Tact Assembly

This repository contains an assembler and disassembler implementation for TVM bitcode.

This implementation provides a complete cycle Text -> Internal representation -> Cells -> BoC -> Cells -> Internal representation -> Text, this means that the same text assembly can be obtained from a text assembly, going through all the compilation and decompilation steps.

The internal representation in this implementation is much more convenient for manipulations than the IR in https://github.com/tact-lang/ton-opcode. Thanks to this, we can, for example, build a bitcode optimizer on top of this assembler, which can work with raw bitcode, and not with some high-level language like Tact or FunC.

During compilation, the assembler collects additional mappings that can be used to convert the TVM log into a full trace that will refer to specific instructions in the decompiled version of the contract. The proof of concept can be found in the ./src/debugger folder, it collects instructions, gas and stack at each step.

This implementation is also able to generate a coverage report for the contract by BoC and logs from the sandbox. The proof of concept can be found in the ./src/coverage folder.

Validity

The assembler was tested on 106k contracts from the blockchain where it successfully decompiled and compiled all contracts into equivalent Cells.

Why?

Reasons not to use Fift:

• It is difficult to use on the web, we need a WASM compiler build, which greatly increases the package size
• Fift is not just a plain assembler, it handles INLINECALLDICT in a special way, implicitly splits Cell into parts in cases where the reference or code size limit is reached and so on.
• We do not control its code, it is written in itself, which makes its modification extremely difficult
• Fift implicitly combines several instructions into one, where a specific implementation is selected based on its arguments
• To write custom logic for how to lay out code, we need to write a lot of unreadable Fift code (see the selector hack)

In the future, this assembler will be used as a target for a new backend in the Tact compiler.

Example contract

Hex: b5ee9c7201010101005e0000b8ff0020dd2082014c97ba94308168e8e0a4f260810200d71820d31f018168e8baf2e052d32f01f823bef2e04dd74c59f9010182f0871740fee9829718bd7dfddb66c215758869fdb682780dc6298283dd6134d67ff910f2a3f800ed55

SETCP 0
DUP
IFNOTRET
DUP
PUSHINT_LONG 85143
EQUAL
PUSHCONT_SHORT {
    DROP
    PUSHINT_16 26856
}
IFJMP
INC
THROWIF_SHORT 32
PUSHINT_16 512
LDSLICEX
DUP
LDU 32
SWAP
PUSHINT_16 26856
EQUAL
THROWIFNOT 82
LDU 48
SWAP
NOW
GEQ
THROWIFNOT 77
PLDREFIDX 0
ROTREV
HASHSU
SWAP
PUSHINT_LONG 61103320625414998982152303707971264142854662587410920750381746782697495516799
CHKSIGNU
THROWIFNOT_SHORT 35
ACCEPT
POPCTR c5