* Drop write lock on sysvars
* adds env var for demoting sysvar write lock demotion
* moves demote logic to is_writable
* feature gates sysvar write lock demotion
* adds builtins to write lock demotion
* adds system program id to builtins
* adds Feature111...
* adds an abi-freeze test
* mvines set of builtin program keys
Co-authored-by: Michael Vines <mvines@gmail.com>
* update tests
* adds bpf loader keys
* Add test sysvar
* Plumb demote_sysvar to is_writable
* more plumbing of demote_sysvar_write_locks to is_writable
* patches test_program_bpf_instruction_introspection
* hard codes demote_sysvar_write_locks to false for serialization/encoding methods
* Revert "hard codes demote_sysvar_write_locks to false for serialization/encoding methods"
This reverts commit ae3e2d2e777437bddd753933097a210dcbc1b1fc.
* change the hardcoded ones to demote_sysvar_write_locks=true
* Use data_as_mut_slice
Co-authored-by: behzad nouri <behzadnouri@gmail.com>
Co-authored-by: Michael Vines <mvines@gmail.com>
* Add some code for special local testing
* Add comment to store_account_and_update_capitalization
* Simplify account.rent_epoch handling for sysvar rent
* Introduce *_for_test functions
* Add deprecation messages to existing api
* Add load_largest_accounts bench
* Check lamports before address filter
* Use BinaryHeap, add Accounts test
* Use pubkey reference in the min-heap
Also, flatten code with early returns
Co-authored-by: Greg Fitzgerald <greg@solana.com>
* Move to module-level clippy allowance
* Fix stake_weighted_timestamp math
* Fix genesis_config
* Fix shred_version
* Fix hard_forks
* Fix process_instruction
* Add ArithmeticOverflow ix error
* Fix nonce_keyed_account
* Update BankSlotDelta frozen abi due to new ix error
* Document AppendVec
* Remove the almost-duplicate state in AppendVec
AppendVec was maintaining two offsets, `current_len` and
`append_offset`. Despite the different looking names, the two
values have the same meaning, but were updated at slightly different
times. When appending a batch of accounts, `current_len` updates would
be immediately available to other threads after each append, whereas
`append_offset` would only be updated after its mutex was unlocked.
`append_offset` is redundant. By removing it, we eliminate potential
bugs and no longer need to suppress clippy warnings.
* Remove get_mut() from AppendVec design
Only the offset into the AppendVec memory is thread-safe. The memory
itself is only threadsafe because it is append-only and is otherwise
unprotected. Adding get_mut() would only be safe if the memory was
protected by a ReadWrite lock.