apr 21, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Description

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

codecov-commenter commented Apr 26, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

crodriguezvega left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

damiannolan left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

seantking Apr 27, 2022 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

AdityaSripal left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

seantking commented Apr 21, 2022 •

edited

Loading

codecov-commenter commented Apr 26, 2022 •

edited

Loading

crodriguezvega left a comment •

edited

Loading

seantking Apr 27, 2022 •

edited

Loading

colin-axner Apr 27, 2022 •

edited

Loading