Solang Test Suite¶
Solang has a few test suites. These are all run on each pull request via github actions.
Solidity parser and semantics tests¶
In the tests directory, there are a lot of tests which call fn parse_and_resolve(). This function parses Solidity, and returns the namespace: all the resolved contracts, types, functions, etc (as much as could be resolved), and all the compiler diagnositics, i.e. compiler warnings and errors. These tests check that the compiler parser and semantic analysis work correctly.
Note that Solidity can import other soldity files using the
import statement. There are further
tests which create a file cache with filenames and and their contents, to ensure that imports
work as expected.
The stage after semantic analysis is codegen. Codegen generates an IR which is a CFG, so it is simply called CFG. The codegen tests ensure that the CFG matches what should be created. These tests are inspired by LLVM lit tests. The tests can found in codegen_testcases.
These tests do the following:
- Look for a comment
// RUN:and then run the compiler with the given arguments and the filename itself
- After that the output is compared against special comments:
// CHECK:means that following output must be present
// BEGIN-CHECK:means check for the following output but scan the output from the beginning
// FAIL:will check that the command will fail (non-zero exit code) with the following output
Mock contract virtual machine¶
For Substrate, ewasm, and Solana there is a mock virtual machine which should work exactly like the real on-chain virtual machine. For ewasm and Substrate, this uses the wasmi crate and for Solana it uses the Solana RBPF crate.
These tests consist of calling a function call fn build_solidity() which compiles the given solidity source code and then returns a VM. This VM can be used to deploy one of the contract, and test various functions like contract storage, accessing builtins such as block height, or creating/calling other contracts. Since the functionality is mocked, the test can do targeted introspection to see if the correct function was called, or walk the heap of the contract working memory to ensure there are no corruptions.