fix: use abi.encode for sender in CCIPLocalSimulatorFork v1.6 messages - #63

Thank you for looking into this. I am not the maintainer, but, based on the repo’s conventions, there are a few things before this commit can be merged.

The release playbook states that fixes are pushed to develop, whilst main receives updates from it.

It makes sense to add a regression test: verify that sender in a v1.6 message are a 32-byte word decoded into address. CCIPLocalSimulatorForkRouting.t.sol shows harness/mock templates.
Also, worth adding lastSender to MinimalStringReceiver and the check in test_routerDirectSend_routesWhenMultipleOffRampsShareSourceSelector. However, Sepolia->Arb Sepolia seems to be on CCIP 2.0 already. Maybe pin the block until for now? The simulator does not support OnRamp 2, I suppose it will be updated later by the maintainers.
Besides, the tests mentioned in the PR description are unaffected by this change as they do not go through _executePostV1dot6.

And add an entry to the ### Fixed section of the changelog (given the format for 0.2.9). The version bump is likely handled by the maintainers.

Hello @Yurii3721 , thanks for raising the issue, using Chainlink Local and apologies for inconvenience. @Nalon is owning this fix and we will cut the new release as soon as possible.

@kurogami-ibrahim79 thank you for opening this PR and willingness to contribute to Chainlink Local. @Yurii3721 is right, we can not merge this to the main branch directly, there is a process of merging to develop, cutting a beta release, testing and then finally merging to main and cutting a stable release. More details here and here.

@Nalon can you please assist with this PR and make sure this is very well tested? Thanks in advance!

@andrejraki Thanks for the explanation regarding the release process. I’ve changed the PR target to the develop branch and will follow the guidelines for the release from beta to stable. Thanks for the guidance. Sorry for the delayed reply.

Thanks for the review, @Yuri3721. I will add a regression test to CCIPLocalSimulatorForkRouting.t.sol to ensure the sender is correctly decoded from a 32-byte word into an address, and then add an entry to the ### Fixed section of CHANGELOG.md for version 0.2.9.

Done. Thanks for the detailed review, @Yurii3721.

Regression test: added test_executePostV1dot6_senderIs32ByteABIWord in CCIPLocalSimulatorForkRouting.t.sol. It drives _executePostV1dot6 with a mock OffRamp (MockOffRampCapture) and asserts the captured sender is a 32-byte ABI word. Before the fix, the test fails; after the fix, it passes.

I also added MockOffRampCapture (recording mock) alongside the existing OffRamp mocks, following the harness/mock pattern in that file. Note: I kept the assertion at the mock boundary (checking message.sender.length == 32) instead of going through MinimalStringReceiver, because the mock OffRamp intercepts before the router dispatch. This avoids the fork dependency entirely and runs cleanly in CI without RPC env vars.

CHANGELOG: added an entry under ### Fixed in the Unreleased section (no version bump, left to maintainers).

Target branch: retargeted to develop.

On the fork test / Sepolia → Arb Sepolia CCIP 2.0 note: you were right, the E2E fork test hit UnsupportedDestinationChain for the Arb Sepolia selector. I've removed that fork test entirely in favor of the mock-based regression test above, which is stable across chain upgrades.

All tests pass locally:

• CCIPLocalSimulatorUnitTest: 8/8
• CCIPLocalSimulatorForkRoutingTest: 8/8 (including the new regression test)

Ready for re-review when you have time. Thanks!