fix(unmerged): contract caller as a chain extension origin
#301
Conversation
# This is the 1st commit message: refactor: general # This is the commit message #2: init # This is the commit message #3: begin refactor # This is the commit message #4: refactor: error handling # This is the commit message #5: tests: add error handling tests # This is the commit message #6: WIP # This is the commit message #7: finalise error handling # This is the commit message #8: refactor: easier review
Co-authored-by: Frank Bell <frank@r0gue.io>
chungquantin
added
bug
chungquantin
changed the title
caller as a chain extension origin
chungquantin
requested review from
evilrobot-01,
Daanvdplas and
peterwht
and removed request for
evilrobot-01 and
Daanvdplas
chungquantin
added
ready-for-final-review
ready-for-high-level-review
and removed
ready-for-final-review
labels
There was a problem hiding this comment.
I think this is definitely something that we need to consider. Both serve a different use case. When we use the contract address to make the calls, the contract serves somewhat as a different entity (e.g. a DAO). When we use the caller's address, the contract serves somewhat as a wrapper (e.g. a wrapper to make something standard compliant). Not sure how we can provide this different use case for each api. Creating a different function id sounds a bit like an overkill to differ between calling with the contract's address or the caller to the contract.
Providing the account which to call the pallet from was another idea but then we would have to filter out these bytes before decoding into the dispatch call or runtime read so that is not ideal either.
| @@ -168,14 +171,14 @@ pub trait Ext { | |||
| type AccountId; | |||
|
|
|||
| /// Returns a reference to the account id of the current contract. | |||
| fn address(&self) -> &Self::AccountId; | |||
| fn address(&self) -> Self::AccountId; | |||
There was a problem hiding this comment.
Why are we changing this?
| } | ||
|
|
||
| impl Ext for () { | ||
| type AccountId = (); | ||
|
|
||
| fn address(&self) -> &Self::AccountId { | ||
| &() | ||
| fn address(&self) -> Self::AccountId { |
There was a problem hiding this comment.
Again, why are we changing this?
There was a problem hiding this comment.
This was changed just to test if it works (will be reverted later when we decide the next step with the extension origin). Reason why I don't add the method caller() because it will bring back stuffs that @evilrobot-01 removed before #265 that removed the Config associated type. The caller() method returns OriginFor<Config> so if we want that method, we need to revert the changed Frank made.
There was a problem hiding this comment.
To be clear, I'm not against adding stuff that is needed, but a useful thing to consider is that taking a trait bound on the whole Config trait makes mocking overly complex. An example is needing to add a mock pallet-contracts impl just to do basic things.
Only expose what is required, nothing more. It may be easier, but has consequences downstream for implementors.
| @@ -185,12 +188,16 @@ pub(crate) struct ExternalEnvironment<'a, T: pallet_contracts::chain_extension:: | |||
| impl<'a, E: pallet_contracts::chain_extension::Ext> Ext for ExternalEnvironment<'a, E> { | |||
| type AccountId = AccountIdOf<E::T>; | |||
|
|
|||
| fn address(&self) -> &Self::AccountId { | |||
| self.0.address() | |||
| fn address(&self) -> Self::AccountId { | |||
There was a problem hiding this comment.
Can we not add a function caller instead? This function is to get the address and should not be changed tmo
There was a problem hiding this comment.
Explained here: #301 (comment)
chungquantin
changed the title
caller as a chain extension origincaller as a chain extension origin
chungquantin
force-pushed
the
chungquantin/feat-psp22_example
branch
from
411ff66
to
8c72e2c
Compare
In the extension code, we have set the
contractitself as the default origin for all calls. However, when it comes to methods likeapproveandtransfer_from, there's a potential problem. Typically, the approve function allows a user (likeALICE) to approve another entity (like a contract) to spend their funds. But since we've set the contract as the origin for all calls, transfer_from would only be callable by the contract itself.Shouldn't it be
self.ext().caller()instead so the caller of the extrinsic is the caller of the contract, not the contract by itself. Would love to have some view on this!Proposed changes
Payload sent to extension must have an extra flag to distinguish between which method should be called with
contractas origin orcalleras origin. Maybe the last byte of the encoding scheme?Guideline for the reviewers
transfer_from_works()inpop-api/examples/fungibleshowALICEcan be the dispatch origin fortransfer_frominstead of a hard-coded contract address. We did not cover this in the contract integration test because it we use the dispatch functions frompallet-assetsdirectly to handle these logics. But as a smart contract developer, everything must be possibly handled on the contract side.