This optimizes out DB access for non-deployed contracts under the assumption
that deployed ones are always loaded via `LoadScriptWithHash` (and if they're
not --- it's a bug anyway with the new hashing model) which actually is a very
popular case (every entry script does that).
When native method calls other contract result should be put
on the stack of current context. With oracles this problem wasn't
noticed because of void return type.
At the moment we should have 3 possible options to check return state
during vm context unloading:
* no check
* ensure the stack is empty
* ensure the stack is not empty
It is necessary to distinguish them because new _deploy method shouldn't
left anything on stack. Example: if we use _deploy method before some
ordinary contract method which returns one value. Without these changes
the contract invocation will fail due to 2 elements on stack left after
invocation (the first `null` element is from _deploy, the second element
is return-value from the ordinary contract method).
1. `Run()` must be able to continue execution after a breakpoint.
2. VM must stop right before the breakpoint, not after.
3. Initial vm state is NONE, not HALT.
When calling external contracts we expect exactly 1 value to be on
stack. For methods returning nothing, `Null` value is pushed, otherwise
it is an error.`
Implement 3 new instructions: TRY,ENDTRY,ENDFINALLY.
1. TRY marks the start of the block where exceptions are catched.
It has 2 arguments which are relative offsets of exception handler
and the end of the whole try/catch construction.
2. ENDTRY denotes either end of try or catch block.
3. ENDFINALLY denotes end of finally block which is executed
irregardless of whether an exception has occured.
Regular CALLs don't update it, only Contract.Call does. Fixes the following
mismatch on preview2 testnet:
file BlockStorage_100000/dump-block-49000.json: block 48644, changes number mismatch: 6 vs 4
1. Slot is a new mechanism for storing variables during execution
which is more convenient than alt.stack. This commit implements
support for slot opcodes in both vm and compiler.
2. Remove old alt.stack opcodes.
3. Do not process globals at the start of every function, but instead
load them single time at main.
In NEO3 SYSCALL opcode has 4-byte ID parameter.
This commit removes support for string-based syscalls and
changes SYSCALL's parameter to be fixed 4-byte value.
Old implementation could view 0x62 byte in
a script as a JMP instruction irregardless of whether it is
a real opcode or a part of a parameter of another instruction.
In this commit instructions are decoded together with parameters
during jump label rewriting.
Creating a new BinReader for every instruction is a bit too much and it adds
about 1% overhead on block import (and actually is quite visible in the VM
profiling statistics). So use a bit more ugly but efficient method.
go vet is not happy about them:
pkg/io/binaryReader.go:92:21: method ReadByte() byte should have signature ReadByte() (byte, error)
pkg/io/binaryWriter.go:75:21: method WriteByte(u8 byte) should have signature WriteByte(byte) error
This seriously improves the serialization/deserialization performance for
several reasons:
* no time spent in `binary` reflection
* no memory allocations being made on every read/write
* uses fast ReadBytes everywhere it's appropriate
It also makes Fixed8 Serializable just for convenience.
This allows easier reuse of opcodes and in some cases allows to eliminate
dependencies on the whole vm package, like in compiler that only needs opcodes
and doesn't care about VM for any other purpose.
And yes, they're opcodes because an instruction is a whole thing with
operands, that's what context.Next() returns.
