Core Benchmarking Infra (#175)

* Implement a solution for the pre-fund account limit

* Update the account pre-funding handling

* Fix the lighthouse node tracing issue

* refactor existing dt infra

* Implement the platform driver

* Wire up the cleaned up driver implementation

* Implement the core benchmarking components

* Remove some debug logging

* Fix issues in the benchmarks driver

* Implement a global concurrency limit on provider requests

* Update the concurrency limit

* Update the concurrency limit

* Cleanups

* Update the lighthouse ports

* Ignore certain tests

* Update the new geth test
This commit is contained in:
Omar
2025-10-05 18:09:01 +03:00
committed by GitHub
parent f9dc362c03
commit 74fdeb4a2e
51 changed files with 4308 additions and 1990 deletions
+375
View File
@@ -0,0 +1,375 @@
//! A wrapper around the compiler which allows for caching of compilation artifacts so that they can
//! be reused between runs.
use std::{
borrow::Cow,
collections::HashMap,
path::{Path, PathBuf},
sync::{Arc, LazyLock},
};
use futures::FutureExt;
use revive_dt_common::{iterators::FilesWithExtensionIterator, types::CompilerIdentifier};
use revive_dt_compiler::{Compiler, CompilerOutput, Mode, SolidityCompiler};
use revive_dt_core::Platform;
use revive_dt_format::metadata::{ContractIdent, ContractInstance, Metadata};
use alloy::{hex::ToHexExt, json_abi::JsonAbi, primitives::Address};
use anyhow::{Context as _, Error, Result};
use revive_dt_report::ExecutionSpecificReporter;
use semver::Version;
use serde::{Deserialize, Serialize};
use tokio::sync::{Mutex, RwLock, Semaphore};
use tracing::{Instrument, debug, debug_span, instrument};
pub struct CachedCompiler<'a> {
/// The cache that stores the compiled contracts.
artifacts_cache: ArtifactsCache,
/// This is a mechanism that the cached compiler uses so that if multiple compilation requests
/// come in for the same contract we never compile all of them and only compile it once and all
/// other tasks that request this same compilation concurrently get the cached version.
cache_key_lock: RwLock<HashMap<CacheKey<'a>, Arc<Mutex<()>>>>,
}
impl<'a> CachedCompiler<'a> {
pub async fn new(path: impl AsRef<Path>, invalidate_cache: bool) -> Result<Self> {
let mut cache = ArtifactsCache::new(path);
if invalidate_cache {
cache = cache
.with_invalidated_cache()
.await
.context("Failed to invalidate compilation cache directory")?;
}
Ok(Self {
artifacts_cache: cache,
cache_key_lock: Default::default(),
})
}
/// Compiles or gets the compilation artifacts from the cache.
#[allow(clippy::too_many_arguments)]
#[instrument(
level = "debug",
skip_all,
fields(
metadata_file_path = %metadata_file_path.display(),
%mode,
platform = %platform.platform_identifier()
),
err
)]
pub async fn compile_contracts(
&self,
metadata: &'a Metadata,
metadata_file_path: &'a Path,
mode: Cow<'a, Mode>,
deployed_libraries: Option<&HashMap<ContractInstance, (ContractIdent, Address, JsonAbi)>>,
compiler: &dyn SolidityCompiler,
platform: &dyn Platform,
reporter: &ExecutionSpecificReporter,
) -> Result<CompilerOutput> {
let cache_key = CacheKey {
compiler_identifier: platform.compiler_identifier(),
compiler_version: compiler.version().clone(),
metadata_file_path,
solc_mode: mode.clone(),
};
let compilation_callback = || {
async move {
compile_contracts(
metadata
.directory()
.context("Failed to get metadata directory while preparing compilation")?,
metadata
.files_to_compile()
.context("Failed to enumerate files to compile from metadata")?,
&mode,
deployed_libraries,
compiler,
reporter,
)
.map(|compilation_result| compilation_result.map(CacheValue::new))
.await
}
.instrument(debug_span!(
"Running compilation for the cache key",
cache_key.compiler_identifier = %cache_key.compiler_identifier,
cache_key.compiler_version = %cache_key.compiler_version,
cache_key.metadata_file_path = %cache_key.metadata_file_path.display(),
cache_key.solc_mode = %cache_key.solc_mode,
))
};
let compiled_contracts = match deployed_libraries {
// If deployed libraries have been specified then we will re-compile the contract as it
// means that linking is required in this case.
Some(_) => {
debug!("Deployed libraries defined, recompilation must take place");
debug!("Cache miss");
compilation_callback()
.await
.context("Compilation callback for deployed libraries failed")?
.compiler_output
}
// If no deployed libraries are specified then we can follow the cached flow and attempt
// to lookup the compilation artifacts in the cache.
None => {
debug!("Deployed libraries undefined, attempting to make use of cache");
// Lock this specific cache key such that we do not get inconsistent state. We want
// that when multiple cases come in asking for the compilation artifacts then they
// don't all trigger a compilation if there's a cache miss. Hence, the lock here.
let read_guard = self.cache_key_lock.read().await;
let mutex = match read_guard.get(&cache_key).cloned() {
Some(value) => {
drop(read_guard);
value
}
None => {
drop(read_guard);
self.cache_key_lock
.write()
.await
.entry(cache_key.clone())
.or_default()
.clone()
}
};
let _guard = mutex.lock().await;
match self.artifacts_cache.get(&cache_key).await {
Some(cache_value) => {
if deployed_libraries.is_some() {
reporter
.report_post_link_contracts_compilation_succeeded_event(
compiler.version().clone(),
compiler.path(),
true,
None,
cache_value.compiler_output.clone(),
)
.expect("Can't happen");
} else {
reporter
.report_pre_link_contracts_compilation_succeeded_event(
compiler.version().clone(),
compiler.path(),
true,
None,
cache_value.compiler_output.clone(),
)
.expect("Can't happen");
}
cache_value.compiler_output
}
None => {
let compiler_output = compilation_callback()
.await
.context("Compilation callback failed (cache miss path)")?
.compiler_output;
self.artifacts_cache
.insert(
&cache_key,
&CacheValue {
compiler_output: compiler_output.clone(),
},
)
.await
.context(
"Failed to write the cached value of the compilation artifacts",
)?;
compiler_output
}
}
}
};
Ok(compiled_contracts)
}
}
async fn compile_contracts(
metadata_directory: impl AsRef<Path>,
mut files_to_compile: impl Iterator<Item = PathBuf>,
mode: &Mode,
deployed_libraries: Option<&HashMap<ContractInstance, (ContractIdent, Address, JsonAbi)>>,
compiler: &dyn SolidityCompiler,
reporter: &ExecutionSpecificReporter,
) -> Result<CompilerOutput> {
// Puts a limit on how many compilations we can perform at any given instance which helps us
// with some of the errors we've been seeing with high concurrency on MacOS (we have not tried
// it on Linux so we don't know if these issues also persist there or not.)
static SPAWN_GATE: LazyLock<Semaphore> = LazyLock::new(|| Semaphore::new(100));
let _permit = SPAWN_GATE.acquire().await?;
let all_sources_in_dir = FilesWithExtensionIterator::new(metadata_directory.as_ref())
.with_allowed_extension("sol")
.with_use_cached_fs(true)
.collect::<Vec<_>>();
let compilation = Compiler::new()
.with_allow_path(metadata_directory)
// Handling the modes
.with_optimization(mode.optimize_setting)
.with_pipeline(mode.pipeline)
// Adding the contract sources to the compiler.
.try_then(|compiler| {
files_to_compile.try_fold(compiler, |compiler, path| compiler.with_source(path))
})?
// Adding the deployed libraries to the compiler.
.then(|compiler| {
deployed_libraries
.iter()
.flat_map(|value| value.iter())
.map(|(instance, (ident, address, abi))| (instance, ident, address, abi))
.flat_map(|(_, ident, address, _)| {
all_sources_in_dir
.iter()
.map(move |path| (ident, address, path))
})
.fold(compiler, |compiler, (ident, address, path)| {
compiler.with_library(path, ident.as_str(), *address)
})
});
let input = compilation.input().clone();
let output = compilation.try_build(compiler).await;
match (output.as_ref(), deployed_libraries.is_some()) {
(Ok(output), true) => {
reporter
.report_post_link_contracts_compilation_succeeded_event(
compiler.version().clone(),
compiler.path(),
false,
input,
output.clone(),
)
.expect("Can't happen");
}
(Ok(output), false) => {
reporter
.report_pre_link_contracts_compilation_succeeded_event(
compiler.version().clone(),
compiler.path(),
false,
input,
output.clone(),
)
.expect("Can't happen");
}
(Err(err), true) => {
reporter
.report_post_link_contracts_compilation_failed_event(
compiler.version().clone(),
compiler.path().to_path_buf(),
input,
format!("{err:#}"),
)
.expect("Can't happen");
}
(Err(err), false) => {
reporter
.report_pre_link_contracts_compilation_failed_event(
compiler.version().clone(),
compiler.path().to_path_buf(),
input,
format!("{err:#}"),
)
.expect("Can't happen");
}
}
output
}
struct ArtifactsCache {
path: PathBuf,
}
impl ArtifactsCache {
pub fn new(path: impl AsRef<Path>) -> Self {
Self {
path: path.as_ref().to_path_buf(),
}
}
#[instrument(level = "debug", skip_all, err)]
pub async fn with_invalidated_cache(self) -> Result<Self> {
cacache::clear(self.path.as_path())
.await
.map_err(Into::<Error>::into)
.with_context(|| format!("Failed to clear cache at {}", self.path.display()))?;
Ok(self)
}
#[instrument(level = "debug", skip_all, err)]
pub async fn insert(&self, key: &CacheKey<'_>, value: &CacheValue) -> Result<()> {
let key = bson::to_vec(key).context("Failed to serialize cache key (bson)")?;
let value = bson::to_vec(value).context("Failed to serialize cache value (bson)")?;
cacache::write(self.path.as_path(), key.encode_hex(), value)
.await
.with_context(|| {
format!("Failed to write cache entry under {}", self.path.display())
})?;
Ok(())
}
pub async fn get(&self, key: &CacheKey<'_>) -> Option<CacheValue> {
let key = bson::to_vec(key).ok()?;
let value = cacache::read(self.path.as_path(), key.encode_hex())
.await
.ok()?;
let value = bson::from_slice::<CacheValue>(&value).ok()?;
Some(value)
}
#[instrument(level = "debug", skip_all, err)]
pub async fn get_or_insert_with(
&self,
key: &CacheKey<'_>,
callback: impl AsyncFnOnce() -> Result<CacheValue>,
) -> Result<CacheValue> {
match self.get(key).await {
Some(value) => {
debug!("Cache hit");
Ok(value)
}
None => {
debug!("Cache miss");
let value = callback().await?;
self.insert(key, &value).await?;
Ok(value)
}
}
}
}
#[derive(Clone, Debug, PartialEq, Eq, Hash, Serialize)]
struct CacheKey<'a> {
/// The identifier of the used compiler.
compiler_identifier: CompilerIdentifier,
/// The version of the compiler that was used to compile the artifacts.
compiler_version: Version,
/// The path of the metadata file that the compilation artifacts are for.
metadata_file_path: &'a Path,
/// The mode that the compilation artifacts where compiled with.
solc_mode: Cow<'a, Mode>,
}
#[derive(Clone, Debug, Serialize, Deserialize)]
struct CacheValue {
/// The compiler output from the compilation run.
compiler_output: CompilerOutput,
}
impl CacheValue {
pub fn new(compiler_output: CompilerOutput) -> Self {
Self { compiler_output }
}
}
+33
View File
@@ -0,0 +1,33 @@
use revive_dt_config::CorpusConfiguration;
use revive_dt_format::{corpus::Corpus, metadata::MetadataFile};
use tracing::{info, info_span, instrument};
/// Given an object that implements [`AsRef<CorpusConfiguration>`], this function finds all of the
/// corpus files and produces a map containing all of the [`MetadataFile`]s discovered.
#[instrument(level = "debug", name = "Collecting Corpora", skip_all)]
pub fn collect_metadata_files(
context: impl AsRef<CorpusConfiguration>,
) -> anyhow::Result<Vec<MetadataFile>> {
let mut metadata_files = Vec::new();
let corpus_configuration = AsRef::<CorpusConfiguration>::as_ref(&context);
for path in &corpus_configuration.paths {
let span = info_span!("Processing corpus file", path = %path.display());
let _guard = span.enter();
let corpus = Corpus::try_from_path(path)?;
info!(
name = corpus.name(),
number_of_contained_paths = corpus.path_count(),
"Deserialized corpus file"
);
metadata_files.extend(corpus.enumerate_tests());
}
// There's a possibility that there are certain paths that all lead to the same metadata files
// and therefore it's important that we sort them and then deduplicate them.
metadata_files.sort_by(|a, b| a.metadata_file_path.cmp(&b.metadata_file_path));
metadata_files.dedup_by(|a, b| a.metadata_file_path == b.metadata_file_path);
Ok(metadata_files)
}
+9
View File
@@ -0,0 +1,9 @@
mod cached_compiler;
mod metadata;
mod pool;
mod test;
pub use cached_compiler::*;
pub use metadata::*;
pub use pool::*;
pub use test::*;
+59
View File
@@ -0,0 +1,59 @@
//! This crate implements concurrent handling of testing node.
use std::sync::atomic::{AtomicUsize, Ordering};
use anyhow::Context as _;
use revive_dt_config::*;
use revive_dt_core::Platform;
use revive_dt_node_interaction::EthereumNode;
/// The node pool starts one or more [Node] which then can be accessed
/// in a round robbin fashion.
pub struct NodePool {
next: AtomicUsize,
nodes: Vec<Box<dyn EthereumNode + Send + Sync>>,
}
impl NodePool {
/// Create a new Pool. This will start as many nodes as there are workers in `config`.
pub async fn new(context: Context, platform: &dyn Platform) -> anyhow::Result<Self> {
let concurrency_configuration = AsRef::<ConcurrencyConfiguration>::as_ref(&context);
let nodes = concurrency_configuration.number_of_nodes;
let mut handles = Vec::with_capacity(nodes);
for _ in 0..nodes {
let context = context.clone();
handles.push(platform.new_node(context)?);
}
let mut nodes = Vec::with_capacity(nodes);
for handle in handles {
nodes.push(
handle
.join()
.map_err(|error| anyhow::anyhow!("failed to spawn node: {:?}", error))
.context("Failed to join node spawn thread")?
.context("Node failed to spawn")?,
);
}
let pre_transactions_tasks = nodes
.iter_mut()
.map(|node| node.pre_transactions())
.collect::<Vec<_>>();
futures::future::try_join_all(pre_transactions_tasks)
.await
.context("Failed to run the pre-transactions task")?;
Ok(Self {
nodes,
next: Default::default(),
})
}
/// Get a handle to the next node.
pub fn round_robbin(&self) -> &dyn EthereumNode {
let current = self.next.fetch_add(1, Ordering::SeqCst) % self.nodes.len();
self.nodes.get(current).unwrap().as_ref()
}
}
+325
View File
@@ -0,0 +1,325 @@
use std::collections::BTreeMap;
use std::sync::Arc;
use std::{borrow::Cow, path::Path};
use futures::{Stream, StreamExt, stream};
use indexmap::{IndexMap, indexmap};
use revive_dt_common::iterators::EitherIter;
use revive_dt_common::types::PlatformIdentifier;
use revive_dt_config::Context;
use revive_dt_format::mode::ParsedMode;
use serde_json::{Value, json};
use revive_dt_compiler::Mode;
use revive_dt_compiler::SolidityCompiler;
use revive_dt_format::{
case::{Case, CaseIdx},
metadata::MetadataFile,
};
use revive_dt_node_interaction::EthereumNode;
use revive_dt_report::{ExecutionSpecificReporter, Reporter};
use revive_dt_report::{TestSpecificReporter, TestSpecifier};
use tracing::{debug, error, info};
use crate::Platform;
use crate::helpers::NodePool;
pub async fn create_test_definitions_stream<'a>(
// This is only required for creating the compiler objects and is not used anywhere else in the
// function.
context: &Context,
metadata_files: impl IntoIterator<Item = &'a MetadataFile>,
platforms_and_nodes: &'a BTreeMap<PlatformIdentifier, (&dyn Platform, NodePool)>,
reporter: Reporter,
) -> impl Stream<Item = TestDefinition<'a>> {
stream::iter(
metadata_files
.into_iter()
// Flatten over the cases.
.flat_map(|metadata_file| {
metadata_file
.cases
.iter()
.enumerate()
.map(move |(case_idx, case)| (metadata_file, case_idx, case))
})
// Flatten over the modes, prefer the case modes over the metadata file modes.
.flat_map(move |(metadata_file, case_idx, case)| {
let reporter = reporter.clone();
let modes = case.modes.as_ref().or(metadata_file.modes.as_ref());
let modes = match modes {
Some(modes) => EitherIter::A(
ParsedMode::many_to_modes(modes.iter()).map(Cow::<'static, _>::Owned),
),
None => EitherIter::B(Mode::all().map(Cow::<'static, _>::Borrowed)),
};
modes.into_iter().map(move |mode| {
(
metadata_file,
case_idx,
case,
mode.clone(),
reporter.test_specific_reporter(Arc::new(TestSpecifier {
solc_mode: mode.as_ref().clone(),
metadata_file_path: metadata_file.metadata_file_path.clone(),
case_idx: CaseIdx::new(case_idx),
})),
)
})
})
// Inform the reporter of each one of the test cases that were discovered which we expect to
// run.
.inspect(|(_, _, _, _, reporter)| {
reporter
.report_test_case_discovery_event()
.expect("Can't fail");
}),
)
// Creating the Test Definition objects from all of the various objects we have and creating
// their required dependencies (e.g., compiler).
.filter_map(
move |(metadata_file, case_idx, case, mode, reporter)| async move {
let mut platforms = BTreeMap::new();
for (platform, node_pool) in platforms_and_nodes.values() {
let node = node_pool.round_robbin();
let compiler = platform
.new_compiler(context.clone(), mode.version.clone().map(Into::into))
.await
.inspect_err(|err| {
error!(
?err,
platform_identifier = %platform.platform_identifier(),
"Failed to instantiate the compiler"
)
})
.ok()?;
reporter
.report_node_assigned_event(
node.id(),
platform.platform_identifier(),
node.connection_string(),
)
.expect("Can't fail");
let reporter =
reporter.execution_specific_reporter(node.id(), platform.platform_identifier());
platforms.insert(
platform.platform_identifier(),
TestPlatformInformation {
platform: *platform,
node,
compiler,
reporter,
},
);
}
Some(TestDefinition {
/* Metadata file information */
metadata: metadata_file,
metadata_file_path: metadata_file.metadata_file_path.as_path(),
/* Mode Information */
mode: mode.clone(),
/* Case Information */
case_idx: CaseIdx::new(case_idx),
case,
/* Platform and Node Assignment Information */
platforms,
/* Reporter */
reporter,
})
},
)
// Filter out the test cases which are incompatible or that can't run in the current setup.
.filter_map(move |test| async move {
match test.check_compatibility() {
Ok(()) => Some(test),
Err((reason, additional_information)) => {
debug!(
metadata_file_path = %test.metadata.metadata_file_path.display(),
case_idx = %test.case_idx,
mode = %test.mode,
reason,
additional_information =
serde_json::to_string(&additional_information).unwrap(),
"Ignoring Test Case"
);
test.reporter
.report_test_ignored_event(
reason.to_string(),
additional_information
.into_iter()
.map(|(k, v)| (k.into(), v))
.collect::<IndexMap<_, _>>(),
)
.expect("Can't fail");
None
}
}
})
.inspect(|test| {
info!(
metadata_file_path = %test.metadata_file_path.display(),
case_idx = %test.case_idx,
mode = %test.mode,
"Created a test case definition"
);
})
}
/// This is a full description of a differential test to run alongside the full metadata file, the
/// specific case to be tested, the platforms that the tests should run on, the specific nodes of
/// these platforms that they should run on, the compilers to use, and everything else needed making
/// it a complete description.
pub struct TestDefinition<'a> {
/* Metadata file information */
pub metadata: &'a MetadataFile,
pub metadata_file_path: &'a Path,
/* Mode Information */
pub mode: Cow<'a, Mode>,
/* Case Information */
pub case_idx: CaseIdx,
pub case: &'a Case,
/* Platform and Node Assignment Information */
pub platforms: BTreeMap<PlatformIdentifier, TestPlatformInformation<'a>>,
/* Reporter */
pub reporter: TestSpecificReporter,
}
impl<'a> TestDefinition<'a> {
/// Checks if this test can be ran with the current configuration.
pub fn check_compatibility(&self) -> TestCheckFunctionResult {
self.check_metadata_file_ignored()?;
self.check_case_file_ignored()?;
self.check_target_compatibility()?;
self.check_evm_version_compatibility()?;
self.check_compiler_compatibility()?;
Ok(())
}
/// Checks if the metadata file is ignored or not.
fn check_metadata_file_ignored(&self) -> TestCheckFunctionResult {
if self.metadata.ignore.is_some_and(|ignore| ignore) {
Err(("Metadata file is ignored.", indexmap! {}))
} else {
Ok(())
}
}
/// Checks if the case file is ignored or not.
fn check_case_file_ignored(&self) -> TestCheckFunctionResult {
if self.case.ignore.is_some_and(|ignore| ignore) {
Err(("Case is ignored.", indexmap! {}))
} else {
Ok(())
}
}
/// Checks if the platforms all support the desired targets in the metadata file.
fn check_target_compatibility(&self) -> TestCheckFunctionResult {
let mut error_map = indexmap! {
"test_desired_targets" => json!(self.metadata.targets.as_ref()),
};
let mut is_allowed = true;
for (_, platform_information) in self.platforms.iter() {
let is_allowed_for_platform = match self.metadata.targets.as_ref() {
None => true,
Some(required_vm_identifiers) => {
required_vm_identifiers.contains(&platform_information.platform.vm_identifier())
}
};
is_allowed &= is_allowed_for_platform;
error_map.insert(
platform_information.platform.platform_identifier().into(),
json!(is_allowed_for_platform),
);
}
if is_allowed {
Ok(())
} else {
Err((
"One of the platforms do do not support the targets allowed by the test.",
error_map,
))
}
}
// Checks for the compatibility of the EVM version with the platforms specified.
fn check_evm_version_compatibility(&self) -> TestCheckFunctionResult {
let Some(evm_version_requirement) = self.metadata.required_evm_version else {
return Ok(());
};
let mut error_map = indexmap! {
"test_desired_evm_version" => json!(self.metadata.required_evm_version),
};
let mut is_allowed = true;
for (_, platform_information) in self.platforms.iter() {
let is_allowed_for_platform =
evm_version_requirement.matches(&platform_information.node.evm_version());
is_allowed &= is_allowed_for_platform;
error_map.insert(
platform_information.platform.platform_identifier().into(),
json!(is_allowed_for_platform),
);
}
if is_allowed {
Ok(())
} else {
Err((
"EVM version is incompatible for the platforms specified",
error_map,
))
}
}
/// Checks if the platforms compilers support the mode that the test is for.
fn check_compiler_compatibility(&self) -> TestCheckFunctionResult {
let mut error_map = indexmap! {
"test_desired_evm_version" => json!(self.metadata.required_evm_version),
};
let mut is_allowed = true;
for (_, platform_information) in self.platforms.iter() {
let is_allowed_for_platform = platform_information
.compiler
.supports_mode(self.mode.optimize_setting, self.mode.pipeline);
is_allowed &= is_allowed_for_platform;
error_map.insert(
platform_information.platform.platform_identifier().into(),
json!(is_allowed_for_platform),
);
}
if is_allowed {
Ok(())
} else {
Err((
"Compilers do not support this mode either for the provided platforms.",
error_map,
))
}
}
}
pub struct TestPlatformInformation<'a> {
pub platform: &'a dyn Platform,
pub node: &'a dyn EthereumNode,
pub compiler: Box<dyn SolidityCompiler>,
pub reporter: ExecutionSpecificReporter,
}
type TestCheckFunctionResult = Result<(), (&'static str, IndexMap<&'static str, Value>)>;