Add get_processed_sibling_instruction syscall

[jackcmay]

Problem

There are scenarios where instructions (top-level or via CPI) would like to know what the last instruction processed was. Some examples are verification of assert instructions or to check for memo instructions. Programs can already check tx level instructions via the Sysvar1nstructions1111111111111111111111111 but there is no way to get information about what inner instructions have been processed.

Summary of Changes

Add a syscall that allows a program to query what sibling instructions have been successfully processed.

/// Returns a sibling instruction from the processed sibling instruction list.
///
/// The processed sibling instruction list is a reverse-ordered list of
/// successfully processed sibling instructions. For example, given the call flow:
///
/// A
/// B -> C -> D
/// B -> E
/// B -> F
///
/// Then B's processed sibling instruction list is: [A]
/// Then F's processed sibling instruction list is: [E, C]

This PR supersedes #22515

Fixes #22437

[Lichtso]

This is close to what I had in mind. Why not include top_level_instruction_trace as the first entry in each inner_instruction_trace? deconstruct() in bank.rs would then need to skip the first entry, but it might be a bit more compact and easier to understand.

(I can also refactor that later myself)

[jackcmay]

Updated, check it out

[codecov]

Codecov Report

Merging #22859 (6bdf7b0) into master (75563f6) will decrease coverage by 0.0%.
The diff coverage is 47.3%.

@@            Coverage Diff            @@
##           master   #22859     +/-   ##
=========================================
- Coverage    81.3%    81.2%   -0.1%     
=========================================
  Files         560      560             
  Lines      152043   152268    +225     
=========================================
+ Hits       123623   123725    +102     
- Misses      28420    28543    +123     

[Lichtso]

I think I missed this the last time I reviewed it, but InstructionAccount should not be exposed. It is only meant to be a constructor for the tests. Otherwise access the fields of the InstructionContext separately, because the encoding is still suspect to heavy changes.

[jackcmay]

You thinking we replace it with a trace-specific instruction struct?

[jackcmay]

I'm exposing it in the instruction trace

[jackcmay]

We could expose an iterator of a trait that exposes the AccountMeta infos

[Lichtso]

It just occurred to me that you can't know what I refer to as I haven't merged that branch yet.
I meant exposing the individual fields in InstructionAccount separately like this:
Lichtso@c34020d

[jackcmay]

Yeah, thats what I thought you meant, but we need a way to discover the account indicies from the InstructiionContext. So instead of get_instruction_accounts() we need a get_instruction_account_indicies_in_transaction()

[jackcmay]

And make InstructionAccount private

[Lichtso]

We already have a get_instruction_account_indicies_in_transaction()
it is called InstructionContext::get_index_in_transaction()

[jackcmay]

Yup, way past that :-)

Pull request has been modified.

[jackcmay]

@CriesofCarrots Try this out and let us know, I'll work on porting it to v1.9

[CriesofCarrots]

@CriesofCarrots Try this out and let us know, I'll work on porting it to v1.9

Awesome, will do!

[CriesofCarrots]

@jackcmay , this is working great. Here's the branch where I have it in use: https://github.com/CriesofCarrots/solana-program-library/tree/finish-memo-extension
Needs patch to local master solana, then cd token/prgoram-2022 && cargo test-bpf realloc

I'll keep my eyes out for your v1.9 backport, and patch to it for testing when it's ready.