Sync changes from Agave master (Feb 23) by mircea-c · Pull Request #12 · anza-xyz/svm

mircea-c · 2026-02-20T19:12:28Z

Syncs 19 commits from Agave master that touched SVM-owned files, plus maintenance commits.

Cherry-picks

Agave commits cherry-picked into SVM-owned paths (ordered oldest to newest):

#10241 – #10694: 17 commits spanning svm/, program-runtime/, transaction-context/, programs/system/, programs/bpf_loader/, programs/loader-v4/, programs/compute-budget/
#10197: clippy — update programs/system/ and rustfmt.toml to Rust 2024 style
#10748: fix(svm) — remove redundant Box allocation in program loader

SVM maintenance

Bumps agave git dep pin to cc74332e59 (last cherry-picked commit)
Adds [patch] entries for solana-bpf-loader-program and solana-loader-v4-program so that solana-builtins (agave git dep) uses SVM's local copies. Required because #10529 added #[cfg(feature = "metrics")]-gated parameters to program-runtime's deploy API that agave's copies don't handle.
Adds #[allow(deprecated)] in the system program test mockup for Rent::lamports_per_byte_year, deprecated in solana-rent 3.1.0 and renamed to lamports_per_byte in v4.

* Bumps solana-sbpf to v0.14.1 * Uses sbpf_version.stack_frame_gaps() flag for MemoryRegion::new_writable_gapped(). * Rekeys enable_sbpf_v3_deployment_and_execution. * Shifts the tests guest address space up to avoid the difference in access_violation_err between EbpfError::StackAccessViolation and EbpfError::AccessViolation.

…(#10155) sbpf-debugger: control the debug port through the invocation context.

* Create transaction.rs * Update toml files * Do not build everything for sbf * Sort toml file

* Fix function * Add unit test

* SIMD-0464: Vote Account Initialize V2 * do not disable legacy instruction

* refactor: remove reloading bool argument thread Large majority of invocations of program loading have `reloading` set to `false`. The only non-test use where reloading is used is in deployment, where an inlined call to `reload` makes more sense anyhow. Presence of this argument/option was making the code much more difficult to grok for me, so I went ahead and removed the argument in most places. It can be added back on a case-by-case basis as needed in the future, though due to the soundness requirements, I believe calling `reload` directly will always make more sense anyway. * ledger-tool

… transaction context (#10524) * Switch unsafe cell * Use into_inter

These are detected by the `clippy::redundant_clone` lint which I hope to deny by default in near future. ref. anza-xyz/agave#10121

* svm: add metrics feature * svm: mask out `execute_timings` when `metrics` is enabled

* sbpf-debugger: pass debug_metadata to sbpf Take advantage of the recent merge at solana-sbpf (anza-xyz/sbpf#80) where VM creators could pass debug metadata that can later be visible to the debugger client. * Remove index_in_trace from debug_metadata. With (anza-xyz/agave#10557) the order of the instruction trace would change. It's possible to bring it back with a follow-up PR if necessary.

Split changes for reordering instruction trace

* transaction_processor: improve comments * inline `load_program_from_bytes` * reinline variable

Create configure_instruction_at_index

Rename payer

Revert "refactor: non-functional program-runtime changes for clarity (#10510)" This reverts commit ad004b758407ff95b68c33a904862fc01d0c246a.

This reverts commit 5db07ad5ed356a69b7d00c4d7be2c148c2119937. Type annotation added to address type inference failure with some nightly versions of rustc.

…#10694) * Renames serialize_parameters_unaligned => serialize_parameters_for_abiv0. * Renames serialize_parameters_aligned => serialize_parameters_for_abiv1. * Adds more comments.

* Fix and improve tests * Split test * Merge macros * Fix lint

…rams Bumps the agave pin to the last cherry-picked commit. Adds patch entries for solana-bpf-loader-program and solana-loader-v4-program to use SVM's local copies, fixing a signature mismatch caused by the metrics feature (#10529) adding #[cfg]-gated params that agave's copies don't handle.

solana-rent 3.1.0 deprecated lamports_per_byte_year (renamed to lamports_per_byte in v4), exemption_threshold, and burn_percent. Suppress warnings in test-only code until the workspace bumps to solana-rent 4.x.

buffalojoec

Mind adding more SVM commits through anza-xyz/agave#10679?

mircea-c · 2026-02-23T19:45:28Z

@buffalojoec I added all the commits that were merged until today: anza-xyz/agave#10197 and anza-xyz/agave#10748.

As for anza-xyz/agave#10679, programs/vote is from an agave dep. The rev pin is now at cc74332e59 which is downstream of that commit so the changes are included.

buffalojoec

I think we might be missing anza-xyz/agave#10398 from the history.

It's older than this range, but I just happened to spot it in my review.

mircea-c · 2026-02-25T14:35:13Z

Done. Added anza-xyz/agave#10398

buffalojoec · 2026-02-25T17:31:17Z

Done. Added anza-xyz/agave#10398

Do we care about the ordering?

mircea-c · 2026-02-25T17:51:12Z

Do we care about the ordering?

Well, I clearly didn't think about it 😆

I rewrote the commits to be in the right order.

buffalojoec

Sorry @mircea-c but this is a lot to review and I'm not 100% sure the best way to ensure we haven't a) missed any commits from Agave or b) included all changes from each commit that touched all files in SVM. Plus, the PR is interleaved with CI and other patch-fixes, making this a pretty big beast.

Can we break this up?

nagisa · 2026-03-04T11:44:21Z

I don't think human review is the right way to go about this in the first place.

Could we figure out what's the command to verify that the tree here and tree in agave are the same? Perhaps doesn't really matter if the history is fully preserved, although git filter-tree would be able to provide the relevant history on the agave side at least. Then the review would be largely a "run this command and you'll see that both contents and history are the way we expect them to be."

buffalojoec · 2026-03-05T10:55:46Z

I don't think human review is the right way to go about this in the first place.

Could we figure out what's the command to verify that the tree here and tree in agave are the same? Perhaps doesn't really matter if the history is fully preserved, although git filter-tree would be able to provide the relevant history on the agave side at least. Then the review would be largely a "run this command and you'll see that both contents and history are the way we expect them to be."

100% agree. I'll see if I can put a script together that can do this and I'll open a PR.

…e arrays)

buffalojoec · 2026-03-09T19:49:06Z

I put together some tooling to programmatically keep Agave in sync with SVM, as well as proposed a new workflow.
#14

Lichtso and others added 22 commits January 23, 2026 08:57

perf: avoid cloning errored error units in ProgramTiming (#10243)

e391677

clippy: fix or silence rust 1.92 signaled warnings (#10245)

3eec1c6

sbpf-debugger: control the debug port through the invocation context …

d125df8

…(#10155) sbpf-debugger: control the debug port through the invocation context.

Prepare transaction-context to be compiled for SBF/BPF (#10241)

4d226d3

* Create transaction.rs * Update toml files * Do not build everything for sbf * Sort toml file

Fix get_current_instruction_index (#10387)

82aa839

* Fix function * Add unit test

svm-test-harness: add more program helpers (#10398)

47b5e70

SIMD-0464: Vote Account Initialize V2 (#10397)

f7ccb05

* SIMD-0464: Vote Account Initialize V2 * do not disable legacy instruction

clippy(svm): fix sort_by lint (#10531)

4c982c8

Switch from UnsafeCell<Box<[ITEM]>> to Box<[UnsafeCell<ITEM>]> in…

97c93ca

… transaction context (#10524) * Switch unsafe cell * Use into_inter

clippy: clean up redundant clones in runtime/svm (#10570)

ada00ff

These are detected by the `clippy::redundant_clone` lint which I hope to deny by default in near future. ref. anza-xyz/agave#10121

svm: add metrics feature (#10529)

03be4cc

* svm: add metrics feature * svm: mask out `execute_timings` when `metrics` is enabled

ABIv2: Rename variables and add new method (#10637)

62a6503

Split changes for reordering instruction trace

refactor: non-functional program-runtime changes for clarity (#10510)

3b3ea62

* transaction_processor: improve comments * inline `load_program_from_bytes` * reinline variable

ABIv2: Create configure_instruction_at_index (#10647)

e6cffb4

Create configure_instruction_at_index

programs/system: Rename payer to from (#10658)

fd35266

Rename payer

Revert #10510 (#10667)

c25970a

Revert "refactor: non-functional program-runtime changes for clarity (#10510)" This reverts commit ad004b758407ff95b68c33a904862fc01d0c246a.

Revert "Revert #10510 (#10667)" (#10670)

4fa8da7

This reverts commit 5db07ad5ed356a69b7d00c4d7be2c148c2119937. Type annotation added to address type inference failure with some nightly versions of rustc.

Refactor - Rename serialization functions according to ABI versions (…

722b512

…#10694) * Renames serialize_parameters_unaligned => serialize_parameters_for_abiv0. * Renames serialize_parameters_aligned => serialize_parameters_for_abiv1. * Adds more comments.

ABIv2: Improve test coverage (#10668)

6e879e4

* Fix and improve tests * Split test * Merge macros * Fix lint

mircea-c self-assigned this Feb 20, 2026

sashass1315 and others added 2 commits February 23, 2026 14:21

fix(svm): remove redundant Box allocation in program loader (#10748)

9e67d46

mircea-c force-pushed the sync-agave-feb20 branch 3 times, most recently from 5557a1e to e2f09f5 Compare February 23, 2026 15:44

mircea-c and others added 2 commits February 23, 2026 10:55

fix: allow deprecated Rent fields in tests

976bb77

solana-rent 3.1.0 deprecated lamports_per_byte_year (renamed to lamports_per_byte in v4), exemption_threshold, and burn_percent. Suppress warnings in test-only code until the workspace bumps to solana-rent 4.x.

clippy: update workspace to Rust 2024 style (#10197)

71e57cd

mircea-c force-pushed the sync-agave-feb20 branch from 1beb279 to c6563a6 Compare February 23, 2026 16:03

mircea-c requested review from a team and buffalojoec February 23, 2026 16:13

buffalojoec reviewed Feb 23, 2026

View reviewed changes

mircea-c changed the title ~~Sync changes from Agave master (Feb 20)~~ Sync changes from Agave master (Feb 23) Feb 23, 2026

mircea-c force-pushed the sync-agave-feb20 branch from ff942a5 to 7850b17 Compare February 23, 2026 19:37

mircea-c added 2 commits February 23, 2026 14:39

fmt: apply Rust 2024 import sorting

a8ce0df

chore: advance agave rev to cc74332e59

fb2fdd4

mircea-c force-pushed the sync-agave-feb20 branch from 7850b17 to e3fd0ac Compare February 23, 2026 19:39

mircea-c requested a review from buffalojoec February 23, 2026 19:45

buffalojoec reviewed Feb 25, 2026

View reviewed changes

mircea-c requested a review from buffalojoec February 25, 2026 14:35

mircea-c force-pushed the sync-agave-feb20 branch 2 times, most recently from 1def558 to 346606b Compare February 25, 2026 17:50

buffalojoec requested changes Mar 2, 2026

View reviewed changes

mircea-c force-pushed the sync-agave-feb20 branch 2 times, most recently from 2598e5c to fb2fdd4 Compare March 6, 2026 02:38

chore: apply cargo-sort 2.1.0 formatting (collapse single-item featur…

2aaa387

…e arrays)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Sync changes from Agave master (Feb 23)#12

Sync changes from Agave master (Feb 23)#12
mircea-c wants to merge 29 commits intoanza-xyz:masterfrom
mircea-c:sync-agave-feb20

mircea-c commented Feb 20, 2026 •

edited

Loading

Uh oh!

buffalojoec left a comment

Uh oh!

mircea-c commented Feb 23, 2026

Uh oh!

buffalojoec left a comment •

edited

Loading

Uh oh!

mircea-c commented Feb 25, 2026

Uh oh!

buffalojoec commented Feb 25, 2026

Uh oh!

mircea-c commented Feb 25, 2026

Uh oh!

buffalojoec left a comment

Uh oh!

nagisa commented Mar 4, 2026 •

edited

Loading

Uh oh!

buffalojoec commented Mar 5, 2026

Uh oh!

buffalojoec commented Mar 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

9 participants

Conversation

mircea-c commented Feb 20, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Cherry-picks

SVM maintenance

Uh oh!

buffalojoec left a comment

Choose a reason for hiding this comment

Uh oh!

mircea-c commented Feb 23, 2026

Uh oh!

buffalojoec left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mircea-c commented Feb 25, 2026

Uh oh!

buffalojoec commented Feb 25, 2026

Uh oh!

mircea-c commented Feb 25, 2026

Uh oh!

buffalojoec left a comment

Choose a reason for hiding this comment

Uh oh!

nagisa commented Mar 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

buffalojoec commented Mar 5, 2026

Uh oh!

buffalojoec commented Mar 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

9 participants

mircea-c commented Feb 20, 2026 •

edited

Loading

buffalojoec left a comment •

edited

Loading

nagisa commented Mar 4, 2026 •

edited

Loading