chore: attempt to improve OriginCall funcs

[tbruyelle]

AssertOriginCall and IsOriginCall relies on the number of frames to determine if the function was called by a gno transaction (definition taken from examples/gno.land/r/demo/banktest/README.md). Since this number is different in the context of test and filetest, the 2 functions are overrided during test setup, but the override works only for filetest files and not for test files.

This change allows AssertOriginCall and IsOriginCall to rely on message's signers instead of frames. If the message has at least one signer, we consider the function was called by a gno tx. As a result, the functions no longer needs to be overrided in test setup, and they work both for filetest files and test files.

Fix #481

[moul]

Hey @tbruyelle, sorry for the delay; everything looks good, however, to make sure that it's stable and stays stable over time.

It would be nice to add tests checking the consistency of this approach in the three main conditions:

• simulating a TX,
• testing with _test.gno,
• testing with _filetest.gno.

We may add a unit test suite simulating the various Machine contexts so we can add more and more tests for std over time.
It may be done in another PR and by someone else, but I was wondering what do you think.

[tbruyelle]

Hey @moul no problem for the delay. You're right some tests are missing for a _test.gno file (which was the root problem addressed by this change actually!).

For _filetest.gno files, only AssertOriginCall is actually tested (for instance in demo/banktest), but not IsOriginCall, I will add a case for that.

About simulating a TX, could you give more details? Is it some kind of integration test ?

[tbruyelle]

I added some tests for AssertOriginCall and IsOriginCall, but that required some changes in the body of AssertOriginCall.

Indeed, the panic in this function occurred in go code, so it wasn't recoverable by gno code. Switching from panic to machine.Panic fixed that. Not sure if this switch has other consequences, I can just assert that in both scenario, the program is stopped.

[tbruyelle]

After discussions with @moul, I realized my attempt to change OriginCall functions actually broke their initial behavior. So I put this PR in draft until I fix that.

My initial understanding on these functions was wrong. I though they were ensuring that the call comes from a transaction, but it's not only that, they also ensure that this call comes directly from a transaction, and not from an other contract. This is why they were relying on the number of frames to determine that origin call property.

[moul]

Please share your thoughts on PR #683.

[tbruyelle]

Closed in favor of #703