Add del_state support #640
Conversation
Signed-off-by: Marcus Brandenburger <[email protected]>
mbrandenburger
added
the
comp/shim
Signed-off-by: Marcus Brandenburger <[email protected]>
Signed-off-by: Marcus Brandenburger <[email protected]>
Signed-off-by: Marcus Brandenburger <[email protected]>
Signed-off-by: Marcus Brandenburger <[email protected]>
|
DO NOT merge until issue in #641 is discovered |
|
Need to check what is the problem here. Will continue after holidays :/ |
Signed-off-by: Marcus Brandenburger <[email protected]>
| Expect(err).ShouldNot(HaveOccurred()) | ||
| Expect(res).Should(Equal([]byte(value))) | ||
|
|
||
| _, err = echoContract.EvaluateTransaction("del_state", key) | ||
| res, err = contract.SubmitTransaction("del_state", key) |
There was a problem hiding this comment.
This fixes the failing integration test.
| ctx->write_set.erase(key); | ||
| ctx->del_set.insert(key); | ||
|
|
||
| ocall_del_state(key, ctx->u_shim_ctx); |
There was a problem hiding this comment.
(this is unrelated to the PR)
Why do we do this ocall? -- we do the same for the put_state.
The ocall for reading is necessary to grab the data. However, the ocalls for putting and deleting keys essentially update the transaction's rwset in Fabric, which we are not using (if I recall correctly). In fact, FPC maintains its own rwset, which is validated and replayed in the second step (of the two-step transaction commit procedure).
There may still be one reason for these ocall: checking for errors, and returning any error to the enclave.
The big problem is that any feedback/error would not be trustworthy. This is likely a showstopper for such error handling.
There was a problem hiding this comment.
very good observation. We should look into that in ideally remove the ocall for put_state and del_state if really not needed. This may positively chaincode execution performance.
This PR adds
del_stateto the FPC shim.Signed-off-by: Marcus Brandenburger [email protected]