ledger-tool: Bubble up error enum instead of eprintln!() and exit()

[steviez]

Problem

The ledger-tool load_and_process_ledger() function performs many sub-operations that can fail. The current error handling of printing text and exiting inline adds extra noise to the code that actually does work. Additionally, all of the separate exit points mean that there might be slightly different variations the common text when an unrecoverable error occurs.

Summary of Changes

• Introduce LoadAndProcessLedgerError enum for load_and_process_ledger() error conditions
• Remove early exits and instead bubble up a Result<_, LoadAndProcessLedgerError>
• Introduce and use wrapper that unwraps or exits load_and_process_ledger()
• Use the wrapper for all previous callers of load_and_process_ledger() to remove boilerplate code

The diff on this one is massive because some match statements were removed and cargo fmt reduced a lot of indentation. All of the changes introduced by cargo fmt are in a single commit, so viewing this PR commit-by-commit OR with whitespace hidden (or both) would be my advice.

All of the struckthrough items will be handled in a separate PR

[codecov]

Codecov Report

Merging #34426 (efbd735) into master (337c233) will increase coverage by 0.0%.
Report is 2 commits behind head on master.
The diff coverage is n/a.

Additional details and impacted files

@@           Coverage Diff           @@
##           master   #34426   +/-   ##
=======================================
  Coverage    81.8%    81.9%           
=======================================
  Files         819      819           
  Lines      221037   221037           
=======================================
+ Hits       181028   181052   +24     
+ Misses      40009    39985   -24     

[brooksprumo]

Lgtm. Love the error changes. Also somewhat surprised by the number of cargo fmt changes.

[brooksprumo]

🤌

[steviez]

Also somewhat surprised by the number of cargo fmt changes.

Yeah, definitely an indicator that we need to break some stuff into sub-functions. For example, the implementation of create-snapshot is >350 lines after load_and_process_ledger() and it was all within this match statement that I have removed

solana/ledger-tool/src/main.rs

Lines 3010 to 3018 in 39f2866

	match load_and_process_ledger(
	arg_matches,
	&genesis_config,
	blockstore.clone(),
	process_options,
	snapshot_archive_path,
	incremental_snapshot_archive_path,
	) {
	Ok((bank_forks, starting_snapshot_hashes)) => {

Similarly, capitalization command looks to have 500 lines after it calls load_and_process_ledger(). These two are most of the bulk.

After typing the above comments about line counts out, I think I'm going to split this into two PR's:

• The first PR will have the first commit that changes eprintln!() & exit!() to bubble up the new error enum
• The second PR will have the other 3 commits
  • The new helper that reduces repeated handling (ie match statements at call site) / using the new helper / running cargo fmt to fix indentation

I don't think we'd backport any of this functionality, but in any case, I think there is a logical split with these two, and it makes the change that actually does a little more (the error enum one) a little less "obfuscated" by being in the same squashed commit as ~1500 lines of cargo fmt

[brooksprumo]

Code looks good to me. Can you update the PR's title before merging, please?