Replaces MockInvokeContext by ThisInvokeContext in tests

[Lichtso]

Problem

MockInvokeContext would need its own ABIv2 compatibility and still uses KeyedAccounts.
So going forward it is easier to just route all the tests to use ThisInvokeContext instead, so that ABIv2 and the inter-operation with the old ABI can both be tested in one structure / implementation.

Summary of Changes

• Replaces use sites of MockInvokeContext by ThisInvokeContext.
• Removes MockInvokeContext, MockComputeMeter and MockLogger.
• Moves ThisInvokeContext to the program-runtime crate to avoid a cyclic dependency with the runtime crate in the program tests.

Fixes #

[codecov]

Codecov Report

Merging #20881 (9c29e44) into master (29ad081) will increase coverage by 0.0%.
The diff coverage is 97.4%.

@@           Coverage Diff           @@
##           master   #20881   +/-   ##
=======================================
  Coverage    81.5%    81.5%           
=======================================
  Files         497      498    +1     
  Lines      140320   140238   -82     
=======================================
- Hits       114440   114379   -61     
+ Misses      25880    25859   -21     

[jackcmay]

Still needs to be removed?

[Lichtso]

Yes, I have another branch for a follow-up PR on this one, which is going to move everything from sdk::process_instruction into the program-runtime-crate. And it will remove that clippy flag.

https://github.com/Lichtso/solana/tree/refactor/move_process_instruction_in_program_runtime

[Lichtso]

Notice: The error code changed as it was previously testing:
process_instruction(&loader_id, &[], &[], &keyed_accounts)
which is incorrect and was changed to:
process_instruction(&loader_id, &[0], &[], &keyed_accounts)

This happened because it originally tested with &loader_id instead of &program_key which the other tests used.

[jackcmay]

I think this looks good, I don't see anything blaring but it's very hard to review when the PR is so large and includes so many unrelated changes.

It would have been helpful if the following had been broken out into separate PRs so that it is easier to identify functional changes:

• Wholesale changes from new_rand() to new_unique()
• Switching to new_ref()
• Renames and code changes unrelated to the problem description
• Moving files around and making large changes

[Lichtso]

had been broken out into separate PRs

Already trying, in fact this is already broken up from a much larger refactoring I have locally.
The other parts are:

• Functional changes: Stop caching sysvars, instead load them ahead of time #21108
• Moving the remaining dyn Traits: https://github.com/Lichtso/solana/tree/refactor/move_process_instruction_in_program_runtime
• Still working on removing the three dyn Traits which had a "mock" version but only have a "this" version now.

Wholesale changes from new_rand() to new_unique()

I agree, that should have been its own PR. I simply forgot that I changed it to have deterministic tests in order to compare that before and after are equal.

Switching to new_ref()

ThisInvokeContext::push() requires Rc<RefCell<AccountSharedData>> which MockInvokeContext didn't so it has to happen in this PR.

Moving files around and making large changes

If you mean the moving of ThisInvokeContext into the program-runtime-crate:
I tried that, but while rebasing this PR onto the immediate state of just moving ThisInvokeContext and separating that from the changes to dyn Trait InvokeContext I decided against it because I still had dependency cycles at the time. But yes, it could have been separated after wards.

---

I do certainly realize how much happened in this PR and that it was hard to review. I am very grateful!
I think the coming follow-up PRs will be a lot cleaner because I have a better idea what they entail.