explicit 8-byte alignment #185
Conversation
cavemanloverboy
requested review from
jkbpvsc,
losman0s and
exdx
as code owners
There was a problem hiding this comment.
Does this change make it possible to finally compile marginfi programs on M-series macs? That would be a huge plus and remove the need for having a x86 VM to compile for me.
I'm a big fan of this change but would like @jkbpvsc to make the final approval. This code is very sensitive and if we change it in a way that's not backwards compatible it will be a pain.
|
yes it makes it compatible with mac. i am confident it doesn't affect on-chain representation, as it just makes the alignment and size of types equal between bpf and arm without changing bpf, but someone should definitely write a test that fetches all types from mainnet to be doubly sure. |
|
@cavemanloverboy can you fix linting, it seems its failing |
this failure was on main haha. i'll fix |
|
the linter failed because the bank and oracles are no longer in the vector holding the lut keys. Just wanna confirm this is intentional |
|
I added a couple of fixture-based tests on You can rebase on |
…mrgnlabs/marginfi-v2 into cavey/explicit-8-byte-alignment
…mrgnlabs/marginfi-v2 into cavey/explicit-8-byte-alignment
| #[macro_export] | ||
| macro_rules! assert_struct_align { | ||
| ($struct: ty, $align: expr) => { | ||
| static_assertions::const_assert_eq!(std::mem::align_of::<$struct>(), $align); |
There was a problem hiding this comment.
I noticed previously static_assertions hasn't been updated in 4 years, I wonder if we can make the same checks using const generics or some other mechanism in latest rust. Just a thought, not for this PR.
There was a problem hiding this comment.
what do you want to check exactly?
|
Closes #181 |
…mrgnlabs/marginfi-v2 into cavey/explicit-8-byte-alignment
|
^ last commit, just wanted to commit latest lock |
* explicit 8-byte alignment * fix linter * update lock files * update locks * upgrade fuzzer's rustc * add toolchain for fuzzer * lower bumpalo version * fix ahash * fix ahash * upgrade rust toolchain for fuzz * Update nightly toolchain for fuzzer * Update test.yaml * latest lock
all padding fields are now 8 byte aligned (on bpf, arm, x86), and all program types are now explciitly 8 byte aligned.
improves devex on mac. bpf in-memory (on-chain) representation is unchanged.