Bumps solana_rbpf to v0.2.14 #18869
Bumps solana_rbpf to v0.2.14 #18869
Conversation
Lichtso
force-pushed
the
rbpf/v0.2.14
branch
3 times, most recently
from
4075902 to
a6723c4
Compare
Codecov Report
@@ Coverage Diff @@
## master #18869 +/- ##
=========================================
- Coverage 82.8% 82.7% -0.1%
=========================================
Files 449 449
Lines 127997 128047 +50
=========================================
- Hits 105993 105983 -10
- Misses 22004 22064 +60 |
|
@jackcmay Can you see if the feature gate looks about right? If so you can merge. |
|
Sounds good, reviewing and will commit if it looks good, thanks! |
| @@ -83,6 +82,8 @@ pub fn create_executor( | |||
| max_call_depth: compute_budget.max_call_depth, | |||
| stack_frame_size: compute_budget.stack_frame_size, | |||
| enable_instruction_tracing: log_enabled!(Trace), | |||
| verify_mul64_imm_nonzero: !invoke_context | |||
| .is_feature_active(&stop_verify_mul64_imm_nonzero::id()), // TODO: Feature gate and then remove me | |||
| ..Config::default() | |||
There was a problem hiding this comment.
Are the stack frame gaps disabled by default now? If so they should be featurized
There was a problem hiding this comment.
Is the Config's instruction_meter_checkpoint_distance default or the compute meter handling change in v0.2.14?? If so should be feature gated
There was a problem hiding this comment.
Still enabled by default: enable_stack_frame_gaps: true
https://github.com/solana-labs/rbpf/blob/302de4d10888dfb76534693308f4ef8060b86fcd/src/vm.rs#L206
Should we disable it in the next release?
There was a problem hiding this comment.
instruction_meter_checkpoint_distance should have no influence on what the program sees, the point where programs are aborted and what error is thrown has to be the same for interpreter and JIT anyway. The option only makes the detection of that condition faster (so it affects the performance only).
There was a problem hiding this comment.
Sounds good as long as there isn't an edge case that's falls in the middle. Also, maybe check with Tao about cost model stuff to make sure this change doesn't result in a behavioral change there.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in #17520.
This reverts commit 9d66458.
This reverts commit 9d66458.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in solana-labs#17520.
* Feature gate for verify_mul64_imm_nonzero as discussed in #17520.
Problem
Summary of Changes
See RBPF release
Fixes #17520
Fixes #18683