Revert "Split compute budget instructions process from struct itself …

[steviez]

Problem

#33513 (which added c73bebe) was found to be a consensus breaking change. A canary (mce5) running the tip of master diverged from consensus; I debugged and found the offending slot and then bisected.

// Good hash (taken from another node)
[2023-10-20T07:02:29.153219516Z INFO  solana_runtime::bank] bank frozen: 224820287 hash: DwtKodrR4LxqZfbP6WtBrjteWQi9do9cSHPNAHEsnJc3 ...

// Bad hash (taken from mce5)
[2023-10-20T07:07:44.718105955Z INFO  solana_runtime::bank] bank frozen: 224820287 hash: E7vuWs9L184A6CwRABRG9hWJdweq8D5LbbK2iucnwxkk ...

Here is the bad hash produced by c73bebe

$ cat c73bebe9847ecd5a1cbffa96bf03e03a7683232f_verify_224820287.log | grep "bank frozen: 224820287"
[2023-10-20T12:24:15.128837014Z INFO  solana_runtime::bank] bank frozen: 224820287 hash: E7vuWs9L184A6CwRABRG9hWJdweq8D5LbbK2iucnwxkk ...

And here is the correct hash produced by c73bebe's parent, 4e5c545:

$ cat 4e5c545e23fb9a74c2fed5b5ea49edb02f64744e_verify_224820287.log | grep "bank frozen: 224820287"
[2023-10-20T12:19:00.173143740Z INFO  solana_runtime::bank] bank frozen: 224820287 hash: DwtKodrR4LxqZfbP6WtBrjteWQi9do9cSHPNAHEsnJc3 ...

Summary of Changes

Revert problematic commit

[steviez]

I generated debug files with each version (attached at bottom of this comment) to help us narrow down. These are human readable so we can do a basic diff:

$ diff 224820287-DwtKodrR4LxqZfbP6WtBrjteWQi9do9cSHPNAHEsnJc3.json 224820287-E7vuWs9L184A6CwRABRG9hWJdweq8D5LbbK2iucnwxkk.json
2c2
<   "version": "1.17.2 (src:00000000; feat:3079302975, client:SolanaLabs)",
---
>   "version": "1.18.0 (src:00000000; feat:3079302975, client:SolanaLabs)",
5c5
<   "bank_hash": "DwtKodrR4LxqZfbP6WtBrjteWQi9do9cSHPNAHEsnJc3",
---
>   "bank_hash": "E7vuWs9L184A6CwRABRG9hWJdweq8D5LbbK2iucnwxkk",
7c7
<   "accounts_delta_hash": "3mJ9VLXJgLFnE8iPF8W9CSLpfxvdC3NUfi6k1fWhjTnu",
---
>   "accounts_delta_hash": "FkoCdzWY4axpa2ZoWbMnLpR8yWXvs5H6XJG6uTjqQers",
1363c1363
<       "hash": "CgHk8GKuyJ4WJ7VJdAZNwndfPGH8KFSL5pnd3iRs6CLR",
---
>       "hash": "5KqCT4hgPgYnZMfFd5bL2ruRCKX3eneye7TMJDDyDUMu",
1365c1365
<       "lamports": 9270460912,
---
>       "lamports": 9270461608,
13216c13216
<       "hash": "GmNZBkdAL6YEjH3DbbCx8dKbYTQ9vNyoZoc81mbxAkdb",
---
>       "hash": "6wd3UCg6YKrCrJwsfRfAZoTgxrNa5as5ZfN3yEkVuGtS",
13218c13218
<       "lamports": 2425999982629,
---
>       "lamports": 2425999982281,

Looking up those hashes in each file, the two accounts that differ are:

4KVBskF8LWptkBCC1Ldj98Ym7ndHhvVfnwS3AJDwUSzw
FBKFWadXZJahGtFitAsBvbqh5968gLY7dMBBJUoUjeNi

4KV appears in 3 transactions in slot 224820287 and has 696 (2 * 348) "extra" lamports with the to-be-reverted commit

y3zB9LyqasD9ANMVdYxFXXpQsBBPneecTTgkdUGDa8WvnN7anMQBxgM6xNN1QhXeeKNYMuWFTr2DZWAifYEqqop
x1Fi6X3FmYUaqpS4DRsk5kR4RjK6jW5dHyyHRMHNLUCSdwwxmxz8RgfrqyYEZRKc5Kwxbxq2TgPYGWrwygAA6qW
YeLzPiVhYpWMZHkVeq1Q8NdkwgoqZTiCy1f6JcCN8fiNq7shWtiKM772ZKSuHKWKthRNnsabGptbP1pGfEfRZde

whereas FBK appears in one transaction and has 348 lamports "missing" with the to-be-reverted commit

2aPVH3MMz2bBfLUoMc4E87eDNC5tGxTuzTzXBn7Zx6pqWZqKcLrHNNsdJgb2XoJs8Ed6qG7RAWQG6vUUMQbk1kZG

224820287-DwtKodrR4LxqZfbP6WtBrjteWQi9do9cSHPNAHEsnJc3.json
224820287-E7vuWs9L184A6CwRABRG9hWJdweq8D5LbbK2iucnwxkk.json

[codecov]

Codecov Report

Merging #33784 (fb69d37) into master (092c213) will decrease coverage by 0.1%.
Report is 1 commits behind head on master.
The diff coverage is 92.3%.

@@            Coverage Diff            @@
##           master   #33784     +/-   ##
=========================================
- Coverage    81.8%    81.8%   -0.1%     
=========================================
  Files         807      806      -1     
  Lines      218029   218059     +30     
=========================================
+ Hits       178385   178399     +14     
- Misses      39644    39660     +16